Posts by Staff

Using Quantitative Medicine to control Type 2 Diabetes

Posted by on Mar 9, 2017 in News | 0 comments

Title: Using Quantitative Medicine to control Type 2 Diabetes
By: eclaireMD Foundation
Date: 10/20/2016 – 3/8/2017

Table of Contents

Abstract
Section 1: Introduction (2 figures)
Section 2: Methods (3 figures)
Section 3: Weight and Diabetes (8 figures)
Section 4: Glucose (8 figures)
Section 5: Glucose and Food & Exercise (17 figures)
Section 6: Glucose and Others (8 figures)
Section 7: Hyperlipidemia and Hypertension (10 figures)
Section 8: Metabolism Index (MI) and General Health
Status Unit (GHSU) (12 figures)
Section 9: Conclusions (0 figure)
Section 10: Acknowledgement (0 figure)

Total 86 pages and 68 figures

Abstract of paper:
“Using Quantitative Medicine (a branch of Translational Medicine) to Control Type 2 Diabetes”
By: eclaireMD Foundation
Date: November 28, 2016

BACKGROUND
The author has had long-term chronic diseases and suffered from type 2 diabetes, hyperlipidemia, and hypertension for a period of 20 years. His primary health data in 2010 are listed as follows:
Weight: 205 lbs.
Waistline: 44 inches
One-time snap check of PPG: 350 mg/dL
90-days of averaged glucose: 280 mg/dL
A1C: 10.0%
ACR: 116 mg/mmol
Triglycerides: 1161 mg/dL

AIMS
The author spent 7 years (2010-2016) conducting research to find an effective way to control his severe diabetic condition. As a result, he developed three mathematical models and various tools to control his chronic diseases, with an emphasis on type 2 diabetes. During this process, he collected approximately one million “clean” data regarding his health condition. He applied multiple disciplines, including advanced mathematics, computer science (e.g. database, big data analytic, cloud computing, and mobile technology), nonlinear and dynamic digitized engineering modeling, and artificial intelligence or “AI” (automation and machine learning) to simulate the human organic metabolic system.

METHOD
The author created two prediction models of weight and glucose values in order to provide patients with an early warning to alter their lifestyle. The glucose prediction model includes factors such as diabetes medication consumed, quality of food and meal, type of dining location (home or restaurant), exercise amount, stress and trauma, current residence location and weather condition, traveling category and frequency, decreasing internal organ function, and time delay impact on glucose measurement. The weight prediction model includes factors such as quantity of food and meal, type of exercise, change in calories, sleep impact, water consumption, and other prominent factors.
The most significant achievement is that he successfully eliminated the use of all his diabetic medications within the past 2 years (2015-2016), while maintaining his A1C level within the range of 6.2% to 6.6%. In addition, his hypertension and hyperlipidemia came under control. His recent summarized health data is listed below:
Weight: 172 lbs.
Waistline: 32 inches
Averaged 90-days glucose: 115 mg/dL
A1C: 6.3 – 6.5% (without any medication)
ACR: 12.6 mg/mmol

RESULTS AND DISCUSSION
His entire research and development efforts have been based on lifestyle management as part of a preventive medicine, by collecting, processing, and analysis of quantitative medical and health data. The results are displayed in more than 70 figures and diagrams which have hundreds to thousands of data within each illustration. This paper has indicated many conclusive correlations among 11 categories. These categories include four health outputs (weight and waistline, glucose, blood pressure, lipid), six health inputs (food and meal, exercise, stress, sleep, water drinking, life pattern regularity), and time effect. All of these 11 categories are composed of approximately 500 elements. All of them are carefully monitored by special-designed computer software via smartphone or personal computer. About 95% of these elements are managed by AI in which 20 to 25 elements are required by the patient’s daily input. The metabolism indices for the past 5 years are highly consistent with the author’s health state found from various lab test results. The accuracy of weight prediction and glucose prediction has reached 99.9% and 99.0%, respectively.

The author’s findings of correlations and conclusions are highly consistent with the commonly available understandings within the medical community. There is no personal prejudice inside this study as it is based on experimental facts. Therefore, the author hopes that the same existing medical conclusions can be further backed up and proven by a scientific big data and analytics approach.
The phase 2 of this project will include the following:
(1) Collecting and presenting data from mass population of worldwide patients with type 2 diabetes and;
(2) Researching and developing more effective ways to influence or alter patients’ health behavior to adapt a better lifestyle management.

Title: Using Quantitative Medicine to control Type 2 Diabetes
By: eclaireMD Foundation
Date: 10/20/2016 – 3/8/2017

Section 1: Introduction
The 7-year research data contained in this paper are based on the work conducted by a long-term diabetic patient, who is also the paper’s author. From here on, I will use the first person to describe my conditions.

I am an entrepreneur and, starting in 1995, was the CEO of a high-tech publicly traded company. I had lived a continuous high-stress life due to the nature of my work. During a routine blood test in 1998, my glucose level was measured at 350 mg/dL. Between 1998 and 2000, I suffered several bouts of severe low blood sugar known as insulin shock. Attached below is a table which summarizes my health examination data from 2000 to 2010. Note that my highest triglycerides level was 1,161 mg/dL, A1C 10.0%, and ACR 116 mg/mmol. On August 3, 2010, my physician advised that I needed to begin using insulin immediately and would most likely end up on dialysis within 3 years. In the following 3 months, I moved to a new city with fewer stressors, shut down all of my business enterprises, and started the journey to save my life. After more than 6 years of doing research on six types of chronic diseases, I have completely changed my lifestyle and have my diseases under control. During this time, I lost 26 lbs. (going from 198 lbs. to 172 lbs.), and reduced my waistline by 12 inches (going from 44 inches to 32 inches). On September 1, 2016, my health examination data indicated the following much improved results: ACR 12.6 mg/mmol, A1C 6.6% (WITHOUT taking any diabetic medications for over one year), triglycerides level 67 mg/dL, HDL 48 mg/dL, LDL 103 mg/dL, total cholesterol 156 mg/dL, and BMI 25.

Figure 1.1

 

 

 

 

 

 

Figure 1-1:

Comparison of health data

My overall diabetes conditions during the past 15 years can be seen in Figure 1-2: A1C over 15 years (2000-2016). Prior to 2012, my lab tested A1C values were out of control, “all over the map”. However, after 2012, my lab tested A1C values were around 6.6 and they have been enveloped and confined by my mathematical simulated A1C curve. It should also be noted that, during this same period, I have gradually and then completely eliminated all kinds of diabetes medications.

Figure 1.2

 

 

 

 

 

 

 

Figure 1-2:

A1C over 15 years (2000-2016)

 

Date:   10/21/2016 17:00

Section 2:   Methods

During the late part of 2010, I came to realize that I had been totally ignorant in the area of chronic diseases, even though I was very well educated in other areas (having studied 7 different fields at various colleges over 17 years).  Therefore, I decided to dedicate my efforts on acquiring the needed “knowledge” to control and improve my health conditions. During the initial 2-year period (2010-2011), I studied internal medicine, with a special interest in 6 chronic diseases which were diabetes, hypertension, hyperlipidemia, heart disease, stroke, and obesity.  During the next 2 years (2012-2013), I focused on food science and nutrition.

After finishing those 4 years of self-studying and preparation, I was ready to fully address my health problems.  I thought about starting with traditional medical research methodology, i.e. basic research starting from the “cell” level.  However, I did not have sufficient financial resources and professional knowledge to go down that route.  I also believed the most difficult hurdle was my age and severity of diseases.  I might not have enough time to go that route. Therefore, I took stock of my strengths, which were mathematics, computer science, and various engineering disciplines.  I have never received any formal training in the biomedical area; therefore, I used a nonlinear dynamic engineering model and Finite Element concept of structural engineering for inorganic materials to simulate the human body’s organic metabolism system.  And then, I applied advanced mathematics to develop this model’s governing equation.  I defined 4 inter-connected body output categories of weight, glucose, lipid, blood pressure; and 6 inter-connected body input categories of food, exercise, stress, sleep, water hydration, and life pattern or regularity for longevity.  These 10 categories contain several hundred detailed elements.  For example, just “Stress” category contains 33 different stressors for both “normal” person and “abnormal” person (e.g. people who suffer personality disorders).  I also included “Time” as my 11th category since human body conditions evolve over time, i.e. “Dynamic”.  The human body’s organic characteristics must be dealt with using Artificial Intelligence (AI), through trial-and-error and other techniques.  In the modeling process, I excluded all environmental factors such as pollution, radiation, toxic chemicals, poison, hormonal therapy, viral infection, and others due to their complexity and the difficulty of data collection.  These factors are important for cancer (cancer is also one kind of chronic diseases).  It should be noted that my research is focused on preventive medicine, therefore, drugs for treatment medicine is only included as a part of the glucose prediction tool.

Given the average 3-4 month lifespan of blood cells, which carry glucose and lipids throughout the human body, I defined the data collected during the first 3-month period as the initial conditions for solving these mathematical governing equations.  Therefore, it is important for any patient to use my tool to collect his/her initial 3-months data as complete as possible.  After applying these initial conditions on the mathematical system, I can then “solve the equation” (in a mathematical sense), and afterward, the system starts to learn by itself through AI.  In 2014 and 2015, I began building this “organic” bio-medical math model and two practical prediction models for both body weight (after one night of sleep) and glucose levels (2 hours after a meal) by using AI.  On January 1, 2012, I started collecting my own body health data and used that data to continuously test and improve my mathematical system. To date, I have collected and processed near 1 million “clean” data on myself.  Without including AI in system capabilities, the human mind would not be able to deal with such a large and complicated database.  As shown in Figure 2-1 and 2-2, I have reached a 99.9% accuracy on body weight prediction (4/11/2015-10/21/2016) and a 99.0% accuracy on glucose prediction (6/1/2015-10/21/2016).

Using my iPhone APP “Tool”, I can easily manage the massive health data over the cloud, predict my vital signs such as weight and glucose, and be able to monitor my overall health status via the Metabolism Index (MI) and General Health Status Unit (GHSU), which is defined as a three-month running average of the MI value.  From the chart, by mid-2014 (both MI and GHSU had dropped below the dividing health level of 73.5% for my case), it was clear that my overall health conditions had improved significantly (as of now, my MI and GHSU are at around 57%) through a better lifestyle management.

Figure 2.1

 

 

 

 

 

 

 

Figure 2-1:

Predicted and actual body weight (4/11/2015 – 10/21/2016)

Figure 2.2

 

 

 

 

 

 

 

Figure 2-2:

Predicted and actual daily averaged glucose (6/1/2015 – 10/21/2016)

Figure 2.3

 

 

 

 

 

 

 

Figure 2-3:

Metabolism Index (MI) and
General Health Status Unit (GHSU)

 

Date:   10/22/2016 – 10/23/2016

Section 3:   Weight & Diabetes

It is well known how difficult it is to reduce body weight and maintain for an extended period of time.  In 2000, my average weight was 198 lbs., in 2010 it was 194 lbs., and in 2016 it was 172 lbs.  From 2013 to 2014, my weight fluctuated around 180 lbs. due to my busy travel schedule and not managing my lifestyle and health well.  After developing the metabolism model in 2014, I started to use this tool to manage my overall lifestyle along with reducing the number of long-distance flying trips.  Combined with my newly developed weight prediction model on April 11, 2015, I have achieved significant weight reduction.  See Figure 3-1 of weight reduction.

Figure 3.1

 

 

 

 

 

 

 

Figure 3-1

Weight (2012-2016)

Reducing my waistline was a much tougher problem to address than weight reduction. From 2000 to 2014, my waistline ranged from 42 to 44 inches. Only until mid-2015, when I started to watch my overall metabolism and use the weight prediction tool, I finally achieved my waistline reduction goal of 32 inches. See Figure 3-2 for waistline reduction.

Figure 3.2

 

 

 

 

 

 

 

Figure 3-2

Waistline (2012-2016)

There are many factors that affect body weight; however, diet and exercise are the two primary components that people can control. While types and quality of food have a strong correlation with types of chronic diseases (to be discussed in future sections), the quantity of food has a strong correlation with weight as indicated in Figures 3-3 & 3-4. For example, from June 2015 to October 2016, my weight has been around 172 lbs. while my food and meal quantity has been around 91% of my normal portions. But, if I want to further reduce my average weight below 170 lbs., I must cut down my food & meal quantity to around 80% of my normal portions.

Figure 3.3

 

 

 

 

 

 

 

Figure 3-3

Food & Meal Quantity (6/1/2015-10/20/2016)

Figure 3.4

 

 

 

 

 

 

 

Figure 3-4

Body Weight in the morning (6/1/2015-10/20/2016)

Now, let us examine the correlation between weight and glucose. Other than genetic factors, being overweight or obese is the main fundamental cause of Type 2 Diabetes (T2D). During my analysis of weight vs. glucose, when I plot out my weight >180 lbs. and daily average glucose >160 mg/dL, I have identified a strong correlation between these two factors. See Figure 3-5 and 3-6 below. My tool’s functionality of predicting post-meal glucose levels before consuming food is very crucial for controlling my PPG (Postprandial Plasma Glucose).

Figure 3.5

 

 

 

 

 

 

 

Figure 3-5

Weight over 180 lbs

Figure 3.6

 

 

 

 

 

 

 

Figure 3-6

Daily averaged glucose over 160 mg/dL

A person’s weight is constantly changing throughout the day and evening due to food and exercise as the most important factors. During the day time, food consumption increases weight and exercise burns off calories. Throughout the night, the processes of vaporization, urination, and bowel movement affect body weight reduction. During the past year, my average weight gain between bedtime and the morning of the same day is 2.8 lbs. My weight reduction between bedtime and the next morning is also 2.8 lbs. That is why I can maintain a constant weight at 172 lbs. (+/- 5 lbs.) over the past year. My tool’s functionality of predicting the next morning’s weight is a crucial component of controlling my weight; therefore, my weight control is a crucial part of controlling my diabetes condition. Figures 3-7 & 3-8 show my two kinds of weight changes.

Figure 3.7

 

 

 

 

 

 

 

Figure 3-7

Weight gain between bedtime and morning of the same day

Figure 3.3

 

 

 

 

 

 

 

Figure 3-8

Weight reduction between bedtime and next morning

 

Date:   10/24/2016  13:00

Section 4:   Glucose

On August 3, 2010, my lab test results showed my A1C as 10.0%, ACR as 116 mg/mmol, and triglyceride level as 1,161 mg/dL.  After receiving a warning from my physician, I decided to change my overall lifestyle.  I began collecting my health and lifestyle data on January 1, 2012.  To date, approximately 5-years’ worth of complete data has been collected.  In my cloud database, I’ve stored about one million data points regarding my body and lifestyle, which includes original inputted data, system calculated data, and AI processed data.  In Figure 4-1: daily glucose and 90-days average glucose, my glucose levels varied between 86 mg/dL and 227 mg/dL with an average value of 129 mg/dL.  In Figure 4-2: my lab-test A1C values were between 6.3% and 7.1% with an average value of 6.6%. My mathematical simulated A1C values (between 6.3% and 8.1% with an average value of 7.1%) had an accuracy of 92.4%.

The laboratory A1C test result reflects the average glucose values for the past 90 days. The eclaireMD Wellness APP calculates a mathematically simulated A1C value based on the user’s daily glucose input data. However, this mathematically simulated A1C value has a built-in artificial intelligence to automatically customize the calculation according to the changes of the user’s biomedical system’s parameters.

Frequent calibration by inputting laboratory test results can improve the accuracy of the mathematically simulated A1C results.  My mathematical A1C curve spectrum has completely “enveloped” the actual lab-tested A1C data from 2012 to 2016. This means that my simulated A1C can provide me with an upper-bound warning before I get tested for the laboratory A1C.

Figure 4-3 lists all of my lab-test A1C values. Although I have taken into account both the lifespan of red blood cells and developed different linear and nonlinear mathematical models, glucose still has a somewhat unpredictable output value due to a highly nonlinear biomedical body system that changes with time. There are many elements that affect glucose values and I will display more of my research results in following discussions regarding these elements.

Although during 2012 through 2016, the overall A1C curve remains at a somewhat steady state, i.e. around 6.6, the most important factor of controlling my diabetes condition is that I have been decreasing the dosage of my medication over two years’ time frame and finally completely removed all of my diabetes medications about one year ago.  Between 2012 and 2013, I was taking three different diabetes medications such as Januvia 100 mg, Actoplus Met 15 mg/850 mg, and Metformin 2000 mg.  I started to decrease the number of drugs and also reduce dosage amounts from the beginning of 2014.  By the end of 2015, I had completely eliminated taking any diabetes medication.

During this period, I witnessed different degrees of “withdrawal symptoms.”  Within one month of removing or reducing medication, my glucose chart fluctuated greatly, with many ups and downs, without any clear and reasonable explanation.

Figure 4.1

 

 

 

 

 

 

 

Figure 4-1:

Daily Glucose and 90-days Averaged Glucose (2012-2016)

 

Figure 4.2

 

 

 

 

 

 

 

Figure 4-2:

Mathematical Simulated A1C and Lab-tested A1C Comparison

 

Figure 4.3

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 4-3:

List of My Past Lab-tested A1C Values and Corresponding Mathematically Simulated A1C Values

I have been a long-term diabetic patient for almost 20 years.  I realized that having knowledge, a reliable tool, and will power are the three main ways to control this disease.  During the past 2 years, I have applied my acquired medical knowledge and developed my APP tools to help control both my weight and glucose.  The main tools are my weight predictor developed on April 11, 2015 and my glucose predictor developed on June 1, 2015.  It should be noted that the metabolism model (MI and GHSU) I developed in 2014 is the foundation for improvements in my overall health.  When I could predict my weight and glucose beforehand, it became much easier for me to adjust the amount and quality of my food, along with the frequency and intensity of my exercise.  In other words, if I can wisely adjust my input parameters, my output values will most likely be automatically adjusted for the better.  It should also be noted that the body is organic (nonlinear) and dynamic (changes with time), so we have to constantly monitor for signs of change.  Two fundamental rules had to be followed in terms of using these two prediction tools.  First, I had to follow the prediction model’s suggestions regarding input value as precisely as possible.  Second, after measuring my weight or glucose, I was not allowed to go back to readjust my original input values in order to change the prediction’s accuracy (unless there were some new findings or facts that I had just learned or realized from applying this prediction experience).  Some degree of artificial intelligence (AI) has been built into the system as well, but the entire bio-medical system needed to be continuously observed and modified along the way.  That is why I created a non-profit medical research foundation to continuously work in depth on this subject even after my death.

On July 1, 2016, I entered the following note in my diary:

“Based on the period of 6/1/2015 to 6/30/2016, 13 months of study on actual data analysis, my Chronic Tool has reached 99.2% accuracy rate on my A1C prediction and 98.4% accuracy rate on my 90-days averaged predicted glucose value.  Therefore, starting today 7/1/2016, I will not use the traditional blood test (test strip method) entirely to measure my glucose.  I will mainly depend on my mathematical model to predict my glucose in order to control my diabetes.  I will then go to a hospital or clinic to measure my A1C after 10/1/2016 to make sure that my diabetes condition is still under control.  During the upcoming 3 months, I must be extremely careful in monitoring my diet and meal contents and do my post-meal exercise diligently.  In addition, there are several primary factors to be monitored as well.  If this test is finally successful, my invented mathematical biomedical model on metabolism to control diabetes can reduce both cost and pain for worldwide diabetes patients.”

Since that day, I decided to decrease the amount of testing using traditional glucose meter (finger piercing and test strip method, or glucose meter / glucometer).  However, I still needed to continue testing my fasting glucose by using the glucometer because of my ongoing research of A1C variation due to different weighting contribution factors of FPG vs. PPG.  Whenever I ate my meals at restaurants where no reliable nutrition information was readily available, or I cooked a new dish at home using new food materials, I would measure my post meal glucose by the finger piercing method.  For the period of July 1, 2016 to October 23, 2016, I have 115 days’ worth of 460 glucose data, 115 FPG data (25% of total glucose data), and another 345 PPG data (33% of total PPG and 25% of total glucose data) as actual measured glucose data.  The remaining 230 data (50% of total data) were relied on my predicted glucose values. The tentative conclusion from this experiment is that I could eliminate the finger-piercing method and still get a 99.6% accuracy on my glucose results.  In order to validate my findings, I plan to continue this experiment for at least another year.  This is an important temporary finding that, if this conclusion is true and my prediction method is proven to be reliable, then most Type 2 diabetes (T2D) patients whose average daily glucose level fall between 100 mg/dL and 400 mg/dL will be able to use my prediction tool to control their diabetes.  It was my original goal of conducting this kind of reliability research to remove the burden and cost associated with finger piercing and test strips.  Hopefully, by using an easy and useful tool as an option, we can further reduce patients’ reluctance on their efforts of controlling diabetes.  Please see Figure 4-4: 50% of glucose data based on predictions (7/1/2016-10/23/2016).

Figure 4.4

 

 

 

 

 

 

 

 

Figure 4-4:

Preliminary Data Analysis to determine the reliability and accuracy of glucose prediction when patient dropping off finger-piercing and testing-strip method. These preliminary data have 50% of glucose values based on predictions (7/1/2016-10/23/2016)

No significant Correlation between FPG and PPG

I have conducted a statistical analysis of the correlation coefficient (r) and coefficient of determination (r2: r square) of FPG vs. PPG during the complete period of 6/30/2015 through 3/24/3017.  The results of both r (3.8%) and r2 (0.14%) from these two sets of data are very low, which means that the FPG values almost have no correlation with PPG values at all, see Figure 4-5.  The outcome makes a lot of sense from the bio-medical point of view.  Fasting glucose is the combination effect of glucose produced by liver and insulin produced by pancreas during sleeping hours.  On the other hand, during the awakening hours, both the diet control and exercise have direct impact on PPG values.

Figure 4.5

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 4-5

No correlated relationship between FPG and PPG during a period (6/30/2015 – 3/7/2017)

Multiple Factor Analysis (Weighted contribution factors) of FPG and PPG on A1C

I have collected my PPG values for more than 5 years and FPG values for almost 2 years.  Therefore, I can conduct a big data analytics on the composite impact of FPG and PPG on A1C value.   First of all, I have defined a new term, “Adjusted A1C”, as an interim processed variable to conduct this study.  The equation is:

Adjusted A1C = a x FPG + b x PPG  (b = 100% – a)

where a and b are A1C’s weighted contribution factor (%) of FPG and PPG, respectively.

I have used the data from the period of 6/1/2015 through 1/3/2017 (my latest A1C lab test date).  During this period, I have collected 9 sets of laboratory tested A1C values which are used as the base for this comparison study.  I then choose 5 sets of different weighted contribution factors for my data sensitivity analysis:

a = 0%, 2%, 5%, 8%, 11.4%;

and the corresponding

b = 100%, 98%, 95%, 92%, 88.6%.

This set of factors is to calculate 5 different adjusted A1C values in order to compare them against the lab tested A1C results which were collected on 9 different lab testing dates.  Figure 4-6 has shown 5 diagrams with 5 sets of adjusted A1C values in comparison with 9 different lab tested A1C values.  These 5 sets of different FPG/PPG weighted contribution factor will result into 5 different Adjusted A1C values.  By comparing Adjusted A1C and lab-tested A1C, I found high correlation coefficients are exited (from 41% to 60%).

Figure 4.6

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 4-6

5 sets of adjusted A1C values in comparison with lab tested A1C data on 9 different lab testing dates

In Figure 4-7, Statistical Comparison study of A1C values based on eclaireMD predicted, lab-tested, and Adjusted, it shows that the case of 0% of FPG and 100% of PPG is the closet mix to eclaireMD predicted A1C which also possess the highest correlation coefficient (60%).  However, the case of 11.4% of FPG and 88.6% of PPG mix gives a perfect match between adjusted A1C and lab-tested A1C.  This means that, in reality, my weighting contribution factors of FPG and PPG to A1C value are probably falling within the range of FPG < 10% and PPG > 90%.  Please note that, I have adopted the common medical community’s convention of using the lab-tested A1C as the basis of all of my glucose comparison study.

During this 21 months period since 6/1/2015, my averaged lab-tested A1C is around 6.55% (real lab-tested values fluctuating between 6.4% and 6.7%) without taking any diabetes medication.  Furthermore, the most probable weighted contribution factors: <11.4% of FPG and >88.6% of PPG, also indicate that my type 2 diabetes is very well under controlled by applying a better lifestyle management without taking any diabetes medications.

Figure 4.7

 

 

 

 

 

 

Figure 4-7

Statistical Comparison study of A1C values based on eclaireMD predicted, lab-tested, and Adjusted

Earlier Fasting Glucose Study

During the course of data collection and analysis, I have noticed some deviation that exists between tested data and simulated data.  My first attempt was trying to identify most of the important elements which affect glucose and A1C.  Some of them will be addressed in the following discussions.  Around April 2015, I became intrigued by the difference between higher fasting glucose in the morning (Dawn Phenomenon) and lower fasting glucose in the morning.  Without possessing a complete and clear knowledge of how the liver and pancreas function in terms of creating and controlling glucose level, I decided to study the fasting glucose phenomena by using a “macro-viewed” big data analytics approach.  Thus far, I have collected close to 700 morning pre-breakfast fasting plasma glucose (FPG) data.  Since the prediction of FPG is very different from the prediction of PPG (postprandial plasma glucose), and the linear contribution weighting factor for FPG to daily averaged glucose value (4 times glucose collection per day) is somewhat around 25%.  At first, I decided to use my 90-day average daily glucose as the initial condition for predicting FPG.  Please see Figure 4-8:  Predicted FPG value based on my preliminary finding of 360-days’ worth of data.  It indicates that, although Daily FPG goes up and down and it is difficult to predict, but after a long period of time, the averaged FPG settles around the 90-days averaged glucose value. The deviation between predicted and actual is only 1.6% and my predictions reach 98.4% accuracy.  It should be noted that the above analysis and tentative conclusion were based on data available prior to October 20, 2016.  During this period, most of my fasting glucose data are very close to my averaged daily glucose value.

Figure 4.8

 

 

 

 

 

 

Figure 4-8:

Predicted FPG value based on my preliminary finding of 360-days’ worth of data.

As shown in Figure 4-5, it clearly indicates that there is no influence between FPG and PPG ( correlation coefficient is 3.8%).  The three correlation coefficients between FPG and all of three separated PPG values (between 111 to 116 mg/dL) are also very low.   I am also puzzled about the quantitative characteristics of FPG, including what the exact causes of the surge, how to predict its pattern, and how high it will jump, particularly how to control the high FPG from happening.

Thus far, I have already collected my FPG data for 21 months.  However, my earlier conclusion of using my averaged daily glucose value as my newly predicted fasting glucose was turned over by my recent observed fasting glucose data since my FPG jumped up suddenly starting from November 22, 2016.   After collecting 4 more months of higher FPG values, I have had sufficient amount of data to investigate FPG.   I have chosen a data set of a period of 244 days, from 6/24/2016 to 3/24/2027.  During this period, I further subdivide them into two “equal length” sub-periods of 122 days each.

In the first sub-period from 7/24/2016 through 11/23/2016, the averaged FPG value is 115 mg/dL, while the second sub-period from 11/23/2016 through 3/24/2017 has an averaged FPG of 130 mg/dL.  It showed a 15 mg/dL increased amount of the averaged FPG value.

During this 4-months sub-period, there are no significant changes in my lifestyle, including food, exercise, stress, sleep, water drinking, life regularity, temperature, living environment, etc.  I have also read many articles and papers regarding FPG, tried many different tricks, e.g. taking snacks before sleep, measuring my 3:00 am glucose level, etc., and conducted numerous correlation analyses.  But, I still could not identify any clear pattern to serve as an useful tool on finding how to control my higher FPG in the morning.  I have also performed statistical analyses to compare current day’s FPG against the same day’s post-breakfast glucose and previous day’s post-dinner glucose.  The purpose of these two studies was to see whether FPG contributes an increase to the same day’s post-breakfast glucose or receives a “left-over” impact from the previous day’s post-dinner glucose.  Please see both Figure 4-9 & 4-10: correlation studies between FPG and two different PPG values.

Figure 4.9

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 4-9

Correlation study between FPG and same day’s post-breakfast PPG

Figure 4.10

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 4-10:

Correlation study between FPG and previous day’s post-dinner PPG

Recent Fasting Glucose Study

In the past, my correlation analyses were focused on relationship between input categories, e.g. diet, exercise, living environment, etc. and output categories, e.g. glucose, weight, blood pressure, etc.   I am a professional engineer with strong mathematics and computer science background.  My past training in engineering disciplines follows this thinking route: develop a governing equation, define prominent input values, and finally derive outputs as the system’s solution.  On March 17, 2017, I suddenly thought about the possibility of one output category could also be served as the influential input factor of another output category.   After I break out from my previously trained thinking constraints, I then be able to look into the FPG problem with a different angle.  In the past, I was trying to identify the correlation and inter-dependency pattern between FPG and other input factors, but totally ignored another output category, weight, even though weight is depending on some inputs, such as food and exercise.   I have already stored my FPG values since 6/1/2015 and all other data in my database.

As shown in Figure 4-11, I have found a high correlation coefficient (47%) exist between FPG and Weight, and a moderate correlation coefficient (24%) exist between Wight & Food Quantity.

Figure 4.11

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 4-11:

Correlation studies between FPG & Weight, and Wight & Food Quantity

 

Figure 4.12

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 4-12:

Weight increase and FPG increase during the second sub-period (11/23/2016 – 3/24/2017)

A very clear correlation picture can be observed in Figure 4-13, when I plot them out under the selection criteria of weight data of >175 lbs and FPG data of >130 mg/dL.  I also realize that, during the past 5 years, I have already controlled my PPG very tightly via diet, exercise, and others.  However, during this 4-months winter sub-period, I have eaten too much “between-meal” snacks, e.g. nuts.  This eating pattern did not contribute much on my PPG value, however, it did result in a 4 to 6 pounds of weight increase which in turn caused the increase of 15 mg/dL on FPG.

Figure 4.13

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 4-13:

Strong correlation for weight > 175 lbs and FPG > 130 mg/dL

Of course, the increased FPG values will most likely also to push up my A1C values.  As I mentioned earlier that my FPG contributes <11.4% and my PPG contributes >88.6% on my predicted A1C value.  Therefore, I am guessing that the forthcoming FPG surge will probably push up my lab-tested results from 6.5 on 1/4/2017 to around 6.7 on 4/1/2017, my scheduled next  medical lab test.

However, if you conduct a statistic analysis (least square mean and correlation coefficient), then you will find that, from Figure 4-14, r is only 14% for weight/PPG comparing against 47% for weight/FPG.  Now, the question is why PPG only has 14% of correlation while FPG has 47%?

Figure 4.14

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 4-14:

Correlation comparison between Weight & FPG and Weight & PPG

My personal interpretation of the correlation coefficient of 14% between post-meal glucose and weight is due to the fact that weight is only a second-level influence on PPG.   The other two lifestyle factors, diet via carbs and sugar control (r=65%) and regular exercise (r=27%), are true primary factors.  Although our body weight (one of body’s output) is controlled by our lifestyle, diet and exercise (two of body’s inputs) as well.  During daytime, we can manipulate both diet and exercise to control our post-meal glucose level to an optimal level.   However, during sleeping hours, it is a different story.  For most cases, during sleep time, other lifestyle factors can not be maneuvered by us and therefore, our brain takes over the total control of operation of our internal organs.  When brain dictates body weight has been increased and then it will give orders to liver to produce glucose around 3AM for storing tomorrow morning’s needed energy.  If liver produces excessive glucose, the brain then gives order to pancreas to produce insulin to balance the glucose level.  But diabetes patient’s pancreas function is malfunctioned, therefore this malfunctioned glucose control mechanism will push up our body’s glucose in the morning (FPG).  Since diet and exercise cannot alter FPG directly during sleep, therefore, the secondary factor, weight, becomes the input factor and comes into play.  This is my understanding and interpretation.

From now on, I will reduce my body weight down to 170-172 lbs range within next month. After achieving this goal, I will continue to monitor both of my weight and FPG for at least another 4-months period of “steady-state” in order to have a complete years of records which composes 3 sub-periods, with averaged FPG of low/high/low.  Through this experiment, I can then confidently draw this conclusion of correlation between weight and fasting glucose and verify it.  In Figure 4-15, I have summarized my predicted FPG values under 6 different body weights in the morning.

Figure 4.15

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 4-15:

My predicted FPG values under 6 different body weights in the morning.

I have selected previous X-months, where X=3,4,5,6,7,8, data of predicted FPG (including the influence from the correlated weight factor) and actual measured FPG to conduct my comparison study.  The results are shown in Figure 4.16 that the accuracy of mean comparison is greater than 98%, while the correlation coefficients fall into the range of 12% to 40%.

Figure 4.16

 

 

 

 

 

 

 

 

 

 

Figure 4.16

Data Sensitivity Study for predicted FPG

Since I developed my post-meal glucose prediction model on 6/1/2015 and my fasting glucose prediction model on 3/18/2017, I have collected 668 days of glucose data.  In the background of predicted glucose, I collected and analyzed data for the following categories: exercise, food, carbs, sugar, stress, weight, etc. which included other calculated and processed information amounting to a total of several hundred thousand data. The final results of FPG along with the PPG comparison between predicted and actual glucose during the period of 7/11/2015 through 3/30/2017 are displayed in Figures 4.17 and 4.18.  As a result, the accuracy of mean values of glucose is greater than 95% while correlation coefficients are within the range of 66% to 89%. This conclusion proves the accuracy and reliability of my glucose prediction models. Furthermore, the lab-tested A1C values also confirmed its accuracy.

 

Figure 4.17

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 4.17

 

Figure 4.18

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 4.18

Please see the following chart, Figure 4.19.  I have used the previous 90-days body weight correlated model to predict my FPG values (7/11/2015-4/8/2017), which is very accurate. The correlation coefficient is almost 73%.

Figure 4.19

 

 

 

 

 

 

 

 

Figure 4.19

Date:   10/25/2016  13:00

Section 5:   Glucose and Food & Exercise

In order to study the relationship between glucose and food, I have developed an APP known as SmartPhoto for the iPhone.  Within SmartPhoto, I constructed a relational database structure to attach with each picture stored in the iPhone album.  The data structure has 5 levels:

1. Group:  (USA, Japan, France, etc.)

2. Category: (home cooking, chain restaurant, individual restaurant, airline, cruise, etc.)

3. File: (Denny’s, McDonald’s, Greek Restaurant, Asian Food, etc.)

4. Name: (restaurant name, dish name, menu item, etc.)

5. Content: (anything you want to keep a record, e.g. Nutrition ingredients, etc.).

 

Once food photos with its data structure are stored in SmartPhoto, they can be sorted and searched any way a person chooses.

Please see Figure 5-1: SmartPhoto Samples of food & meal with glucose level attached with each photo.  From May 1, 2015 through October 20, 2016, I collected a total of 1,591 pictures of food and meal with an average glucose level (PPG) of 121.8 mg/dL.  During the same period, my daily averaged glucose level (including FPG) is 121.41 mg/dL.

Figure 5.1

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 5-1:

SmartPhoto Sample Pictures of Food & Meal

 

From the years of 2012 through 2014, during my glucose analysis, I came to a tentative conclusion that my high glucose periods (close to 140 mg/dL) were contributed to traveling overseas.  Please see Figure 5-2: Glucose results from 2012 to 2014.  I found that the majority dishes of Eastern Asia (excluding Northern China), Hawaii, and Tahiti contain high contents of sugar in their cooking process.  Rice, flour, and/or taro are the main sources of carbohydrates. However, during my recent extended stay in various eastern Asian countries and Hawaii over 8 months has introduced another prominent fact.  My average glucose level dropped below 120 mg/dL – a drop of 20 points from previous periods.  Please see Figure 5-3: Glucose results from 2015 through 2016.

After careful analysis of this average glucose decline, I discovered the following four reasons:

(1) I followed my rule of choosing food material and picking menu items more cautiously when I use the glucose prediction capability of my tool;

(2) I spread my daily walking exercise to three post-meal time frames, averaging 4,000 steps after each meal, instead of concentrating on one walk in the evening (this will be discussed in future sections);

(3) I watched my food and meal intake and walking exercise more carefully on traveling days.  For example, after eating a meal at the airport restaurant or airline lounge, I make the effort to walk 3,000 to 4,000 steps in the aisles connecting the boarding gates;

(4) My SmartPhoto tool’s analysis capability also provides me many insights regarding dining locations, food menus, and cooking material selection.

Figure 5.2

 

 

 

 

 

 

 

Figure 5-2:

Glucose during period of 2012-2014

 

Figure 5.3

 

 

 

 

 

 

 

Figure 5-3:

Glucose during period of 2015-2016

From examining the big picture data in SmartPhoto, I tabulated the results in Figure 5-4: Summary Table of Averaged Glucose at Different Eating Locations.  There are a total of 1,591 food and meal pictures with an average of PPG value of 121.8 mg/dL.  During the same period from May 1, 2015 to October 20, 2016, my average daily glucose from my APP tool, including FPG, is 121.41 mg/dL – this is another supporting point of why I decided to use my average daily glucose value as the initial predicted FPG value – while my 90-day average glucose is 123.75 mg/dL as shown in Figure 5-5: Glucose during SmartPhoto Period from May 1, 2015 to October 20, 2016.

Figure 5.4

 

 

 

 

 

 

 

 

 

Figure 5-4: Summary Table of Averaged Glucose and Different Eating Places

Figure 5.5

 

 

 

 

 

 

 

Figure 5-5:

Glucose during SmartPhoto Period (05/01/2015-10/20/2016)

 

My preliminary explanation and interpretation of causes for these summarized results are as follows:

(1) The average glucose values in all the studied nations are similar, measuring between 119.9 and 125.6 mg/dL.  From 2015 to 2016, I followed strict rules for food and meal intake along with the similar ratio between eating at home and eating outside in every country.

(2) Home cooking equates to a 115.3 mg/dL of glucose value, eating in chain restaurants (where nutritional ingredients information is published) equates to 125.2 mg/dL, eating in individual restaurants (where nutrition information is unavailable) equates to 132.3 mg/dL, eating at an airport, in an airline lounge, and in-flight meals equates to 134.0 mg/dL, and eating ready-cooked food from supermarkets equates to 140.6 mg/dL.

(3) Airline-related food produces high glucose points due to the fact that there are limited options on food items and limited space for post-meal exercising.

(4) After studying the nutrition of major food items, I have tried not to eat processed foods.  However, when I have limited options, I can still eat them provided that I read the nutrition facts on the labels carefully (especially the information regarding carbohydrates and sugar).

(5) Further detailed analysis regarding individual restaurants received the following average glucose values:

USA: 129.8 mg/dL

Japan: 139.6 mg/dL

Taiwan: 136.7 mg/dL

Other Nations: 130.8 mg/dL

Figure 5.6

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 5-6: Measured Average Glucose for Different Eating Places

 

In general, American and Western food do not include sugar in the cooking process (except in desserts).  Japanese, Korean, southern Chinese, and Southeast Asian cultures add both sugar and salt into dishes during the cooking process.

(6) I discovered one interesting observation from analyzing a particular popular brand of chain restaurants.  Usually, I avoid eating lunch or dinner at any chain restaurant.  However, breakfast is an exception since the portions are usually smaller due to economic reasons.  As a result, the portion of carbohydrates and sugar are also greatly reduced in certain chain restaurants’ breakfast foods.  This same particular brand of American chain restaurant has an average glucose value of 122.9 mg/dL, while Japan has 117.4 mg/dL, Taiwan has 125.3 mg/dL, and China has 126.2 mg/dL.  My observation is that this particular chain restaurant in Taiwan and China add some local flavors to the menu items; furthermore, I suspect that its standard operating procedures (SOP) of procurement and cooking may not completely comply with its headquarter’s requirements.

(7) From 2013 to 2014, while I was studying food and nutrition, I drew an incorrect conclusion that I could eat as many vegetables as I wanted.  Later in 2015, after I compiled several million points of data on food nutrition, I discovered the differences among various vegetables.

One way I distinguish between how different vegetables affect my glucose is by color.  Please see Figure 5-7: Summary Contents of Carbs and Sugars in Vegetables.  I came to the conclusion that if I eat large quantities of vegetables, my PPG can increase to a higher value.  I must pay attention to the color of vegetables when I eat them in order to get a more accurate glucose prediction.

(8) When I have a craving for snacks, desserts, and/or fruits, I can definitely consume them, however, I must limit the quantity in order to control both my glucose value and weight.  The best practice for me is to eat limited amount of them between meals, for example at 10 am or 3 pm.  I avoid giving in to my cravings before bedtime to assist with my weight control.  Fruits are important for overall body health, however, it is important to avoid eating high-sugar content fruits (such as pineapples, bananas, etc.) and also limit the quantity consumed.  With this control mechanism, I can maintain a healthy level of glucose.

Fig 5.7

 

 

 

 

 

Figure 5-7:  Summary Contents of Carbs and Sugars in Vegetables

It would be interesting to analyze the “extreme” cases in my records, e.g. studying glucose over 200 mg/dL.  Figure 5-8 displays all of my 17 meals which contributed to glucose over 200 mg/dL from May 1, 2015 to October 20, 2016.

It should be noted that the 3 major sources contributing to my extremely high PPGs are eating at individual restaurants offering East Asian food, American chain restaurants, and meals on airlines and cruises.  I can still eat at these locations provided I have knowledge of the food nutrition, use the right tool to predict post-meal glucose value, and have sufficient willpower to resist giving into cravings at the wrong time of day.

Figure 5.8

 

 

 

 

 

 

 

 

 

Figure 5-8:

17 meals contributed to PPG over 200 (5/1/2015-10/20/2016)

Furthermore, as indicated in the following Figure 5-9: Analysis of Causes for Glucose Values Greater Than 140, it is clear that high carbs & sugar food and Asian food have contributed about 58% of higher glucose values (>140mg/dL) causes.  Another interesting fact is that about 10% of unknown reasons occurred, which means I could not explain the actual causes of those high glucose values.

Figure 5.9

 

 

 

 

 

 

 

 

Figure 5-9:

Analysis of Causes for Glucose Values Greater Than 140

Research has shown that carbohydrates and sugars directly affect glucose levels.  By using the following rules, I can estimate my glucose level before I consume my meal by controlling the food quality and quantity.

(1) I can find the ingredients on the Nutrition Facts label on the food packaging.  I use the amount provided in terms of grams divided by 20 to get the portion estimate.  For example, carbohydrate has 16 grams, then calculate it as 16/20=0.8. The value of 0.8 is entered into the carbs input box of the tool.  I also calculate the sugar amount by using the same method.

(2) When I cook at home, I need to estimate the percentage based on using my open-hand area for estimation or my fist size for volume estimation as 100%.  However, based on my observation for the past few years of portion estimation, I have noticed recently that I need to reduce my 100% estimation to 2/3 of my hand or fist size.  My guess is that my body’s toleration of carbohydrates and sugar has been reduced due to the effects of diabetes.  After collecting more data regarding this phenomenon, I may need to build another layer of artificial intelligence to address this organic change.

(3) My tool can also search each item of my food components from the food bank, and then add them up to get the total consumption of both carbohydrates and sugar.

(4) Most fruits have both carbohydrates and sugar, but some fruits such as bananas, pineapples, and grapes have higher carbohydrates and sugar content.

(5) It is highly recommended not to eat any desserts, since they contains high carbohydrates, sugar, salt, and fat, which are not healthy for you.  Try to eat plenty of green leafy vegetables; but avoid or reduce non-green vegetables such as beets, carrots, corn, onions, and tomatoes which have higher sugar content.

The most important principle for diabetic patients is to “even out” their glucose wave during the entire day, i.e. push the high tide downward (reduce hyperglycemia) and lift the low tide up (avoid shock from low glucose) like an ocean wave.  Once you are able to maintain your target weight and have a balanced nutrition, your diabetes and other chronic disease conditions should be under control.

Correlation between PPG and Diet (3/8/2017)

PPG (post-meal glucose) values are greatly influenced by our lifestyle, including primarily diet (intake of carbohydrates and sugar) and exercise (within a 2-hour period after each meal).  In Figure 5.10 and 5.11, the correlation results of r = 64.7% and r2 = 41.8% which show that a very strong positive correlation exists between daily averaged PPG and the averaged intake of carbs and sugar during daily meals.

Fig 5.10

 

 

 

 

 

 

 

 

Figure 5.10

Correlation between PPG and diet (carbs and sugar in gram)

Fig 5.11

 

 

 

 

 

 

 

 

Figure 5.11

Detailed daily averaged intake of carbs and sugar per meal (around 15 gram per meal)

More detailed analysis were done for investigating dinner’s glucose and intake of carbs + sugar by using both a “Mean (Average) method and Least Square Mean (LSM) method. The results are displayed in both Figure 5.12 and 5.13 which show strong positive correlation values of r = 64.4% and r2 = 41.5%.  These dinner’s correlation results are almost identical as the daily meal averaged values.  Using the Mean method, I get an averaged post-dinner glucose of 115.64 mg/dL and an averaged intake of carbs + sugar of 13.90 gram; however, using the LSM method, I get averaged post-dinner glucose of 115.66 mg/dL while keeping an averaged intake of carbs + sugar of 13.90 gram also.  This shows that there is no significant difference between using Mean or LSM method.

Fig 5.12

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 5.12

Least Square Mean Calculation for Correlation between PPG and Diet

 

Fig 5.13

 

 

 

 

 

 

 

 

Figure 5.13

Averaged Glucose and Carbs/Sugar using Mean Calculation

Glucose and Exercise:

Other than food and meals, exercise was another important factor that contributed to my glucose reduction.  In my APP, I include many different types of exercises.  In my many years of experience, I believe that walking at average speed is the best kind of exercise for many senior citizens.  My average walking speed is 2.5 miles per hour, about 6,000 steps per hour, or 100 steps per minute.

In 2012, I walked an average of 8,000 steps, or 3.3 miles, per day.  During that time, it was difficult for me to walk too long because I was overweight and had weak knees.  By 2016, I gradually increased to 17,200 steps per day, or 7.2 miles per day without any difficulty.  Please see Figures 5-14, 5-15, and 5-16.

In the beginning of 2015, I discovered my PPG would significantly decrease if I spread out my daily walking exercise into 3 segments, i.e. exercising within 2 hours after finishing each meal.  By examining my glucose data for extended periods of time, I also learned that when my average glucose was around 140 mg/dL. With every 1,000 steps taken after a meal, I could reduce my glucose level 7 to 10 points.  However, when my average glucose value dropped to around 120 mg/dL, I could reduce my glucose level 4 to 6 points with every 1,000 steps after a meal.  This difference is due to the assumptions I made in my mathematical models.

Fig 5.14

 

 

 

 

 

 

 

 

 

 

 

Figure 5-14:

Walking Exercise (2012-2016)

Fig 5.15

 

 

 

 

 

 

 

 

 

Figure 5-15:

Waking Exercise Concentrating in the Evening (2012-2014)

 

Fig 5.16

 

 

 

 

 

 

 

Figure 5-16:

Walking Exercise Spreading over After 3 Meals (2015-2016)

 

Correlation between PPG and Exercise

In Figure 5.17 the correlation results of r= 27.9% and r2 = 7.8% show that a weaker but still significant negative correlation between PPG and averaged walking steps after each meal.  Please note that, as indicated in the same figure, I have walked around 4,000 steps within a 2 hour period after each meal.  In this way, I could take full advantage of walking exercise to improve my daily post-meal metabolism and also reduce my PPG values.

At present, I walk approximately 18,000 steps or 7.5 miles per day.  About 2 months ago, I felt some pain on my heels and I worried about over-exercising my joints.  Therefore, I decided to replace a part of my daily walking exercise with Tai-chi, a slower body motion and stretching movements.  I plan to collect big data over a much longer period of time to study the effect of Tai-chi on my glucose control situation more precisely.

 

Fig 5.17

 

 

 

 

 

 

 

 

Figure 5.17

Correlation between PPG and exercise (post-meal walking steps)

Date:   10/28/2016  13:00

Section 6:  Glucose and Others (Stress, Travel, Temperature)

Glucose and Stress:

Stress causes many health problems.  When people go through a long and continuous stressful lifestyle pattern, it can severely affect their health.  During my demanding thirty-year career, I endured a constant stressful lifestyle. This led me to have severe chest pains on 5 different occasions, and severe Type 2 diabetes, which resulted in chronic toe injuries, bladder damage, and kidney damage.  However, after retirement from my business career, I have enjoyed a peaceful, non-eventful lifestyle (except for the year 2014).  During that year, I went through 3 episodes of higher than normal stressful events from March through June, then from September through October, and again from November through December. Please see Figures 6-1: Comparison of Stressful Periods from March to December 2014 and Peaceful Period from January 2015 to October 2016.

Figure 6.4

 

 

 

 

 

 

 

Figure 6-1

Stress Scores Comparison of Stressful period (3/2014-12/2014) and Peaceful Period (1/2015-10/2016)

 

From the following Figures 6-2, 6-3, 6-4, 6-5, and 6-6, we can observe the clear correlation among stress, blood pressure readings, glucose values, and A1C levels.

Since the second and third stressful events occurred back to back, the charts reflect the high stress score, hypertension, and glucose to align with two time spans from March through June and again from September through December. However, the A1C value peaks approximately 3 months later than these time spans because A1C takes 3 to 4 months’ worth of average glucose values.

Figure 6.5

 

 

 

 

 

 

 

Figure 6-2:

Stress Score During 2014

Figure 6.6

 

 

 

 

 

 

 

 

Figure 6-3:

Higher Blood Pressures During Stressful Periods

Figure 6.7

 

 

 

 

 

 

 

Figure 6-4:

Higher Daily Glucose During Higher Blood Pressure and Stressful Periods

Fig 6.5

 

 

 

 

 

 

 

Figure 6-5 :

Higher A1C peaks Appear around 3 months Later of High Glucose

 

Figure 6.9

 

 

 

 

 

 

 

Figure 6-6:

Putting Higher Stress Scores and Higher 90-days Averaged Glucose Together

Furthermore, I suffered two separate physical traumas in 2015.  The first incident occurred on June 23rd, where I fell on a sloped walkway resulting in a face injury and an emergency room visit.  My recorded glucose values for the following three days after the accident were 152 mg/dL, 208 mg/dL, and 154 mg/dL, but then on the 4th day, the value dropped to normal level around 120 mg/dL.  The second incident occurred on December 4th, where I sustained a leg injury on a construction site and went to the emergency room again.  My recorded glucose values for the following three days after the accident were 145 mg/dL, 175 mg/dL, 165 mg/dL and then dropped to normal level around 120 mg/dL on the 4th day.  As a result of these two stressful incidents, my glucose level increased temporarily. Both cases took 4 days to allow the traumatic impact on my glucose to diminish.  On April 8, 2016, I fell after missing a step on a stairway, but I did not sustain any physical injury.  However, two hours after eating my normal low-carbohydrate & low-sugar breakfast, my PPG had spiked at 148 mg/dL. Not only had I ate the same breakfast as I always did, but I had also walked 4,000 steps afterward.  Although this was a minor incident, it demonstrated that stress can affect the glucose level.

Glucose and Travel:

Throughout my life as a businessman, I have traveled extensively worldwide.  To make it simple for my glucose discussion, I have only compiled my traveling record from 2012 to 2016.  I define long trips as air travel time with more than 3 hours (along with +/-2 hours going in and out of the airport) which can affect about 2 meals.  I define short air travel trips as flying time with less than 3 hours (along with +/-2 hours going in and out of the airport) can affect about 1 meal.  During the past 5 years, on average, I have flown every two weeks, or more precisely, every 12.9 days.  From my analysis of my health status on flying days, I noticed that both my glucose and metabolism were affected noticeably by the traveling.  The two main reasons my glucose and metabolism were affected are due to airline food not being the best option for diabetics, and limited space for exercise. Once I figured out the main causes, I managed my air travel meals very carefully by watching what I can eat safely and to walk as much as I can after my meal in the crowded airport space. Therefore, during 2015 to 2016, both my glucose and metabolism index during air travel days have greatly improved, almost reaching my normal level of 120 mg/dL.  These analysis results are in Figure 6-7.

Figure 6.10

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 6-7:

Correlation Among Glucose, Metabolism and Air Travel

 

Glucose and Weather:

I spent 40 years living in different states in the U.S. with less pollution, great weather, and mild climate (with temperatures ranging from 15 degree to 25 degree Celsius).  During the first half of 2016, I stayed in East Asia continuously for more than 6 months, throughout winter, spring, and early summer.  Although I was traveling to different cities, I disciplined myself to maintain a routine lifestyle, which includes monitoring food, exercise, stress, sleep, water intake, other routines, etc.  However, I noticed my glucose level continued an upward trend during February through June, when the temperature was getting hotter in Asia. I could not explain why but I wondered if the hot weather conditions affected my metabolism.  Figure 6-8 provides a preliminary and short period (about 4.5 months) of data observation.   I wrote this information here to invite other researchers’ attention and input on this topic.

Figure 6.11

 

 

 

 

 

 

 

Figure 6-8

Correlation Between Glucose and Atmosphere Temperature

 

Date:   10/29/2016  15:00

Section 7:  Hyperlipidemia and Hypertension

I have compiled my physical examination data and entered them into my tool since the year 2000.  The plotted lipid graphics are reflected in Figure 7-1, 7-2, 7-3, and 7-4.  Results have shown that I suffered from hyperlipidemia from 2000 to 2012.  Since 2012, although my focus was to control my diabetes, my overall strategy was to utilize preventive medicine via a better and effective lifestyle management.  The mathematical-simulated metabolism model developed in 2014 provided an effective tool for lifestyle management.  As a result, while my glucose values were under control, it assisted me in changing my unhealthy lipid data into a healthy state.  Further discussions about metabolism will be found in the next section.

Fig 7.1

 

 

 

 

 

 

 

Figure 7-1:  Triglycerides (2000-2016)

 

Fig 7.2

 

 

 

 

 

 

 

Figure 7-2:  HDL-C (2000-2016)

 

fig 7.3

 

 

 

 

 

 

 

Figure 7-3:  LDL-C (2000-2016

 

Fig 7.4

 

 

 

 

 

 

 

Figure 7-4:  Total Cholesterol (2000-2016)

 

Lipids have a close relationship with the “quality” of food.  In order for me to consume a better quality of food, I included a list to serve as both reminder and record for the food quality.  Please see Figure 7-5: Reminder and Record of Quality for Food & Meal.  In this table, if you follow all the rules, then you will get a score of 0.5. If you violate all the rules, then you will get a score of 1.5. Please see Figure 7-6: Score of Quality for Food & Meal from mid-2014 to October, 20 2016.  My “Quality for Food & Meal” satisfaction level is 96% – reflecting a score of 0.54, satisfaction level = (1.5-0.54)/(1.5-0.5).

Fig 7.5

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 7-5:

Reminder and Record of Quality of Food & Meal

 

Fig 7.6

 

 

 

 

 

 

 

Figure 7-6:

Score of Quality of Food & Meal

 

My blood pressure data is shown in Figures 7-7, 7-8, and 7-9. As I mentioned in an earlier section, from March through December of 2014, I had 3 consecutive stressful periods and their impact on my blood pressure can be seen on these diagrams.  I have already included “limit salt intake” as one of the requirements in “Quality for Food & Meal,” which I firmly follow in my diet.  From Figure 7-10: Analysis of Causes for High Blood Pressure, my data showed that major stressful events are highly correlated to hypertension, which is then followed by overseas traveling, jet-lag, post exercise timing, extreme weather condition, etc.

Figure 7.7

 

 

 

 

 

 

 

Figure 7-7:  Highest Daily SBP & DBP

Figure 7.8

 

 

 

 

 

 

 

Figure 7-8:  Averaged Daily SBP & DBP

Figure 7.9

 

 

 

 

 

 

 

Figure 7-9:  Averaged Daily Heart Rate

Figure 7.10

 

 

 

 

 

 

 

 

Figure 7-10:  Analysis of Causes for High Blood Pressure

 

Date:   11/1/2016  11:00

Section 8:  Metabolism Index (MI) and General Health Status Unit (GHSU)

For the entire year of 2014, I have conducted research and development on the subject of overall health and chronic diseases.  At the beginning, I tried to find a good definition of “Metabolism” but failed.  For example, the Webster Dictionary defines it as “metabolism = the organic processes (in a cell or organism) that are necessary for life.”  Finally, I tried to define metabolism in a quantitative way.

I created a new term called Metabolism Index (Ml). It is based on four categories of human body health’s daily output data and six categories of human body health’s daily input data related to chronic diseases.  The four categories of daily output include body weight, blood sugar, blood pressure, and lipid. The six categories of daily input include exercise, water drinking, sleep, stress, food and meals, and daily routines.  Since input, output, and the biomedical system are dynamic, i.e. they are changing with time; I included “Time” as the eleventh category.

Within each category, there are many more elements.  For example, there are 8 elements in sleep, 33 in stress (not all elements are suitable for everyone), and approximately 100 for food and meal, etc.  At the end, there are several hundred of elements that need to be addressed, recorded, and analyzed.  Of course, it is a huge burden to figure them out effectively on a daily basis.  The biggest challenge is how to solve the inter-connectivity issues among 11 different categories and hundreds of elements.  As a result, I utilized the finite element concept and dynamic plastic behavior of structural engineering to model this system.  I was able to build a set of mathematical governing equations with various boundary conditions.  With these efforts, the remaining problem to solve was to apply computer science, especially computational automation, and artificial intelligence.  This is where big data analysis and analytic come into play.

The General Health Status Unit (or GHSU) is the moving average of Metabolism Index (MI) over the most current 90 days.  Originally, I defined MI to fall within the range of 0.5 (best condition) to 1.5 (the worst condition).  When both MI and GHSU are under 1.0, it means that your health is generally good.  However, if these values are over 1.0, you may have some health issues or related lifestyle problems.  For myself, I finalized an optimal set of elements within each category and also defined my desired healthy level status: 170 lbs. for body weight; 120 mg/dL for glucose; 120/80 for SBP/DBP; and 150/40/130/200 for triglycerides/HDL-C/LDL-C/total cholesterol.  The “break-even” level for both MI and GHSU is actually 73.5%, i.e. above 73.5% is unhealthy whereas below 73.5% is healthy.  Please note that I have adopted the general medical practice of the lower value to represent as better or healthy.

As of October 20, 2016, my MI and GHSU are at 58.45% and 57.1% respectively, which indicate that I am healthy.  My physicians have also confirmed that my general conditions are very healthy based on various laboratory test results of the past 2 years. This is an actual application of how to control chronic diseases via applying quantitative medicine on lifestyle management, i.e. a branch of preventative medicine and translational medicine.  Please see Figures 8-1 and 8-2 regarding my MI and GHSU for a period of 2012 through 2016 and another period of April 11, 2015 to October 20, 2016.

Figure 8.1

 

 

 

 

 

 

 

Figure 8-1:  MI & GHSU (2012-2016)

 

Figure 8.2

 

 

 

 

 

 

 

Figure 8-2:  MI & GHSU (4/11/2015-10/20/2016)

 

With the introduction of basic concepts regarding MI and GHSU, let us examine scores of some major categories. In previous sections, we have already seen many figures of collected data summary, such as weight, waistline, glucose, blood pressure, lipids, food & meal, exercise, and stress.  The remaining missing categories are also important but not so directly linked with the health output data, particularly diabetes related.  I will repeat food & meal data and figures in this section.  Figure 8-3 lists the summary category scores derived from my mathematical computational model and their transformed “satisfaction level” which is a self-explanatory phrase.

Besides the Metabolism Model that was developed in 2014, two other major breakthroughs were produced: the Weight Prediction released on April 11, 2015 and Glucose Prediction released on June 1, 2015. These three models provided tremendous help and accurate guidance to help control my diabetes and other chronic diseases.

Therefore, in this section’s data and figure display, I select the period from April 11, 2015 to October 20, 2016 as the standard common period for comparison.

Figure 8.3

 

 

 

 

 

 

 

 

Figure 8-3:

Conversion Table of MI Category Scores to Satisfaction Levels

 

My drinking water score is 0.74 and its satisfaction level is 95% (100% is defined as drinking 6 bottles or 3,000 ml of water per day). During this period, I have been drinking 5.7 bottles or 2,850 ml of water on average per day.

Figure 8.4

 

 

 

 

 

 

 

Figure 8-4:  Water Score

 

Three major stressful events happened to me in 2014; however, during this period (4/11/2015-10/20/2016), I did not encounter stressful situations. Therefore, my stress score is 0.51 and its satisfaction level is 99%.

Figure 8.5

 

 

 

 

 

 

 

Figure 8-5:  Stress Score

 

Sleep category has 8 elements. Among them, sleep hour and sleep interruption due to waking up are the most important two elements for my case.  My total sleep score is 0.74 and its satisfaction level is 86%, not bad at all.

Figure 8.6

 

 

 

 

 

 

 

 

Figure 8-6:  Sleep Score

 

During this period, I slept 7 hours and 15 minutes per night on average, which is quite sufficient.

Figure 8.7

 

 

 

 

 

 

 

Figure 8-7:  Sleep Hours

For most male senior citizens, night time urination due to prostate enlargement is the most disturbing factor affecting sleep.  In my case, I was told by my physician that my bladder was damaged due to the long term effects and severity of diabetes.  During the years from 2012 to 2014, I used to wake up 4 times at night to use the bathroom. However, during the period from 2015 to 2016, I only wake up 1.8 times (less than 2) per night on average. This improvement was entirely due to a better lifestyle management, not from intaking of urological medication.

Figure 8.8

 

 

 

 

 

 

 

 

Figure 8-8:  Sleep Disturbance due to Wake Up

 

My food and meal score is simply the average of both quantity score and quality score.  It is 0.73 and its satisfaction level is 77%, which is a decent score.

Figure 8.9

 

 

 

 

 

 

 

Figure 8-9:  Food & Meal Score

 

I will repeat both the food quantity and food quality scores in order to emphasize their different roles in controlling chronic diseases.  Food & Meal Quantity control is important for weight control, and in turn to control multiple chronic diseases.  My score is 0.91, or 91% of my normal food consumption (portion size) which allows me to maintain my weight at 172 lbs. I have started another push to drop my weight down to 168-169 lbs level by reducing my portion size to around 80%.

Figure 8.10

 

 

 

 

 

 

 

Figure 8-10:  Food & Meal Quantity Score

 

My food quality score is 0.54 and its satisfaction level is 96%.  You can get the 100% score if you follow the ready-defined 20 rules precisely every day.  From my experience, this score helps me to lower my blood lipid to reflect the data in the healthy level status. Genetically, I was born with low blood pressure, but previously as a businessperson I encountered many stressful events that caused me to have “temporarily but not so severely” hypertension.  I refuse to take medication for my “high” blood pressure, so instead I changed my lifestyle in order to correct this health problem.

Figure 8.11

 

 

 

 

 

 

 

Figure 8-11:  Food & Meal Quality Score

 

Finally, let us examine my daily routine life pattern score of 0.74 and its satisfaction level 95%.  This means that I follow a regular routine in my daily life pattern.  This category has a total of 14 elements to be checked on a daily basis.  Evidence has shown that a simple and regular life pattern contributes a lot to life longevity.  I finally was able to live this kind of life after I retired from a highly competitive business career and to find new interests to pursue, while maintaining a simple but routine life.

Figure 8.12

 

 

 

 

 

 

 

 

Figure 8-12:  Daily Routine Score

 

Date:   11/1/2016  14:00

Section 9:   Conclusions

This project “Using Quantitative Medicine (a branch of Translational Medicine) to control Type 2 diabetes” started in August of 2010 through the end of 2016.  For 4 years, I self-studied several chronic diseases and food nutrition in depth.  In addition, I invested 3 years of research and develop these 3 major mathematical and biomedical prediction models to simulate the human body’s health system.  Along the way, I created an application software for patients to use on their iPhone/iPad or PC.  I was able to produce and store more than one million of “clean” data regarding my health and lifestyle in the cloud server from 2012 to 2016.

Although my findings were no different from other research and discussions that existed in the medical community, I hope the same conclusions are drawn from my personal quantitative data; therefore, it can provide confirmation and credibility for other patients to follow.

As I mentioned previously, I am presenting my personal health data from the past 5 years.  However, I am confident that my findings are highly applicable to many other Type 2 Diabetes (T2D) patients with cases of having glucose values in the range of 90 to 400 mg/dL.  Although there are other patients’ data available in the cloud server through the use of my tool, but I have not spent enough time to analyze their data yet.  This will be a part of phase 2 of this project.

In August of 2010, when my physician informed me that both my A1C and ACR values were dangerously high, I was extremely scared and did not know what to do next.  I thought about my health condition long and hard and then I realized that there is only one person who can really help me.  That person is MYSELF.  For those cases that need medication, operation, or urgent care, a trained medical doctor is definitely needed for treatment and guidance.  However, in the case of chronic diseases, I did not get this disease overnight thus it cannot be cured overnight. The clear way to overcome these conditions is through “preventative medicine,” which requires a lifestyle change. Since I was diagnosed with diabetes, I knew already that there is no way I could cure my disease completely, but I can do my best to control it from getting worse.  From my personal journey, I found 3 fundamental problems associated with most diabetic patients:

(1) Lack of disease knowledge;

(2) Lack of a useful tool;

(3) Lack of willpower and persistence.

That is why it is difficult to change our lifestyle in order to control our chronic diseases.  By now, I have acquired sufficient knowledge regarding diabetes.  Through my research, I have also developed practical tools to control it on a daily basis. However, I am still puzzled about how to influence others to change their lifestyles and behaviors.  I am currently studying this problem via “Social Physics”, i.e. using natural science, including mathematics, physics, computer science, and various engineering methods to address and alter human beings’ social-psychological behaviors.  I know this is another long and tough journey.  I founded eclaireMD Foundation with a goal to address certain diseases along with medical problems and to conduct nonprofit activities to help other patients worldwide.  My plan is to incorporate what I learn through “Social Physics” and incorporate it in future phases of this project.

 

Date:   11/1/2016  15:00

Section 10:   Acknowledgement

I would like to thank the following people:

First and foremost, I wish to express my appreciation to Professor Norman Jones, who is a very important person in my life.  Not only did he give me the opportunity to study at MIT, but he also trained me extensively on how to solve problems and conduct scientific research.

I would also like to thank Professor James Andrews.  He helped and supported me tremendously when I failed academically at the University of Iowa.  He believed in me and prepared me the undergraduate requirements on how to build my engineering foundation.

These two great professors helped me, so that I could one day help others.

Jamie M. Nuwer, MD in Palo Alto, California, is a bright, young physician who has compassion in helping her patients.  She also took me under her care while she worked at the Stanford Network in Menlo Park.  Her encouragement and support are deeply appreciated.

Neal Okamura, MD in San Ramon, California, was my primary physician for 20 years, from 1992 to 2012.  He was the one who warned me in 2010 regarding the severity of my diabetic condition, which triggered me to launch this project.

Jeffrey Guardino, MD and Kristine Sherman, MD of the Stanford Network in Menlo Park, California, have been my primary physicians since 2012.  During my many visits with them, we spoke in detail about how I used my tool to improve my health.  It was Dr. Guardino who encouraged me in 2015 to write this paper for others to read.

Lynne Bui, MD in San Jose, California, has been a trusted friend and medical advisor since 2010 when I started this project.

During my journey, I met two medical doctors in Taiwan, Jia-Lei Loo, MD and Chi-Jen Tseng, MD.  Not only are they my medical advisors, but they are also my social network friends.  We constantly chat about many health-related issues online.  I am grateful to them for their support and encouragement.

My deep appreciation goes to both James Ratcliff, MD and JoEllen VanZander, MD of the Stanford Network in Menlo Park, California, for their care and encouragement.

I also would like to express my appreciation to both Steven Bhimji, MD and Patricia Hsiao, MD for their early participation of the “knowledge” development for this project.

I want to thank Gay Winterringer, who has a PhD in nutrition and food science, for her impressive input and knowledge.

I have known Dennis Heller for many years while working together and we became good friends after my retirement from my business career.  I would like to thank him for his time and feedback throughout the project.

I would like to extend my appreciation to my dear old friend from the MIT days, Dr. Toyohiko Muraki, for his input on my development of the metabolism model, using multi-dimensional nonlinear engineering modeling technique, and many follow-up discussions regarding my research work.

My appreciation also extends to my friend, Professor Toru Nakura at Department of Electrical and Electronic Engineering at Tokyo University in Japan, for his valuable inputs and discussions regarding my mathematical and statistical analysis results interpretation.

My deep appreciation also goes to Janet Kwan, RN for her devoted contribution and whole-hearted support to this project since she joined in early 2013.  I have discussed and exchanged numerous ideas regarding diabetes disease and its care with her.

Last, but not least, I would like to thank my wonderful wife, Li Li, who is a diabetic patient herself and has participated in my program for over 3 years.  Her participation and daily support of my efforts provided additional data points different from mine, resulting in many insightful revelations.

 

 

 

Read More

Using Quantitative Medicine to Control Diabetes (Simplified Chinese)

Posted by on Feb 22, 2017 in News | 0 comments

撰写时间:2017年02月17 – 2017年2月20日

论文(2016年12月2日):
摘要:使用量化医学来控管第二型糖尿病
发表者:美国eclaireMD基金会
日期:2016年11月28日

背景资料
我长期饱受第二型糖尿病、高脂血症和高血压等新陈代谢慢性疾病的折磨已达20年之久。如下是我自2010年开始研究并记录的当时的健康数据:
体重:205磅
腰围:44英寸
一次性血糖突击检查:350毫克/分升
90天平均血糖:280毫克/分升
糖化血红蛋白(A1C):10.0%
尿蛋白检验値 (ACR):116毫克/毫摩尔
甘油三酯:1161毫克/分升

研究目标
我在2010至2016年间一共花了七年的时间来研究糖尿病,最终发现了有效的控制糖尿病病情的方法。并针对第二型糖尿病一共开发出了三种数学模式和各种应用软件工具来控制这些慢性疾病。在此过程中,我大约收集了一百万条“纯或净”的健康数据,并采用多项科技方法,利用所学的不同的学科理论,其中包括了:高等数学、计算机科学(如:数据库、大数据分析、云计算、移动科技)、非线性和动态的塑性工程理论、人工智能或”Artificial Intelligence (AI)”(数据自动化与机械化学习)来模拟人体器官生物医学的新陈代谢系统。

应用方法
我所开发的体重和血糖这两种预测模型的应用,主要是提醒和帮助患者能够及早和有效地改变自己的生活方式。 血糖预测模型包括了:糖尿病药物干预、食物和营养的质与量、就餐地点的差异(居家或餐厅)、运动量、各种压力和创伤,居住地的气候环境,旅游的品质及频率,内脏器官的衰竭度、延时对血糖测量的影响。体重预测模型则包括了:食物的分量、运动的类型与强度、卡路里的变化、睡眠的各种质量、饮水量、及其他有关联的因素。
对于我本人而言,最明显的成效是自2013年起逐渐減药,2015年至2016年这两年间,我己不再服用任何糖尿病药物,还可以将糖化血红蛋白值(A1C)保持在6.2%至6.6%的范围内,同时还有效的控制住了高血压和高血脂。请详见如下的2016年底时期的健康数据:
体重:172磅
腰围:32英寸
90天平均血糖:115毫克/分升
糖化血红蛋白(A1C):6.3 – 6.5%(未服用任何药物)
尿蛋白检验値(ACR):12.6毫克/毫摩尔

结果与讨论
作为预防医学的一部分,这整个研发工作是基于生活方式管理、有关的各种信息收集、大量的健康与疾病的数据处理和分析。在60多个各型各类的图形图表中,每个图中都包含至少数百至数千个数据,并将这些研究成果呈现在您面前。此论文中的输入与输出数据(I/O Data)的11类大范围 (Big Category)有着密不可分的互动关联性(Inter-Connectivity),这包括了四个健康输出的大范围(体重和腰围、血糖、血压、血脂),六个健康输入的大范围(食品和饮食、运动、压力、睡眠、饮水、生活的规律性)和时效(Time)。这11类大范围是由500多个比较细微的元素(Element)所组成,它们通过在智能手机上面安裝的APP, 或是在个人电脑上安装的特定软件,对病患每日每时的生活、健康、行为、疾病状况进行全面性与高精确度的监控。系統內部的人工智能大约能够自动管理95%的元素,剩下的5%(大约20-25个)的元素,则需由患者每日主动输入系统中。结论是: 我过去5年间的健康代谢指标和预测值都能与医院体检的结果几乎达到了一致,而且我的体重和血糖预测的精确度也分别达到了99.9%和99.0%。
我的研究结论基本上能与医学界所提供的常见认知与常闻结论达到高度契合。此项研究完全以实验数据为依据,不带任何个人偏见或者内含个人的预设立场。因此,我希望经由科学化的大数据和量化分析的方法得以再次验证的医学理论能够得到进一步的推广和病患们的支持。
目前我们正在进行的科研项目的第2阶段,包括有如下的內容:
(1)继续收集、分析、改进并且公布全球第二型糖尿病患者的大数据资料;
(2)研究开发更有效的方法来影响或改变患者的健康生活行为,以便能接受并且适应更好的生活方式管理。

论文:使用量化医学来控管第二型糖尿病
发表者:美国eclaireMD基金会
日期:2016年10月20日 16:00

第1节:介绍
作为本文作者,过去7年间的研究数据,都基于本人长期患糖尿病期间积累而来,接下来,我将以第一人称来描述我的病情。
我是一名企业家,1995年在一家高科技上市公司担任CEO。由于工作原因,我一直生活在一个强高压的状态与环境之下。1998年验血时,我的血糖达到350毫克/分升。1998至2000年期间,我经历过几次严重的心绞疼与低血糖(俗称“胰岛素休克”)。下图总结了我自2000至2010年的体检数据。值得注意的是,本人最高的甘油三酯水平是1161毫克/分升,糖化血红蛋白(A1C)是10.0%,尿蛋白检验値(ACR)是116 毫克/毫摩尔。2010年8月3日医生建议我马上开始使用胰岛素注射,并告知有可能在3年之内要进行肾透析。接下来的3个月里,我搬到了一个离群索居没有压力的沙漠城市里,並且关闭了我所有的营利企业公司,开启了我的自救生命之旅。在过去7年的时间里,针对6种慢性病从阅读学习进而研究开发,同时我也彻底地改变了自己的生活方式,最后控制住了病情。这段日子里,我体重减轻了26磅(从198磅降到172磅)。腰围减少了12英寸(从44英寸减到32英寸)。2016年9月1日,我的体检数据有了明显的改善:尿蛋白检验値(ACR)12.6毫克/毫摩尔,糖化血红蛋白(A1C)6.6%(停止服用糖尿病药物超过一年多),甘油三酯水平67毫克/分升,高密度脂蛋白(HDL)48毫克/分升 ,低密度脂蛋白(LDL)103毫克/分升,总胆固醇156毫克/分升,体重指数(BMI)25。

图1-1:健康数据对比图
1
这是我过去15年来整体的糖尿病状况。2012年之前,我在医院化验的糖化血红蛋白(A1C)值曾经长期失控。然而2012年后,化验检测糖化血红蛋白(A1C)值约在6.6,这与我应用我自己开发的数学模型模拟创造出来的糖化血红蛋白值(A1C)曲线高度吻合。还有一点值得注意的是,在此期间我已逐渐减少药物,直到2015年我可以不再服用任何治疗或控制糖尿病的药物。

图1-2:2000至2016年糖化血红蛋白(A1C)记录表

日期:2016年10月21日 17:00
第2节:方法
2010年下半年,我意识到尽管自己在美国接受过良好的高等教育(17年来在不同的大学里学习了7个不同的专业),但对于慢性疾病却还是一无所知。 因此,我决定潜心研究如何控制我的慢性疾病和改善我的健康状况。在2010至2011最初的两年,我学习研究了内科医学里的6种慢性疾病:糖尿病、高血压、高脂血症、心脏病、中风和肥胖症。 2012至2013年间我又将注意力转移到食品科学和食物营养学方面的学习研究。 最后在2014至2016年这三年间,我专注于相关科技的研究和有效控管工具的开发。
这7年的学习和科研期间,我平均每周工作50小时,它所花费的功力与时间就如同攻读了一项博士学位一样。但是我的目标却只是为了彻底的解决我自己的健康问题与拯救自己的生命。最初我本想从传统的生物医学研究方法着手,也就是说从最基本的细胞层开始(Micro-Level)着手做基础研究,但是我没有足够的资金和专业的知识培训来完成这项长期的研究,所以我必须得结合自身已有的知识与经验,也就是在数学、计算机科学和各种工程学科的优势。我以前从未涉及过生物医学领域,也未曾受过该专业的正规培训;因此,我决定用宏观(Macro-Level)的角度着手,我采用了非线性塑性动力学的工程模型、有限元素工程分析的概念,以及无机材料的特性,来模拟人体的有机细胞材料与内脏的新陈代谢系统。我使用了高等数学来开发构建这个模型。我设定了体重、血糖、脂肪、和血压,这4个相互关联的身体输出大范围,还有: 食物、运动、压力、睡眠、饮水和生活方式,这6个輸入的大范围。在这10项大范围里面又包含了几百个细节元素。比如:仅“压力”类一项就分为“正常人”和“非正常人”(如”人格障礙病患者”)两大类型,而且它们的内部还设定了33个不同的压力元素。 我还加入“时间”作为第11项大范围,因为人体健康特征会随着时间而演变,即“动态”的观念。我经过不断的摸索和实验,也应用了人工智能(AI)来模拟人体的有机特性。在模拟的过程中,由于其复杂性和数据收集的困难度,我排除了所有的特殊环境因素,譬如:污染、辐射、有毒化学品、毒品、激素治疗、病毒感染等。这些因素对癌症尤其起到了至关重要的作用(其实癌症也是一种慢性疾病)。请大家注意,我的研究工作主要是集中在预防医学领域,因此,有别于大多数的糖尿病研究论文,在我的这篇论文里面, 药物仅仅作为血糖预测工具的一部分,而不特别去强调分析它。
鉴于红血球细胞的平均寿命是3-4个月,其携带的血糖和脂类遍及人体,我将3个月内收集的数据作为数控方程式的原始状态(Initial Condition)。因此,要求患者务必在使用我的工具时,尽可能详尽的收集最初3个月的数据。有了这个先决原始状态条件,才可以开始“解方程式”(数学意义上的解释),然后系统再通过使用人工智能(AI)来做进一步的机械学习和自动分析。2012年1月1日,我开始收集自己的健康数据,并使用该数据不断测试和改进系统内的数学体系。2014至2015年,我通过了使用人工智能(AI)开始构建“有机”生物医学数学模型,制作出两个实用性极强的预测模型,一个是预测体重模型(夜间休息后),另一个是预测餐后血糖值模型(餐后2小时)。到目前为止,在我们基金会的云服务器里面,我已收集并处理了近100万条“纯或净”的我个人的健康与疾病的数据。如果没有人工智能(AI)系统的分析能力,人脑是无法处理如此庞大的数据的。如图2-1和2-2所示,我的体重预测精准度已达到99.9%(2015年4月11日至2016年10月21日),血糖值预测精准度达到99.0%(2015年6月1日至 2016年10月21日)。
使用我开发的App工具,不仅可以轻松管理云端的大量健康数据,还可以预测体重和血糖的健康状况,并通过新陈代谢指标(MI:Metabolism Index)和综合身体健康状态单位(GHSU: General Health Status Unit)来监测整体的健康状况,这两个专有名词是我自己创立的,新陈代谢指标(MI:Metabolism Index)是聚合了11个大范围和500个元素在每日每时不停地计算出来的,而综合身体健康状态单位(GHSU) 被定义为每天的平均新陈代谢指标(MI)值的三个月的运行平均值。 图表是2012-2016年间我的新陈代谢指标(MI)和总体身体状况单位(GHSU)(它们已经达到了73.5%的健康水平分界线),很显然,通过这一套的自动化健康管理,我的整体健康状况有了明显的改善(到目前2016年10月底为止,我的代谢指标(MI)和综合身体健康状态单位(GHSU) 在57%左右,非常健康)。
图2-1:预测和实际体重(2015年4月11日-2016年10月21日)

图2-2:预测和实际每日平均血糖(2015年6月1日-2016年10月21日)

图2-3:新陈代谢指数(MI)和综合身体健康状态单位(GHSU)

日期:2016年10月22日– 2016年10月23日
第3节:体重和糖尿病
众所周知,要减肥和维持理想的体重是非常困难的一件事,2000年时我的体重是205磅,2010年是198磅,2012年是194磅,2015至2016年间则降到了172磅。 从2013到2014年间我的体重在180磅左右上下波动。 由于我经常去各国出差,所以无法有效的管理好自己的饮食生活和健康状况。 2014年我开发了新陈代谢的数学模型以后,我就开始使用该工具来管理我的整体生活方式,同时尽量减少长途飞行的次数。 结合我在2015年4月11日开发的体重预测模型,自身减重取得了显著成效。详见减肥图3-1。
图 3-1:2012-2016年体重

有经验的人都知道要减少腰围比降低体重更难。 2000年到2014年,我的腰围在42到44英寸之间波动。 直到2015年中,我开始观察自身的新陈代谢指数,并且使用了体重预测工具,终于将我的腰围成功地缩减到32英寸。详见腰围控制图3-2。
图 3-2: 2012-2016 的腰围

影响体重的因素很多, 饮食和运动是控制体重的最有效方式,当然其他的因素(譬如:睡眠的品质与压力)也会影响体重。食物种类的质和慢性疾病的类型(下一节讨论)有很强的相关性,参见图3-3和3-4,显而易见食物的量与体重则有很强的关联性。例如,从2015年6月至2016年10月,我的体重是172磅。而我的食量大约是我正常食量的91%。假如我想将我的体重再减到170磅左右,那么我就必须将食物的量再度减少到正常食量的80%。
图 3-3:食物摄入量 (2015年6月1日-2016年10月20日)

图 3-4:清晨体重 (2015年6月1日-2016年2016年10月20日)

现在让我们分析一下体重和血糖之间的关系。 除遗传因素外,超重(Overweight)与肥胖(Obesity)是第二型糖尿病(T2D)的主因。当我分析体重与血糖时,我标注出体重> 180磅,日均血糖> 160 mg/dL,从图表中可见两者之间有很强的相关性。请参见下图3-5和3-6。若是在就餐以前就能预知你的餐后血糖是多少,你就可以选择好餐食的质和量,如此一来,你就可以同步控管你的体重与血糖了,(我的餐后血糖预测功能工具就是为了能有效的控制住餐后血糖, 但是同时也能帮助体重的控管)。

图 3-5:体重超过180磅

图 3-6 :日均血糖超过160 mg/dL

体重在早晚都会有差异,饮食和运动是其主要的影响因素。日间摄入食物会增加体重,但运动后会燃烧脂肪。 夜晚通过蒸发、排尿和排便等,这三个管道可以减轻体重。去年,我的睡前体重和晨起体重的对比增加值在2.8磅。 夜间休息后到次日晨起体重则也同时减轻了2.8磅。这就是为什么我的体重能够一直保持在172磅(+/- 5磅)左右的原因。次日晨起预测功能工具对控制体重起着至关重要的作用,而控制住体重就是控制住糖尿病的关键。图3-7和3-8呈现了我本人两种截然不同的体重变化。
图3-7:同日睡前和晨起体重增加

图3-8:睡前和次日晨起体重减轻

日期:2016年10月24日 13:00
第4节:血糖
2010年8月3日,我的体检报告显示糖化血红蛋白(A1C)值为10.0%,尿蛋白检验值(ACR)为116 mg/mmol,甘油三酯为1161 mg/dL。医生警告我要立即改变我的生活方式,否则我将必须得做肾透析了! 因此,我立即清醒过来,下定决定洗心革面。2012年1月1日我开始收集自己的健康和生活方式数据。迄今为止,我已收集了约5年的完整数据。 在我的云数据库中,我也存储了大约一百万条身体和生活方式的数据,其中包括:原始输入数据、系统计算出的数据丶和人工智能(AI)处理过的数据。图4-1中:呈现的是每日和90天的平均血糖,我的血糖值变化区间在86mg/dL和227mg/dL,平均值为129mg/dL。图4-2中:体检报告显示的糖化血红蛋白(A1C)值变化区间在6.3%和7.1%,平均值为6.6%。我的数学模拟糖化血红蛋白(A1C)值(6.3%和8.1%之间,平均值为7.1%)准确率达到92.4%。
糖化血红蛋白(A1C)体检结果反映了过去90天的平均血糖值。我的医学研究公益基金会(eclaireMD)把iPhone平台上的Wellness App开发出来,这个软件基于用户每日输入的血糖数据,可以预估出一个数学模拟的糖化血红蛋白(A1C)值。 当然,这种数学模拟糖化血红蛋白(A1C)值的应用工具相当于一个内置的人工智能工具,它可以根据用户生物医学系统参数的变化而自动地进行调整和计算。
通过输入体检结果进行频繁校准可以提高数学模拟A1C值的精确度。自2012到2016年,数学计算的糖化血红蛋白(A1C)曲线已跟实际测试的糖化血红蛋白(A1C)数据完全吻合。这意味着,我的模拟糖化血红蛋白(A1C)值应用工具,在得到体检报告结论前,就起到了上线预警的作用。
图4-3 呈现了我所有的实验测试的糖化血红蛋白(A1C)值。尽管我己经考虑进去了红血球细胞的寿命,而加入不同的线性和非线性计算的数学模型,但鉴于高度非线性的生物医学体系统会随着时间而变化,因此有时候血糖仍会有不可预测的输出值。而直接或间接能够影响血糖值的因素又有很多,接下来我会跟大家讨论更多我的研究成果。
2012至2013年间,我曾服用三种糖尿病药物,如:捷诺维(Januvia)100毫克,艾可拓吡格列酮(Actoplus Met) 15毫克/850毫克,二甲双胍(Metformin)2000毫克。2012至2016年期间,我的总糖化血红蛋白(A1C)曲线保持在一个稳定状态,即6.6左右,2014年初,我开始逐渐减少服药量,到了2015年的时候,我就已经成功地戒断,不再服用任何糖尿病的药物了。这也可以说是我控制住糖尿病最主要的成就之一。
在此期间内,我遭遇了不同程度的“戒断症状”(Withdraw Effect)。在停药、减少药量的一个月内,我的血糖图跌宕起伏,也找不到任何明确的线索和合理的解释,我曾经多次因为恐惧此举会伤害自己的身体而又重新回去服药,但是在反复尝试了一年多以后,最终我用意志力坚持了一个多月的时间,我的血糖折线图终于又回归于平稳状态(Steady State),所以从2015年初以后,我就不再服用任何糖尿病的药物了。
图4-1:日血糖和90天平均血糖(2012-2016年)

图4-2:数学模拟A1C和实验室测试A1C对比图

图4-3:实验室检测A1C值和数学模拟A1C值对比图

在我的数据收集和分析过程中,我注意到测试数据和模拟数据之间存在有一些偏差。当然,我初试时,曾经试图找出影响血糖和糖化血红蛋白(A1C)的主要元素。其中的一部分我将在下面的章节中讨论。2015年4月间,我对“黎明现象”(Dawn Effect高空腹血糖)和空腹血糖(Low Fasting Glucose低空腹血糖)之间存在的差异产生兴趣。尽管我对肝脏和胰腺是如何制造和控制血糖这方面一无所知,但是我决定通过阅读大量的医学论文和分析我自己的大数据来研究空腹血糖。目前为止,我收集了400多个早晨(早餐前)的空腹血糖(FPG)数据。由于空腹血糖(FPG)预测与餐后血糖(PPG)预测不尽相同,而且,一般而言,在线性计算里,空腹血糖(FPG)的线性权重约为日平均血糖数据的25%,因此我决定使用90天平均日血糖作为空腹血糖(FPG)的初始条件。图4-4:空腹血糖(FPG)研究结果显示我的初步发现是基于360天的数据值。这表明,尽管每日空腹血糖(FPG)上下跌宕起伏难以预测,但经过一段时间后,平均空腹血糖(FPG)总是能稳定在90天平均血糖值的附近。预测和实际之间的偏差仅为1.6%,我的预测准确性达到98.4%。

图4-4:空腹血糖(FPG)研究结果

我患糖尿病长达20年之久,并意识到控制这种疾病必须具备三个先决条件:实用的知识、可靠的工具和坚强的意志力与持久力。我运用自学的内分泌科医学知识,开发出了自己的App应用软件工具来控制我的体重和血糖。2015年4月11日我开发了体重预测工具,同年6月1日开发了血糖预测工具。值得注意的是,2014年我所开发的新陈代谢模型(MI和GHSU)的主要目的是为了改善自身的整体健康状况,再加上2个预测系统,通过调整食量、运动频率和运动强度来预测自己的体重和血糖就变得更得心应手了。换句话说,如果能适当调整输入参数,输出值就可能会自动调整至最佳状态。还有一点应注意,身体是有机体(非线性)和动态体(随时间变化),所以我们必须不断监测这种变化的迹象,也必须使用人工智能来不断地调整数学控制方程式(Governing Equation)。在使用这两个预测工具时必须遵循下面两个基本规则:首先,(1) 我必须遵循预测模型的建议,尽可能输入精确的输入值。其次,(2) 测量体重或血糖后,不能再重新调整已经输入了的数值,或者回头手动输入作与输出值吻合的调整,因为这样做会影响到系统预测的准确性(除非从预测经验中获得某些特殊认知或意识到某些新的事实和发现)。人工智能(AI)在系统中已达到一定的级别,但整个生物医学系统仍需要在应用中不断观察和调整。如果有一天即使我含笑九泉,而我所创立的非营利性医学研究基金会,仍可以继续深入地研究这些相关的课题,并且持续地改善我们的工具。
2016年7月1日,在我的日记中记录如下:
“2015年6月1日至2016年6月30日,通过13个月的实际数据分析,我研发出的用来预测A1C的慢性病工具的准确率已达到99.2%,90天的平均血糖值预测准确率已达到了98.4%。因此从2016年7月1日开始,我不再使用血糖仪验血(试纸法)的方法来测量血糖,而是用自己开发的数学模型取而代之,来预测血糖以控制自身的糖尿病,为了确保糖尿病病情得以控制,我还于2016年10月1日亲自去医院做了一次A1C检查。接下来的3个月,我非常注意自己的饮食和生活,每天坚持餐后运动。当然,还有几个其他主要因素,如果这个测试最终成功,那么我发明的数学生物医学模型对全球的糖尿病患者而言,不仅降低治疗成本,还减少了疾病测试所带来的痛苦“。
从那天起,我决定减少使用传统血糖仪(手指穿刺和血糖试纸法)的检测量。但是,我仍需使用血糖仪继续测试空腹血糖,因为我正在进一步研究因空腹血糖(FPG)和餐后血糖(PPG)权重因子所引起的A1C的变化。每当我去不能提供营养数据的个体餐厅就餐时,因为不能完全确定就餐食物中的淀粉和糖的分量,我就会使用手指穿刺法测量餐后血糖。2016年7月1日至2016年10月23日期间,115天内,我一共收集了460笔血糖数据,115笔空腹血糖(FPG)数据(总数的25%),另外115笔餐后血糖(PPG)数据(总餐后血糖的33%和总数据的25%)作为实测血糖数据。剩余的230笔数据(总数据的50%)全部来自于我的预测血糖值。该实验所得的初步结论是,有一半的情况下我不使用手指穿刺法,然而我血糖预测结果仍能达到99.6%的准确率。为了证实我的发现,我打算在进行一年的实验。如果此推论能得到证实,那说明我的预测方法是行得通的,大多数第二型糖尿病(T2D)患者平均日血糖水平在100 mg/dL和400 mg/dL之间,他们能使用我的预测工具来控制糖尿病。我希望能用一个简便的工具,以省去患者使用血糖监测仪和试纸的负担和成本,让患者采用更便捷更无痛的方式,而也能更好地控制住糖尿病。请参见图4-5:50%的预测血糖数据(2016年7月1日-2016年10月23日)。
在东亚、欧洲和美国,约9-10%的人有糖尿病。 中国与东亚国家约有10%,拉丁美洲约有8.5%,而非洲约有5%的人口患有糖尿病。 特别是发展中国家的国民收入有限,每日频繁的血糖测试对一般家庭来说是个经济负担,更不用说更昂贵的糖化血红蛋白(A1C)检查了。而我的预测工具对这些糖尿病患者来说,无疑提供了很大的帮助。 当然,苹果设备对大多数人而言不能触手可得,而且大多数糖尿病患者又趋于老龄化,他们可能难于掌握现代科技和苹果手机应用(App)。因此,我的eclaireMD慈善医学研究基金会也开发了一款面向发展中国家和老龄患者的应用工具,那就是台式电脑(PC)和网络工具,该工具并与云服务器同步进行大数据运算。
我的体重和血糖预测工具是一款免费软件,可通过苹果APP Store下载。或使用搜索工具栏输入关键字“eclaireMD”,在关键字下搜索“Wellness”产品。

图4-5:当患者不再使用手指穿刺和试纸发方法时,初步数据分析可给出血糖预测值的可靠性和准确性分析。 这些基础数据有50%是基于血糖预测值(2016年7月1日-2016年10月23日)

日期:2016年10月25日 13:00
第5节:血糖和食物
为了研究血糖和食物之间的关系,我开发了一款苹果手机应用软件SmartPhoto。在该应用中,我构建了一个关系型数据库结构,以便将每张照片储存在iPhone相册中。其中数据结构分为5层, 我同时列举一些范例:
1.分组:(美国、日本、法国等)
2.分类:(居家料理、连锁餐厅、个体餐厅、航空公司、邮轮等)
3.文件:(丹尼斯餐厅,麦当劳,希腊餐厅,亚洲食物等)
4.名称:(餐厅名称、菜名、菜单项等)
5.内容:(存储记录,如:营养成分等)
一旦食物照片存储到SmartPhoto应用数据库中,您可以按照自己的喜好方式进行存储分类和搜索。请参见图5-1:SmartPhoto应用中每张食物照片附有血糖值。2015年5月1日至2016年10月20日,我收集了1591张食物和就餐照片,它们的平均餐后血糖水平(PPG)为121.8 mg/dL。同期,我的日平均血糖水平(包括空腹血糖FPG)为121.41 mg/dL。

图5-1:SmartPhoto软件应用中的食物样图

2012至2014年根据我的血糖分析,得出一个初步结论: 高血糖期(接近140 mg/dL)均处于在海外旅行阶段。 请参见图5-2:2012至2014年的血糖测定结果。我发现东亚(不包括中国华北地区)、夏威夷和塔希提岛大多数菜肴含糖量较高,而碳水化合物主要来源于米饭、面粉、芋头等, 但是我在2016年以后在东亚国家和夏威夷逗留超过8个月之后,我发现到后期的平均血糖水平下降到120 mg/dL以下,比前期下降了20点。 请参见图5-3:2015至2016年血糖测定结果。
四个主因可降低平均血糖值:
(1)当我使用血糖预测工具时,会更加严谨的选择食材和餐厅菜单里面的食物;
(2)每日三餐后健走,平均4000步/次,而不是将运动集中在晚上(有关细节详见如下内容);
(3)旅行期间我会对食物的摄入量和健走运动倍加小心谨慎。 如:在机场餐厅或贵宾室就餐后,我会在登机口过道附近健走3000到4000步;
(4)SmartPhoto工具的分析功能可提供许多帮助,比如:就餐地、菜单和烹饪食材的选择。

图5-2:2012-2014年期间血糖

图5-3:2015-2016年期间血糖

请参见图5-4,根据SmartPhoto中的大量照片数据,将不同就餐地所得的平均血糖做了统计和汇总。1591份食物和就餐图片的餐后血糖水平(PPG)为121.8 mg/dL。2015年5月1日到2016年10月20日同期,我开发的APP工具呈现的平均日血糖,包括空腹血糖(FPG)为121.41 mg/dL,这就是为什么我决定采用平均日血糖值作为初始预测空腹血糖(FPG)值的原因 ,我的90天平均血糖为123.75 mg/dL,如图5-5:SmartPhoto中2015年5月1日至2016年10月20期间的血糖。

图5-4:平均血糖和不同就餐地汇总表

图5-5:SmartPhoto中2015年5月1日至2016年10月20日期间的血糖

我的初步研究心得汇总如下:
(1)所研究国家的平均血糖值相似,测量值介于119.9和125.6 mg/dL之间。2015年到2016年,不论外出就餐,还是在家用餐,我都严格控制食物摄入量的比例。
(2)在家用餐的血糖值为115.3mg/dL,连锁餐厅(有营养数据的餐厅)用餐血糖值为125.2mg/dL;个体餐厅(无营养数据)用餐血糖值为132.3mg/dL;机场、航空公司休息室、飞机上用餐血糖值为134.0 mg/dL;食用超市购买的速食品血糖值为140.6 mg/dL。
(3)航空公司提供的食品会让血糖升高,因为机上食物可选性有限,又无餐后运动的空间。
(4)学习和研究了主要食材的营养成分后,我尽量不食用加工类食品。除非无选择时,不得不吃,但食用之前我会仔细阅读标签上的营养成分(特别是碳水化合物和糖的含量信息)。
(5)个体餐厅经分析所得的平均血糖值如下:
美国:129.8 mg/dL
日本:139.6 mg/dL
台湾:136.7 mg/dL
其他国家(包括中国):130.8 mg/dL

图5-6:不同就餐地的平均血糖值

总的来说,美国和西方食物在烹饪过程中不放糖(除甜点外)。 日本、韩国、中国长江以南和东南亚地区(从越南一直延伸到新加坡)在烹饪过程中都会添加糖和盐。
(6)我在分析某一知名品牌的连锁餐厅时发现了一个有趣的现象。通常情况下,我不在任何一家连锁餐厅吃午餐或晚餐,但是早餐例外,因为它们为了薄利多销,就必须把早餐的份量做小,再加上我再把它们所提供的含碳水化合物的面包类减半进食,这样我就可以控制好我的早餐后血糖值了。我在同一品牌的美国连锁餐厅用餐的平均血糖值为122.9mg/dL,在日本则为117.4mg/dL(日本是一个例外,因为他们非常注重并且服从总公司颁布的烹饪标准作业程序(SOP),但是到了台湾则变成125.3mg/dL,中国则变成126.2mg/dL。我的观察结果是:在中国和台湾虽然是同一个品牌,但是,它们到了当地以后,食材多少都会发生变化。而且,我怀疑其采购方式和烹饪的标准作业程序(SOP)都有可能不完全与总部的要求一致。
(7)2013至2014年间,我学习食材营养成分过程中,曾得出了一个错误的观点,那就是: 可以”尽量多吃蔬菜”。但是在2015年,当我收集了六百万个不同食材的各种营养数据并加以研读后,我发现各种蔬菜的营养成分其实都有差异。有一天忽然间我的灵感闪现,我通过检查不同颜色的蔬菜,来辨别它的碳水化合物和糖分的含量,得出了一个便捷的蔬菜与血糖关系的预估方式,那就是使用蔬菜的颜色来分类。请参见图5-7:蔬菜中碳水化合物和糖含量汇总。我目前的结论是:假如我吃很多蔬菜,餐后血糖水平(PPG)仍有可能会升高。因此我必须注意选择蔬菜的颜色以及食用的总量,这样才能获得更准确的血糖预测值。
(8)当我想吃零食、甜点或水果时,我会选择恰当的时间,并且限量,以控制血糖和体重。最佳的方法就是在两餐之间少量食用,例如:上午10点或下午3点。为了要控制体重,睡前尽量少吃。水果对身体健康有益,但要避免高糖份的水果(如:菠萝,香蕉等),而且一定要限量。有了这种控制机制,血糖就会健康无比。

图5-7:蔬菜中碳水化合物和糖含量汇总

我也仔细观察并分析记录下“极端”情况。 特别是血糖超过200 mg/dL 时。 请参见图5-8 是我17次的用餐情况记录,2015年5月1日至2016年10月20日血糖值超过200 mg/dL。
值得一提的是,有3个主要因素会让我的餐后血糖水平(PPG)升高,第一是在个体独立餐厅内食用东亚食物,第二是在美国连锁餐厅吃午餐或晚餐,第三是在飞机或游轮上进餐。 如果我能事先获知这些食物的营养成分,然后使用正确的工具来预测餐后血糖值,再结合自身的意志力与持久力,那么在这种地方就餐仍然是会安全的。

图5-8:2015年5月1日至2016年10月20日的17次用餐记录显示餐后血糖水平超过200 mg/dL

如下图所示:图5-9:分析血糖值大于140的原因,显而易见高碳水化合物、高糖食物、亚洲食物等三项原因是将血糖值增加了约58%(> 140 mg/dL)。 另一个发现是,仍有10%的未知因素,无法解释为何会引发高血糖。

图5-9:分析血糖值大于140的原因

经由大数据的研究可以很容易地发现碳水化合物、糖和血糖之间的关系。使用以下的几个方法,可以很容易地通过控制食物的质量和数量来预测餐后血糖值。
(1)我用食品包装袋上所提供的营养成分和配料成分的克数除以20来估算。例如:碳水化合物16克,16÷20 = 0.8,并将0.8输入到carbs栏内。同样地,糖分也是这样来估算。
(2) 在家烹饪时,我会以手掌的面积和厚度,或是拳头的体积大小,来当作是100%的量用来估算判断淀粉分和糖的量。但依我过去几年(得糖尿病的15年后)的估算、判断、比较,我最近注意到,我必须以我手掌或拳头大小的三分之二或者二分之一来当作是100% 的量来估算判断,我猜测可能是因为受到糖尿病的长期影响,自身内部器官对于碳水化合物(淀粉)和糖分的耐受力已大为降低。在收集更多关于这种现象的大数据之后,我或许需要构建另外一个层次的人工智能来解决这种身体内部器官的质性变化。
(3)我的应用工具可从食材库中搜索每项食材的营养成分,然后将它们相加,以获得碳水化合物和糖的总消耗量。
(4)大多数水果都含碳水化合物和糖分,但有一些水果,如:香蕉、菠萝和葡萄它们的碳水化合物和糖含量更高。
(5)在用餐时,请一定不要食用点心类甜食,因为它们含有高碳水化合物、高糖、高盐和高油脂,这些对身体的健康都是很不利的。多吃绿叶蔬菜,非绿色蔬菜尽量少吃,如:茄子,糖含量高的甜菜、胡萝卜、玉米、洋葱和番茄。
对糖尿病患者而言,最重要的原则是要在一天中保持“均衡”的血糖值,也就是说:要随时降低高血糖,及随时提高低血糖(避免低血糖休克)。当你能保持住目标体重,营养摄入均衡,自身的糖尿病和其他慢性疾病就会得以控制。

日期:2016年10月28日 13:00
第6节:血糖和其他因素(运动、压力、旅行、温度)
血糖和运动
除食物和饮食外,运动也是至关重要的,它有助于降低血糖。我的APP应用软件中涵盖了各式各样不同类型的运动。但是据我自己多年来的经验,对老年人而言,保持匀速健走是最好的锻炼方法。我的健走均速是2.5英里/小时,约步行6000步/小时,或100步/分钟。
2012年,我平均每天健走8000步,即3.3英里。那段时间,因为体重是194磅,因超重而导致膝盖负担太重,我不能走太久。到2016年,我逐步增加到每天17200步,或每天7.2英里,到了那个时候,我的体重已经降到172磅,对我而言真是轻而易举。请参见图6-1,6-2和6-3。但是,如果持续走路太多,有时候会给膝部或者足底造成负担,所以最好再进行其他类型的运动,譬如太极拳,来辅助每天的运动量。
2015年初,我发现如果将健走分成三部分,即每餐后2小时内锻炼,餐后血糖会明显降低。通过长时间分析我的血糖数据,我还了解到,当平均血糖约140 mg/dL 时,如餐后健走1000步,血糖会降低7至10个点。因此,当我的平均血糖值下降到约120 mg/dL 时,餐后每1000步可将血糖降低4至6个点。这种差异是源于我的数学模型内部所出的初始状况的假设。

图6-1:健走记录(2012-2016年)

图6-2:晚上健走记录(2012-2014年)

图6-3:三餐后健走记录(2015-2016年)

血糖和压力
压力会导致很多健康问题。当人们经历长期、持续性的压力时,它会严重影响身心健康。在我长达三十年的职业生涯中,一直生活在高压状态下。正因为此原因,我分别在五个不同的时间和场合,经历过多次严重的心绞痛,并始入了第二型糖尿病,这最终也导致了我患上了慢性脚趾损伤、膀胱损伤和肾脏损伤。 然而在我从高压的企业负责人的位置退下来以后,我喜欢上了平静、无扰的生活方式(2014年除外)。 2014那一年我分别经历了三次生活事件的压力(3-6月、9-10月、11-12月)。 请参见图6-4:2014年3月至12月的压力期与2015年1月至2016年10月的平和期比较图。
图6-4:压力期(2014年3月至12月)与平和期(2015年1月至2016年10月)比较图

下图6-5、6-6、6-7、6-8和6-9,可见压力、血压、血糖值和A1C水平之间的明显相关性。
由于接连发生第二和第三次压力事件,图表可见压力评分、高血压和血糖与3-6月和9-12月的两个时间跨度的一致性。因A1C需要3-4个月的平均血糖值,所以A1C值的峰值约在3个月后体现出来。

图6-5:2014年压力评分

图6-6:压力期血压升高

图6-7:高血压和压力期的日高血糖

图6-8:高A1C峰值在3个月后的高血糖出现

图6-9:高压力分数和较高的90天平均血糖对比图

2015年我受了两次外伤。第一次发生在6月23日,我跌倒在一个路边的斜坡上,结果面部受伤,进了急诊室。发生此事故后,三天之内的血糖值分别为152 mg/dL、208 mg/dL和154 mg/dL,第4天才恢复平稳约120 mg/dL。第二次事故发生在12月4日,我在一个建筑工地上受了腿伤,再次进了急诊室,在这次事故后,三天的血糖值分别为145 mg/dL、175 mg/dL、165 mg/dL,也是到第4天才恢复平稳约120 mg/dL。由于这两次突发事故,造成我的血糖水平暂时性升高。两次事故都是经过了4天,创伤状态平稳后,我的血糖才逐渐降低恢复平稳状态。2016年4月8日,我在楼梯上踏空,万幸身体没有受伤,但是就在我食用完低碳水化合物和低糖分的早餐两个小时后,餐后血糖值竟为148 mg/dL。我那天不仅是和平时吃的是一样的早餐,餐后还健走了4000步。虽然这是小事一桩,但很明显,压力的确会影响血糖的水平。

血糖和旅行
在我过往忙碌的企业生涯中,曾走遍世界各地。此处的分析我做了一些简化,只编译了2012至2016年的旅行记录。大于3.5个小时的飞行时间(再加上三到四个小时的进出机场)被我定义为长途旅行,它会影响两顿餐食的血糖。小于3.5个小时的飞行时间(再加上三到四个小时的进出机场)被定义为短途旅行,它只会对一顿餐食的血糖造成影响。过去5年中,平均每两周(更准确地说了每12.9天),我就要旅行一次。从我乘机旅行期间的健康状况分析来看,我注意到旅行期间血糖和新陈代谢的变化都很明显。它们受到的影响主要是来自于两个因素,第一是大多数的航空公司與相關部門所提供的食物對糖尿病患者来说都是不友善的,第二是在飞机上或者在机场里的运动空间非常有限。当我找出了这些原因以后,旅行时段内为了保持良好的血糖值,我会非常小心地选择对血糖安全性高的食物,而且尽可能用餐之后在机场内的通道上健走。因此,2015至2016年期间,凡空中旅行期间我的血糖和代谢指数都得到大大的改善,几乎接近正常水平120 mg/dL。请参见6-10图分析结果。

图6-10:血糖、新陈代谢和航空旅行的关联性

血糖和天气
我在美国成长,也在不同的州内生活了多年,我所居住的大部地区都是无污染、天气好、气候温和(气温15-25摄氏度)。 2016年的上半年,我在东亚地区逗留了6个月,跨越了冬季、春季和初夏。虽然我穿梭于不同的城市,但是我严格律己,尽量保持正常的生活方式,对食物、运动、压力、睡眠、饮水量和其他等项都进行监督。但我注意到,2至6月份,当亚洲的温度越来越高时,血糖值也会随之而上升。我无法解释原因,但我很想知道,炎热的天气是否会影响身体的新陈代谢。图6-11 可见一个短期间(约四个半月)的数据观察。我之所以写这个课题,是为了邀请其他研究人员的参与,以得到对这个课题广泛的关注和研究。

Please don’t 图6-11:血糖和气温的关联性

日期:2016年10月29日 15:00
第7节:高脂血症和高血压
2000年开始我就一直收集自己的体检数据,并将它们输入到自己的工具中。图7-1、7-2、7-3和7-4 是绘制的脂质图形。图中可见,2000至2012年我患有高脂血症。2012年我主要专注于控制自身的糖尿病,但是我的总方针是利用预防医学的理念来管理自我,使用最佳的生活方式管理自我,用更好、更有效的方法来控制各种慢性病。2014年我开发了一款以数学模拟代谢模型为生活方式管理的有效应用工具。 因此,最终不仅我的血糖得以控制,我的血脂质数据也变得健康了。 新陈代谢我们将在下一章节进一步探讨。

图7-1:甘油三酯(2000-2016年)

图7-2:高密度脂蛋白-胆固醇(2000-2016年)

图7-3:低密度脂蛋白-胆固醇(2000-2016年)

图7-4:总胆固醇(2000-2016年)

脂类与食物的“质”有着密不可分的关系。 为了饮食健康,我列出了一份食物质量提醒和记录清单以供参考。 请参见图7-5:食物质量提醒和记录。如果您能完全遵守此表内的规定,会得到0.5分。 如果您完全违反规定,则会得到1.5分。 请参阅图7-6:2014年中到2016年10月20日的食物质量评分。我的“食物质量”满意度为96% , 得分为0.54,满意度=(1.5- 0.54)/(1.5-0.5)。

图7-5:食物质量提醒和记录

图7-6:食物质量评分

我的血压数据如图7-7、7-8和7-9所示。 正如我前一章节提到的,自2014年3月到12月,本人经历了三次连续性的生活事件压力期,如下图所示,您可以看到它们对血压的影响。 我已经把“摄入低盐量”作为“食品质量”的要求之一,并在自己的日常饮食中严格律己。 从图7-10所示高血压病因的分析中,数据清晰显示了生活上的事件压力与高血压有着密不可分的关系,此外,还有海外旅行、时差、饭后运动时间、极端天气条件等。

图7-7:最高日收缩压(SBP)和日舒张压(DBP)

图7-8:平均日收缩压(SBP)和日舒张压(DBP)

图7-9:平均日心率

图7-10:高血压病因分析

日期:2016年11月1日 11:00
第8节:新陈代谢指数(MI)和 综合身体健康状态单位(GHSU)
2014年全年,我进行了研究并开发了以整体健康和慢性疾病为主题的产品。起初我尝试从定义“新陈代谢”入手,但是失败了。 譬如:在韦氏词典中定义“新陈代谢 = 生物体内物质和能量的自我更新过程(细胞或生物体中)”。我也曾尝试咨询了好几位医师,也无法得到一个准确清楚的定义。总之,这个名词对许多人来说,只有一个模糊的定性描述(Qualitative Description),而缺乏一个明确的定量描述(Quantitative Description)。后来,我尝试系统化的收集数据,再加以科学化的整理分析,中间还导出了一套AI监管的数字控制方程式(Governing Equation)去重新定义“新陈代谢”这个名词。
2014年我花了一整年的时间来研究这个领域,其间我还自创了一个新的术语,它被称为“新陈代谢指数”(MI: Metabolism Index)。它是基于四大类的人体健康日常输出数据,和六大类的人体健康日常输入数据,而这些数据都与慢性病有着密不可分的关系。四类日常输出数据包括:体重、血糖、血压和血脂。六类日常输入包括:运动、饮水、睡眠、压力、食物及日常生活作息。由于输入、输出和生物医学系统之间的互动是动态的,也就是说它们是会随音时间而变化的,所以我把“时间”作为第十一大类。
更细一点的定义:每个大分类都是由许多细微的元素所组成。譬如:睡眠有9个元素(睡前思绪万千、睡眠时间、醒来次数、晨间清新感觉、醒后头疼、梦境多寡、环境舒适度、睡眠期间生病或身不舒适、失眠的程度),压力有33个元素(并非所有因素都适合每个人),而食物类则有约100个元素。再加上系统运算推出与人工智能产生的新元素,最终会有大约五百个因素需加以组织、记录和分析。当然,每日有效地解决这个问题,对人脑而言,绝对是一个巨大的负担。在数学理论上,最大的挑战是如何解决这11个不同分类的大范围和五百个细微元素之间的互相连动性的问题(Inter-Connectivity)。因此,我利用结构工程的有限元素概念 (Finite Element Method)和动态塑性工程概念 (Dynamic Plastic Engineering Modeling)来模拟这个生物医学系统 (Bio-Medical system)。我建立了以人工智能为基础的多组具有各种边界条件(Boundary Conditions)的数学控制方程式(Governing Equation)。通过这个过程的努力,剩下的要解决的问题就是要应用计算机科学去开发实用的产品,特别是使用自动化计算和人工智能,这也就是大数据分析和云运算的用武之地了。
综合身体健康状态单位(GHSU: General Health Status Unit)是最近90天内代新陈谢指数(MI)的动态平均值。 最初,我采用了医学界的定义规范,也就是数值越低越好,我定义了新陈代谢指数在0.5(最佳条件)到1.5(最坏条件)的范围内。 当新陈代谢指数和综合身体健康状态单位都低于1.0时,表示健康状态良好。 但如果该值超过1.0,可能预示健康和生活方式上出问题了。就我自身而言,我在每个大分类范围中确定一组最佳元素,并且先确定自身要求达到的最佳健康状况(Targeted Health Status):体重为170磅、 血糖为120 mg/dL、收缩压和舒张压为120/80、 甘油三酯/高密度脂蛋白-胆固醇/低密度脂蛋白-胆固醇/总胆固醇为150/40/130/200。新陈代谢指数和综合身体健康状态单位的“均衡”水平实际上是在73.5%,也就是说:高于73.5%代表不健康,低于73.5%代表健康。 再提示一次,我使用一般医疗实际的做法,也就是用较低的数值来代表更好的健康。
在2016年10月20日这天,我的新陈代谢指数和综合身体健康状态单位分别为58.45%和57.1%,这表明在过去的3个月里我很健康。 我在过去2年内的各种医院的实验室测试结果,也证实了我身体非常健康。 通过对生活方式管理,应用量化医学来控制慢性病是一项有效的措施,我们也可以称之为“预防医学“的一部分。 请参阅8-1和8-2图中2012-2016年间及2015年4月11日至2016年10月20日期间我的综合身体健康状态单位和新陈代谢指数。

图8-1:代谢指数和综合健康状况单位(2012-2016年)

图8-2:代谢指数和综合健康状况单位(2015年4月11日-2016年10月20日)

做完新陈代谢指数和综合身体健康状态单位的基本概念介绍之后,让我们研究一些主要分类的评分数值。前面章节中我们已看到许多收集的数据汇总,如:体重、腰围、血糖、血压、血脂、食物、运动和压力。剩余的缺失部分也很重要,但与健康输出数据没有直接关联。本节我还得重申一下食物数据。图8-3列出了源自我的数学计算模型及其转化的“满意度”得出的汇总分类分数,而“满意度”一词一看就一目了然。
除了2014年间开发的新陈代谢模型,还有另外两个重大突破,一个是2015年4月11日开发的体重预测软件和2015年6月1日开发的血糖预测软件。这三个模型为我提供了巨大的帮助和准确的指导,以帮助我控制糖尿病。本节的数据和图形显示中,我选择了2015年4月11日至2016年10月20日之前的资料作为标准区间进行比较。

图8-3:代谢指数转换成满意度级别换算表

我的饮水分数是0.74,其满意度为95%(100%定义为每天饮用6瓶或3000毫升水)。 在此期间,我每天平均喝5.7瓶或2850毫升的水。

图8-4:饮水

2014年我经历了三次生活压力事件。 然而近期内我完全没有任何生活上的压力。 因此,我的压力得分为0.51,其满意度为99%。

图8-5:压力评分

睡眠分类有9个因素。 其中,睡眠时间和睡眠中断是我研究中最重要的两个元素。 我的总睡眠分数是0.74,满意度是86%,总体还算可以。

图8-6:睡眠评分

此期间我平均每晚睡7小时15分钟,睡眠充足。
图8-7:睡眠时间

对于大多数中老年男性来说,前列腺增大导致夜间排尿是影响睡眠质量的最主要原因。以我为例,我的夜尿频繁则是另外一个问题,我的泌尿科医生告诉我,我的膀胱由于长期受糖尿病的影响已遭损伤。所以在2012至2014年间,我每晚平均起夜4次。然而到了2015至2016年间,这个问题得以改善,每晚平均起夜1.8次(不到2次),这个数据多少显示了是通过调整自我生活方式改善的结果,已经使我的膀胱功能得到了某种程度上的修复。

图8-8:起夜引起睡眠障碍

食物分数仅是数量分数和质量分数的平均值。我的分数是0.73,满意度为77%,就我个人而言,这个得分已经相当不错。

图8-9:食物评分

我要重申一下食物数量和质量的分数在控制慢性疾病中的作用。食物数量是控制体重的关键,这样才能控制多种慢性疾病。我的分数是0.91,或91%为正常食物消耗量(份量),这样才能维持体重在172磅。我下一个目标是将体重减到168-169磅,因此要降低食量在80%左右。

图8-10:食物质量评分

另一方面,我的食物质量分数为0.54,其满意度为96%。如果您每天遵循这系统内的20条规定,就可得到满分100%。依我的经验,此分数可有助于降低血脂及血压,并且记录在总体身体健康状态单位的数据上。由于基因,也可以说是遗传,我天生就是低血压,但是在创业上市的过程中,也经历过许多重压事件,导致了我患上了“暂时性”的高血压。但是我拒绝服用任何治疗高血压的药物,而是通过改变自我生活方式来纠正这个健康问题。
图8-11:食物质量评分

最后,让我们看一下我的日常生活模式的规律性的得分是0.74,满意度为95%。这说明我在过去的7年里,一直遵循着一套规律而健康的日常生活模式。该类别每天共检查14个元素。 世界各地都有实际的证据表明简单健康而又规律的生活方式是长寿的重要关键之一。从商场退休以后,我最终找到了自己新的追求和爱好(用我学的自然科学去研究预防医学及病患的社会心理人格学),同时也过着简单快乐而又有规律的生活。

图8-12:日常生活习惯方式评分

日期:2016年11月1日 14:00
第9节:结论
2010年8月至2016年底,我置身于研究“采用量化医学控制第二型糖尿病”这个课题。4年来,自我研读并且深入理解了六种慢性疾病和食物营养。 此外,我还用了3整年的功夫来研究和开发3个主要以应用数学、计算机科学和工程模型来模拟人体器官的生物医学系统。 这几年来,我专门为患者开发了一套应用软件,他们可在 iPhone/iPad或电脑(PC)上使用。 2014至2016年在我的云服务器上已记录和存储了超过一百万笔我个人的疾病、健康和生活方式的“纯或净”数据(Clean Data)。
虽然我的研究成果与其他医学界的研究异曲同工,但我仍然希望能够从我个人的量化数据中得出相同的有效结论,因为它是使用大数据、云运算和近代科技导出的相同结论,所以我希望我个人的大数据可以为其他患者提供更多的参考价值和结论的可信度。
如前所述,我提供了过去5年的个人健康数据。但我相信,我的发现和工具非常适用于血糖值在 90-400 mg/dL范围内的众多第二型糖尿病(T2D)患者, 因为我个人过去的血糖值就是在这个范围内上下浮动。我的云服务器中尽管还有其他患者的数据可供参考,但是我还没有太多的闲暇时间来分析这些数据。接下来我会将此项目作为下一个课题,留到第二阶段继续深入研究。
2010年8月医生告诉我糖化血红蛋白(A1C)和尿检(ACR)都处于高危状态,需要注射胰岛素以及做肾透析,我听后非常害怕,不知如何是好。回顾自己的健康管理的状况一直都不好,寻医求助得到的只是更多样化和更大量的药物,而这些药物的副作用也无法完全得知。我终于意识到: 我只能寻求自身的帮助,凡事还是要靠自己。当然那些需要药物、手术或紧急护理的病例,还是需要在专业医生的指导下治疗,因为治疗医学还有它的必要性与一定性。患慢性疾病并非一夜之间的事情,也不可能在一夜之间就治愈。但是通过“预防医学”来改变生活方式是显而易见的有效控制方法。自从我被诊断患有糖尿病之后,我就知道这种病是不可能彻底治愈的,但我能做的是尽我所能去控制它,使它不再恶化。依我个人经验,我发现大多数糖尿病患者有3个根本问题:
(1)疾病知识匮乏
(2)缺乏有效的应用工具
(3)缺乏意志力和坚持
控制慢性疾病必须从改变生活方式做起。现在我已掌握足够的糖尿病知识。通过研究,我也开发了实用性工具,并做好每日自身的控制工作。但让我困惑的是,我无法影响并改变身边其他人的生活方式和行为,我相信医生们也有同样的困惑。目前我正在研究“社会物理学”这个课题,也就是:使用自然科学,包括:数学、物理、计算机科学和各种工程方法来解决和改变人类的个性行为与社会大众的心理互动行为,它不仅包括了病患本人的个性,也包含了与他人互动的行为影响力。我计划通过“社会物理学”并结合自身的理工学识与经验,将其纳入到今后第二期的研究项目课题中。我了解这将是另外一个漫长而又艰辛的研究过程,而我成立eclaireMD医学基金会的目标就是为了解决某些疾病以及其相关的医疗护理问题,并且希望能够通过非营利性的公益活动来帮助全世界的其他慢性病的广大患者。

日期:2016年11月1日 15:00
第10节:鸣谢
我要诚挚地感谢以下人员:
首先,我要感谢Norman Jones教授,他是我生命中的贵人。他不仅给我提供了在麻省理工学院博士班深造的机会,而且还培养了我如何解决问题,如何进行科学研究的修养。
我还要感谢 James Andrews 教授。我在爱荷华大学硕士班学习失败时,他给予了我极大的鼓励、帮助和支持。他信任我,并为我安排入读工程科的基础本科,同时还在计算机科学的众多课程中,做了充足的准备工作,他还带领我,第一次进入了生物力学的研究领域。
这两位伟大的教授给予了我很大的帮助和鼓舞,就是因为他们两位恩师造就了今天的我,所以我才有能力来回馈社会,才有能力伸出援手去帮助他人。
Jamie M. Nuwer医学博士来自史坦福大学和加州大学洛杉矶分校医学院,她是一名聪明、年轻的女医师,对她的病人们非常有同情心,而且热情。她在门罗公园的史坦福大学医疗中心工作时给予了我很多的照顾,对于她的鼓励和支持我深表谢意。
Neal Okamura 博士来自加州的圣拉蒙地区医院,过去曾经是我的主治医生,自1992年至2012年,这20年来我深受他的细心照顾。2010年他再次对我的糖尿病严重程度给予警告,正因如此,触发了我开发这个项目的想法与决心。
Jeffrey Guardino 和 Kristine Sherman,这两位来自加州门罗公园的史坦福大学医疗中心的医师,自2012年以来一直是我的主治医生。经过多次拜访、交谈中,我们详细谈到了如何使用我开发的工具来改善自身的健康状况。也就是2015年的时候, Guardino 博士鼓励我写出这篇论文以供病患和其他的医生们参考阅读。
Lynn Bui 博士来自加州大学伯克莱分校和洛杉矶分校医学院的癌症肿瘤专家,自2010年启动这个项目以来,她一直是一位值得信赖的朋友和医学顾问。
旅行中我在台湾遇到了罗嘉雷(Jia-Lei Loo)和曾啓祯(Chi-Jen Tseng)两位医学博士。他们不仅是我的医疗顾问,还是我的网友,我们经常在线上聊很多关于健康的话题,在此我也非常感谢他们的支持和鼓励。尤其是罗博士在过去的5年里,不断地为我解说我的年度健康检查数据,他也是第一个见证了我的努力与进步的人。
我也对来自加州门罗公园的史坦福大学医疗中心的James Ratcliff 和 JoEllen VanZander两位医师深表谢意,感谢他们一直以来的关心和鼓励。
我还要感谢 Steven Bhimji 和 Patricia Hsiao 两位医学博士,感谢这两位医生在早期开发这个产品时候给予我的医学知识与协助。
感谢 Gay Winterringer,她是一位营养学和食品学博士,感谢她给予的食物营养的知识和令人深刻的大量食品营养数据。
Dennis Heller 是我以前在半导体企业的同事,我们曾是在商场上并肩作战的战友,离开半导体产业以后我们成为好朋友。在此我要感谢他在整个项目期间的付出和支持。
我更要感谢我在麻省理工学院读书时认识的老朋友村木丰彦(Toyohiko Muraki)工学博士与教授,他对新陈代谢模型的开发投入了很多精力,并采用多维非线性工程建模技术,对我的研究工作进行了大量的跟进和后续讨论工作。
另外,对美国注册护士 Janet Kwan 的贡献我也深表谢意,她自2013年初加入该项目以来一直全心全意的支持和付出。通过和她的交流探讨,也让我学到了很多关于糖尿病的实际护理常识。
最后,我还要感谢我的妻子:莉莉,虽然她不是一个至关重要的科技研发人物,但是她也是一名糖尿病的长期患者,并且参与到了我的开发项目和实验当中来。她每天和我分享我的工作心得,并给予我热情的、无怨无悔的支持,还提供了与我截然不同的数据以供参考,让我得以撰写我的第二篇医学论文,我从她那里也深受启发与鼓舞。

Read More

Using Quantitative Medicine to Control Diabetes (Traditional Chinese)

Posted by on Feb 22, 2017 in News | 0 comments

撰寫時間:2017年02月17 – 2017年2月20日

論文(2016年12月2日):
摘要:使用量化醫學來控管第二型糖尿病
發表者:美國eclaireMD基金會
日期:2016年11月28日

背景資料
我長期飽受第二型糖尿病、高脂血症和高血壓等新陳代謝慢性疾病的折磨已達20年之久。如下是我自2010年開始研究並記錄的當時的健康數據:
體重:205磅
腰圍:44英寸
壹次性血糖突擊檢查:350毫克/分升
90天平均血糖:280毫克/分升
糖化血紅蛋白(A1C):10.0%
尿蛋白檢驗値 (ACR):116毫克/毫摩爾
甘油三酯:1161毫克/分升

研究目標
我在2010至2016年間一共花了七年的時間來研究糖尿病,最終發現了有效的控制糖尿病病情的方法。並針對第二型糖尿病一共開發出了三種數學模式和各種應用軟件工具來控制這些慢性疾病。在此過程中,我大約收集了一百萬條“純或凈”的健康數據,並采用多項科技方法,利用所學的不同的學科理論,其中包括了:高等數學、計算機科學(如:數據庫、大數據分析、雲計算、移動科技)、非線性和動態的塑性工程理論、人工智能或”Artificial Intelligence (AI)”(數據自動化與機械化學習)來模擬人體器官生物醫學的新陳代謝系統。

應用方法
我所開發的體重和血糖這兩種預測模型的應用,主要是提醒和幫助患者能夠及早和有效地改變自己的生活方式。 血糖預測模型包括了:糖尿病藥物干預、食物和營養的質與量、就餐地點的差異(居家或餐廳)、運動量、各種壓力和創傷,居住地的氣候環境,旅遊的品質及頻率,內臟器官的衰竭度、延時對血糖測量的影響。體重預測模型則包括了:食物的分量、運動的類型與強度、卡路里的變化、睡眠的各種質量、飲水量、及其他有關聯的因素。
對於我本人而言,最明顯的成效是自2013年起逐漸減藥,2015年至2016年這兩年間,我己不再服用任何糖尿病藥物,還可以將糖化血紅蛋白值(A1C)保持在6.2%至6.6%的範圍內,同時還有效的控制住了高血壓和高血脂。請詳見如下的2016年底時期的健康數據:
體重:172磅
腰圍:32英寸
90天平均血糖:115毫克/分升
糖化血紅蛋白(A1C):6.3 – 6.5%(未服用任何藥物)
尿蛋白檢驗値(ACR):12.6毫克/毫摩爾

結果與討論
作為預防醫學的一部分,這整個研發工作是基於生活方式管理、有關的各種信息收集、大量的健康與疾病的數據處理和分析。在60多個各型各類的圖形圖表中,每個圖中都包含至少數百至數千個數據,並將這些研究成果呈現在您面前。此論文中的輸入與輸出數據(I/O Data)的11類大範圍 (Big Category)有著密不可分的互動關聯性(Inter-Connectivity),這包括了四個健康輸出的大範圍(體重和腰圍、血糖、血壓、血脂),六個健康輸入的大範圍(食品和飲食、運動、壓力、睡眠、飲水、生活的規律性)和時效(Time)。這11類大範圍是由500多個比較細微的元素(Element)所組成,它們通過在智能手機上面安裝的APP, 或是在個人電腦上安裝的特定軟件,對病患每日每時的生活、健康、行為、疾病狀況進行全面性與高精確度的監控。系統內部的人工智能大約能夠自動管理95%的元素,剩下的5%(大約20-25個)的元素,則需由患者每日主動輸入系統中。結論是: 我過去5年間的健康代謝指標和預測值都能與醫院體檢的結果幾乎達到了一致,而且我的體重和血糖預測的精確度也分別達到了99.9%和99.0%。
我的研究結論基本上能與醫學界所提供的常見認知與常聞結論達到高度契合。此項研究完全以實驗數據為依據,不帶任何個人偏見或者內含個人的預設立場。因此,我希望經由科學化的大數據和量化分析的方法得以再次驗證的醫學理論能夠得到進一步的推廣和病患們的支持。
目前我們正在進行的科研項目的第2階段,包括有如下的內容:
(1)繼續收集、分析、改進並且公布全球第二型糖尿病患者的大數據資料;
(2)研究開發更有效的方法來影響或改變患者的健康生活行為,以便能接受並且適應更好的生活方式管理。

標題:使用量化醫學來控管第二型糖尿病
作者:美國eclaireMD基金會
日期:2016年10月20日 16:00

第1節:介紹
作為本文作者,過去7年間的研究數據,都基於本人長期患糖尿病期間積累而來,接下來,我將以第一人稱來描述我的病情。
我是一名企業家,1995年在一家高科技上市公司擔任CEO。由於工作原因,我一直生活在一個強高壓的狀態與環境之下。1998年驗血時,我的血糖達到350毫克/分升。1998至2000年期間,我經歷過幾次嚴重的心絞疼與低血糖(俗稱“胰島素休克”)。下圖總結了我自2000至2010年的體檢數據。值得註意的是,本人最高的甘油三酯水平是1161毫克/分升,糖化血紅蛋白(A1C)是10.0%,尿蛋白檢驗値(ACR)是116 毫克/毫摩爾。2010年8月3日醫生建議我馬上開始使用胰島素註射,並告知有可能在3年之內要進行腎透析。接下來的3個月裏,我搬到了一個離群索居沒有壓力的沙漠城市裏,並且關閉了我所有的營利企業公司,開啟了我的自救生命之旅。在過去7年的時間裏,針對6種慢性病從閱讀學習進而研究開發,同時我也徹底地改變了自己的生活方式,最後控制住了病情。這段日子裏,我體重減輕了26磅(從198磅降到172磅)。腰圍減少了12英寸(從44英寸減到32英寸)。2016年9月1日,我的體檢數據有了明顯的改善:尿蛋白檢驗値(ACR)12.6毫克/毫摩爾,糖化血紅蛋白(A1C)6.6%(停止服用糖尿病藥物超過一年多),甘油三酯水平67毫克/分升,高密度脂蛋白(HDL)48毫克/分升 ,低密度脂蛋白(LDL)103毫克/分升,總膽固醇156毫克/分升,體重指數(BMI)25。

圖1-1:健康數據對比圖
1
這是我過去15年來整體的糖尿病狀況。2012年之前,我在醫院化驗的糖化血紅蛋白(A1C)值曾經長期失控。然而2012年後,化驗檢測糖化血紅蛋白(A1C)值約在6.6,這與我應用我自己開發的數學模型模擬創造出來的糖化血紅蛋白值(A1C)曲線高度吻合。還有一點值得注意的是,在此期間我已逐漸減少藥物,直到2015年我可以不再服用任何治療或控制糖尿病的藥物。

圖1-2:2000至2016年糖化血紅蛋白(A1C)記錄表

日期:2016年10月21日 17:00
第2節:方法
2010年下半年,我意識到盡管自己在美國接受過良好的高等教育(17年來在不同的大學裏學習了7個不同的專業),但對於慢性疾病卻還是一無所知。 因此,我決定潛心研究如何控制我的慢性疾病和改善我的健康狀況。在2010至2011最初的兩年,我學習研究了內科醫學裏的6種慢性疾病:糖尿病、高血壓、高脂血症、心臟病、中風和肥胖症。 2012至2013年間我又將注意力轉移到食品科學和食物營養學方面的學習研究。 最後在2014至2016年這三年間,我專注於相關科技的研究和有效控管工具的開發。
這7年的學習和科研期間,我平均每周工作50小時,它所花費的功力與時間就如同攻讀了一項博士學位一樣。但是我的目標卻只是為了徹底的解決我自己的健康問題與拯救自己的生命。最初我本想從傳統的生物醫學研究方法著手,也就是說從最基本的細胞層開始(Micro-Level)著手做基礎研究,但是我沒有足夠的資金和專業的知識培訓來完成這項長期的研究,所以我必須得結合自身已有的知識與經驗,也就是在數學、計算機科學和各種工程學科的優勢。我以前從未涉及過生物醫學領域,也未曾受過該專業的正規培訓;因此,我決定用宏觀(Macro-Level)的角度著手,我采用了非線性塑性動力學的工程模型、有限元素工程分析的概念,以及無機材料的特性,來模擬人體的有機細胞材料與內臟的新陳代謝系統。我使用了高等數學來開發構建這個模型。我設定了體重、血糖、脂肪、和血壓,這4個相互關聯的身體輸出大範圍,還有: 食物、運動、壓力、睡眠、飲水和生活方式,這6個輸入的大範圍。在這10項大範圍裏面又包含了幾百個細節元素。比如:僅“壓力”類一項就分為“正常人”和“非正常人”(如”人格障礙病患者”)兩大類型,而且它們的內部還設定了33個不同的壓力元素。 我還加入“時間”作為第11項大範圍,因為人體健康特征會隨著時間而演變,即“動態”的觀念。我經過不斷的摸索和實驗,也應用了人工智能(AI)來模擬人體的有機特性。在模擬的過程中,由於其複雜性和數據收集的困難度,我排除了所有的特殊環境因素,譬如:汚染、輻射、有毒化學品、毒品、激素治療、病毒感染等。這些因素對癌症尤其起到了至關重要的作用(其實癌症也是一種慢性疾病)。請大家注意,我的研究工作主要是集中在預防醫學領域,因此,有別於大多數的糖尿病研究論文,在我的這篇論文裏面, 藥物僅僅作為血糖預測工具的一部分,而不特別去強調分析它。
鑒於紅血球細胞的平均壽命是3-4個月,其攜帶的血糖和脂類遍及人體,我將3個月內收集的數據作為數控方程式的原始狀態(Initial Condition)。因此,要求患者務必在使用我的工具時,盡可能詳盡的收集最初3個月的數據。有了這個先決原始狀態條件,才可以開始“解方程式”(數學意義上的解釋),然後系統再通過使用人工智能(AI)來做進一步的機械學習和自動分析。2012年1月1日,我開始收集自己的健康數據,並使用該數據不斷測試和改進系統內的數學體系。2014至2015年,我通過了使用人工智能(AI)開始構建“有機”生物醫學數學模型,制作出兩個實用性極強的預測模型,一個是預測體重模型(夜間休息後),另一個是預測餐後血糖值模型(餐後2小時)。到目前為止,在我們基金會的雲服務器裏面,我已收集並處理了近100萬條“純或凈”的我個人的健康與疾病的數據。如果沒有人工智能(AI)系統的分析能力,人腦是無法處理如此龐大的數據的。如圖2-1和2-2所示,我的體重預測精準度已達到99.9%(2015年4月11日至2016年10月21日),血糖值預測精準度達到99.0%(2015年6月1日至 2016年10月21日)。
使用我開發的App工具,不僅可以輕松管理雲端的大量健康數據,還可以預測體重和血糖的健康狀況,並通過新陳代謝指標(MI:Metabolism Index)和綜合身體健康狀態單位(GHSU: General Health Status Unit)來監測整體的健康狀況,這兩個專有名詞是我自己創立的,新陳代謝指標(MI:Metabolism Index)是聚合了11個大範圍和500個元素在每日每時不停地計算出來的,而綜合身體健康狀態單位(GHSU) 被定義為每天的平均新陳代謝指標(MI)值的三個月的運行平均值。 圖表是2012-2016年間我的新陳代謝指標(MI)和總體身體狀況單位(GHSU)(它們已經達到了73.5%的健康水平分界線),很顯然,通過這一套的自動化健康管理,我的整體健康狀況有了明顯的改善(到目前2016年10月底為止,我的代謝指標(MI)和綜合身體健康狀態單位(GHSU) 在57%左右,非常健康)。
圖2-1:預測和實際體重(2015年4月11日-2016年10月21日)

圖2-2:預測和實際每日平均血糖(2015年6月1日-2016年10月21日)

圖2-3:新陳代謝指數(MI)和綜合身體健康狀態單位(GHSU)

日期:2016年10月22日– 2016年10月23日
第3節:體重和糖尿病
眾所周知,要減肥和維持理想的體重是非常困難的一件事,2000年時我的體重是205磅,2010年是198磅,2012年是194磅,2015至2016年間則降到了172磅。 從2013到2014年間我的體重在180磅左右上下波動。 由於我經常去各國出差,所以無法有效的管理好自己的飲食生活和健康狀況。 2014年我開發了新陳代謝的數學模型以後,我就開始使用該工具來管理我的整體生活方式,同時盡量減少長途飛行的次數。 結合我在2015年4月11日開發的體重預測模型,自身減重取得了顯著成效。詳見減肥圖3-1。
圖 3-1:2012-2016年體重

有經驗的人都知道要減少腰圍比降低體重更難。 2000年到2014年,我的腰圍在42到44英寸之間波動。 直到2015年中,我開始觀察自身的新陳代謝指數,並且使用了體重預測工具,終於將我的腰圍成功地縮減到32英寸。詳見腰圍控制圖3-2。

圖 3-2:2012-2016年腰圍

影響體重的因素很多, 飲食和運動是控制體重的最有效方式,當然其他的因素(譬如:睡眠的品質與壓力)也會影響體重。食物種類的質和慢性疾病的類型(下一節討論)有很強的相關性,參見圖3-3和3-4,顯而易見食物的量與體重則有很強的關聯性。例如,從2015年6月至2016年10月,我的體重是172磅。而我的食量大約是我正常食量的91%。假如我想將我的體重再減到170磅左右,那麽我就必須將食物的量再度減少到正常食量的80%。

圖 3-3:食物攝入量 (2015年6月1日-2016年10月20日)

圖 3-4:清晨體重 (2015年6月1日-2016年2016年10月20日)

現在讓我們分析一下體重和血糖之間的關系。 除遺傳因素外,超重(Overweight)與肥胖(Obesity)是第二型糖尿病(T2D)的主因。當我分析體重與血糖時,我標註出體重> 180磅,日均血糖> 160 mg/dL,從圖表中可見兩者之間有很強的相關性。請參見下圖3-5和3-6。若是在就餐以前就能預知你的餐後血糖是多少,你就可以選擇好餐食的質和量,如此一來,你就可以同步控管你的體重與血糖了,(我的餐後血糖預測功能工具就是為了能有效的控制住餐後血糖, 但是同時也能幫助體重的控管)。

圖 3-5:體重超過180磅

圖 3-6 :日均血糖超過160 mg/dL

體重在早晚都會有差異,飲食和運動是其主要的影響因素。日間攝入食物會增加體重,但運動後會燃燒脂肪。 夜晚通過蒸發、排尿和排便等,這三個管道可以減輕體重。去年,我的睡前體重和晨起體重的對比增加值在2.8磅。 夜間休息後到次日晨起體重則也同時減輕了2.8磅。這就是為什麽我的體重能夠壹直保持在172磅(+/- 5磅)左右的原因。次日晨起預測功能工具對控制體重起著至關重要的作用,而控制住體重就是控制住糖尿病的關鍵。圖3-7和3-8呈現了我本人兩種截然不同的體重變化。

圖3-7:同日睡前和晨起體重增加

圖3-8:睡前和次日晨起體重減輕

日期:2016年10月24日 13:00
第4節:血糖
2010年8月3日,我的體檢報告顯示糖化血紅蛋白(A1C)值為10.0%,尿蛋白檢驗值(ACR)為116 mg/mmol,甘油三酯為1161 mg/dL。醫生警告我要立即改變我的生活方式,否則我將必須得做腎透析了! 因此,我立即清醒過來,下定決定洗心革面。2012年1月1日我開始收集自己的健康和生活方式數據。迄今為止,我已收集了約5年的完整數據。 在我的雲數據庫中,我也存儲了大約一百萬條身體和生活方式的數據,其中包括:原始輸入數據、系統計算出的數據、和人工智能(AI)處理過的數據。圖4-1中:呈現的是每日和90天的平均血糖,我的血糖值變化區間在86mg/dL和227mg/dL,平均值為129mg/dL。圖4-2中:體檢報告顯示的糖化血紅蛋白(A1C)值變化區間在6.3%和7.1%,平均值為6.6%。我的數學模擬糖化血紅蛋白(A1C)值(6.3%和8.1%之間,平均值為7.1%)準確率達到92.4%。
糖化血紅蛋白(A1C)體檢結果反映了過去90天的平均血糖值。我的醫學研究公益基金會(eclaireMD)把iPhone平臺上的Wellness App開發出來,這個軟件基於用戶每日輸入的血糖數據,可以預估出一個數學模擬的糖化血紅蛋白(A1C)值。 當然,這種數學模擬糖化血紅蛋白(A1C)值的應用工具相當於一個內置的人工智能工具,它可以根據用戶生物醫學系統參數的變化而自動地進行調整和計算。
通過輸入體檢結果進行頻繁校準可以提高數學模擬A1C值的精確度。自2012到2016年,數學計算的糖化血紅蛋白(A1C)曲線已跟實際測試的糖化血紅蛋白(A1C)數據完全吻合。這意味著,我的模擬糖化血紅蛋白(A1C)值應用工具,在得到體檢報告結論前,就起到了上線預警的作用。
圖4-3 呈現了我所有的實驗測試的糖化血紅蛋白(A1C)值。盡管我己經考慮進去了紅血球細胞的壽命,而加入不同的線性和非線性計算的數學模型,但鑒於高度非線性的生物醫學體系統會隨著時間而變化,因此有時候血糖仍會有不可預測的輸出值。而直接或間接能夠影響血糖值的因素又有很多,接下來我會跟大家討論更多我的研究成果。
2012至2013年間,我曾服用三種糖尿病藥物,如:捷諾維(Januvia)100毫克,艾可拓吡格列酮(Actoplus Met) 15毫克/850毫克,二甲雙胍(Metformin)2000毫克。2012至2016年期間,我的總糖化血紅蛋白(A1C)曲線保持在一個穩定狀態,即6.6左右,2014年初,我開始逐漸減少服藥量,到了2015年的時候,我就已經成功地戒斷,不再服用任何糖尿病的藥物了。這也可以說是我控制住糖尿病最主要的成就之一。
在此期間內,我遭遇了不同程度的“戒斷癥狀”(Withdraw Effect)。在停藥、減少藥量的一個月內,我的血糖圖跌宕起伏,也找不到任何明確的線索和合理的解釋,我曾經多次因為恐懼此舉會傷害自己的身體而又重新回去服藥,但是在反復嘗試了一年多以後,最終我用意志力堅持了一個多月的時間,我的血糖折線圖終於又回歸於平穩狀態(Steady State),所以從2015年初以後,我就不再服用任何糖尿病的藥物了。
圖4-1:日血糖和90天平均血糖(2012-2016年)

圖4-2:數學模擬A1C和實驗室測試A1C對比圖

圖4-3:實驗室檢測A1C值和數學模擬A1C值對比圖

在我的數據收集和分析過程中,我注意到測試數據和模擬數據之間存在有一些偏差。當然,我初試時,曾經試圖找出影響血糖和糖化血紅蛋白(A1C)的主要元素。其中的一部分我將在下面的章節中討論。2015年4月間,我對“黎明現象”(Dawn Effect高空腹血糖)和空腹血糖(Low Fasting Glucose低空腹血糖)之間存在的差異產生興趣。盡管我對肝臟和胰腺是如何制造和控制血糖這方面一無所知,但是我決定通過閱讀大量的醫學論文和分析我自己的大數據來研究空腹血糖。目前為止,我收集了400多個早晨(早餐前)的空腹血糖(FPG)數據。由於空腹血糖(FPG)預測與餐後血糖(PPG)預測不盡相同,而且,一般而言,在線性計算裏,空腹血糖(FPG)的線性權重約為日平均血糖數據的25%,因此我決定使用90天平均日血糖作為空腹血糖(FPG)的初始條件。圖4-4:空腹血糖(FPG)研究結果顯示我的初步發現是基於360天的數據值。這表明,盡管每日空腹血糖(FPG)上下跌宕起伏難以預測,但經過一段時間後,平均空腹血糖(FPG)總是能穩定在90天平均血糖值的附近。預測和實際之間的偏差僅為1.6%,我的預測準確性達到98.4%。

圖4-4:空腹血糖(FPG)研究結果

我患糖尿病長達20年之久,並意識到控制這種疾病必須具備三個先決條件:實用的知識、可靠的工具丶和堅強的意志力與持久力。我運用自學的內分泌科醫學知識,開發出了自己的App應用軟件工具來控制我的體重和血糖。2015年4月11日我開發了體重預測工具,同年6月1日開發了血糖預測工具。值得注意的是,2014年我所開發的新陳代謝模型(MI和GHSU)的主要目的是為了改善自身的整體健康狀況,再加上2個預測系統,通過調整食量、運動頻率和運動強度來預測自己的體重和血糖就變得更得心應手了。換句話說,如果能適當調整輸入參數,輸出值就可能會自動調整至最佳狀態。還有一點應注意,身體是有機體(非線性)和動態體(隨時間變化),所以我們必須不斷監測這種變化的跡象,也必須使用人工智能來不斷地調整數學控制方程式(Governing Equation)。在使用這兩個預測工具時必須遵循下面兩個基本規則:首先,(1) 我必須遵循預測模型的建議,盡可能輸入精確的輸入值。其次,(2) 測量體重或血糖後,不能再重新調整已經輸入了的數值,或者回頭手動輸入作與輸出值吻合的調整,因為這樣做會影響到系統預測的準確性(除非從預測經驗中獲得某些特殊認知或意識到某些新的事實和發現)。人工智能(AI)在系統中已達到一定的級別,但整個生物醫學系統仍需要在應用中不斷觀察和調整。如果有一天即使我含笑九泉,而我所創立的非營利性醫學研究基金會,仍可以繼續深入地研究這些相關的課題,並且持續地改善我們的工具。
2016年7月1日,在我的日記中記錄如下:
“2015年6月1日至2016年6月30日,通過13個月的實際數據分析,我研發出的用來預測A1C的慢性病工具的準確率已達到99.2%,90天的平均血糖值預測準確率已達到了98.4%。因此從2016年7月1日開始,我不再使用血糖儀驗血(試紙法)的方法來測量血糖,而是用自己開發的數學模型取而代之,來預測血糖以控制自身的糖尿病,為了確保糖尿病病情得以控制,我還於2016年10月1日親自去醫院做了一次A1C檢查。接下來的3個月,我非常注意自己的飲食和生活,每天堅持餐後運動。當然,還有幾個其他主要因素,如果這個測試最終成功,那麽我發明的數學生物醫學模型對全球的糖尿病患者而言,不僅降低治療成本,還減少了疾病測試所帶來的痛苦“。
從那天起,我決定減少使用傳統血糖儀(手指穿刺和血糖試紙法)的檢測量。但是,我仍需使用血糖儀繼續測試空腹血糖,因為我正在進一步研究因空腹血糖(FPG)和餐後血糖(PPG)權重因子所引起的A1C的變化。每當我去不能提供營養數據的個體餐廳就餐時,因為不能完全確定就餐食物中的澱粉和糖的分量,我就會使用手指穿刺法測量餐後血糖。2016年7月1日至2016年10月23日期間,115天內,我一共收集了460筆血糖數據,115筆空腹血糖(FPG)數據(總數的25%),另外115筆餐後血糖(PPG)數據(總餐後血糖的33%和總數據的25%)作為實測血糖數據。剩餘的230筆數據(總數據的50%)全部來自於我的預測血糖值。該實驗所得的初步結論是,有一半的情況下我不使用手指穿刺法,然而我血糖預測結果仍能達到99.6%的準確率。為了證實我的發現,我打算在進行一年的實驗。如果此推論能得到證實,那說明我的預測方法是行得通的,大多數第二型糖尿病(T2D)患者平均日血糖水平在100 mg/dL和400 mg/dL之間,他們能使用我的預測工具來控制糖尿病。我希望能用一個簡便的工具,以省去患者使用血糖監測儀和試紙的負擔和成本,讓患者采用更便捷更無痛的方式,而也能更好地控制住糖尿病。請參見圖4-5:50%的預測血糖數據(2016年7月1日-2016年10月23日)。
在東亞、歐洲和美國,約9-10%的人有糖尿病。 中國與東亞國家約有10%,拉丁美洲約有8.5%,而非洲約有5%的人口患有糖尿病。 特別是發展中國家的國民收入有限,每日頻繁的血糖測試對一般家庭來說是個經濟負擔,更不用說更昂貴的糖化血紅蛋白(A1C)檢查了。而我的預測工具對這些糖尿病患者來說,無疑提供了很大的幫助。 當然,蘋果設備對大多數人而言不能觸手可得,而且大多數糖尿病患者又趨於老齡化,他們可能難於掌握現代科技和蘋果手機應用(App)。因此,我的eclaireMD慈善醫學研究基金會也開發了一款面向發展中國家和老齡患者的應用工具,那就是個人電腦(PC)和網絡工具,該工具並與雲服務器同步進行大數據運算。
我的體重和血糖預測工具是一款免費軟件,可通過蘋果APP Store下載。或使用搜索工具欄輸入關鍵字“eclaireMD”,在關鍵字下搜索“Wellness”產品。

圖4-5:當患者不再使用手指穿刺和試紙發方法時,初步數據分析可給出血糖預測值的可靠性和準確性分析。 這些基礎數據有50%是基於血糖預測值(2016年7月1日-2016年10月23日)

日期:2016年10月25日 13:00
第5節:血糖和食物
為了研究血糖和食物之間的關系,我開發了一款蘋果手機應用軟件SmartPhoto。在該應用中,我構建了一個關係型數據庫結構,以便將每張照片儲存在iPhone相冊中。其中數據結構分為5層, 我同時列舉一些範例:
1.分組:(美國、日本、法國等)
2.分類:(居家料理、連鎖餐廳、個體餐廳、航空公司、郵輪等)
3.文件:(丹尼斯餐廳,麥當勞,希臘餐廳,亞洲食物等)
4.名稱:(餐廳名稱、菜名、菜單項等)
5.內容:(存儲記錄,如:營養成分等)
當食物照片存儲到SmartPhoto應用數據庫中,您可以按照自己的喜好方式進行存儲分類和搜索。請參見圖5-1:SmartPhoto應用中每張食物照片附有血糖值。2015年5月1日至2016年10月20日,我收集了1591張食物和就餐照片,它們的平均餐後血糖水平(PPG)為121.8 mg/dL。同期,我的日平均血糖水平(包括空腹血糖FPG)為121.41 mg/dL。

圖5-1:SmartPhoto軟件應用中的食物樣圖

2012至2014年根據我的血糖分析,得出一個初步結論: 高血糖期(接近140 mg/dL)均處於在海外旅行階段。 請參見圖5-2:2012至2014年的血糖測定結果。我發現東亞(不包括中國華北地區)、夏威夷和大溪地島大多數菜肴含糖量較高,而碳水化合物主要來源於米飯、麵粉、芋頭等, 但是我在2016年以後在東亞國家和夏威夷逗留超過8個月之後,我發現到後期的平均血糖水平下降到120 mg/dL以下,比前期下降了20點。 請參見圖5-3:2015至2016年血糖測定結果。
四個主因可降低平均血糖值:
(1)當我使用血糖預測工具時,會更加嚴謹的選擇食材和餐廳菜單裏面的食物;
(2)每日三餐後健走,平均4000步/次,而不是將運動集中在晚上(有關細節詳見如下內容);
(3)旅行期間我會對食物的攝入量和健走運動倍加小心謹慎。 如:在機場餐廳或貴賓室就餐後,我會在登機口通道附近健走3000到4000步;
(4)SmartPhoto工具的分析功能可提供許多幫助,比如:就餐地、菜單和烹飪食材的選擇。

圖5-2:2012-2014年期間血糖

圖5-3:2015-2016年期間血糖

請參見圖5-4,根據SmartPhoto中的大量照片數據,將不同就餐地所得的平均血糖做了統計和匯總。1591份食物和就餐圖片的餐後血糖水平(PPG)為121.8 mg/dL。2015年5月1日到2016年10月20日同期,我開發的APP工具呈現的平均日血糖,包括空腹血糖(FPG)為121.41 mg/dL,這就是為什麽我決定采用平均日血糖值作為初始預測空腹血糖(FPG)值的原因 ,我的90天平均血糖為123.75 mg/dL,如圖5-5:SmartPhoto中2015年5月1日至2016年10月20期間的血糖。

圖5-4:平均血糖和不同就餐地點總表

圖5-5:SmartPhoto中2015年5月1日至2016年10月20日期間的血糖

我的初步研究心得匯總如下:
(1)所研究國家的平均血糖值相似,測量值介於119.9和125.6 mg/dL之間。2015年到2016年,不論外出就餐,還是在家用餐,我都嚴格控制食物攝入量的比例。
(2)在家用餐的血糖值為115.3mg/dL,連鎖餐廳(有營養數據的餐廳)用餐血糖值為125.2mg/dL;個體餐廳(無營養數據)用餐血糖值為132.3mg/dL;機場、航空公司休息室、飛機上用餐血糖值為134.0 mg/dL;食用超市購買的速食品血糖值為140.6 mg/dL。
(3)航空公司提供的食品會讓血糖升高,因為機上食物可選性有限,又無餐後運動的空間。
(4)學習和研究了主要食材的營養成分後,我盡量不食用加工類食品。除非無選擇時,不得不吃,但食用之前我會仔細閱讀標簽上的營養成分(特別是碳水化合物和糖的含量信息)。
(5)個體餐廳經分析所得的平均血糖值如下:
美國:129.8 mg/dL
日本:139.6 mg/dL
臺灣:136.7 mg/dL
其他國家(包括中國):130.8 mg/dL

圖5-6:不同就餐地的平均血糖值

總的來說,美國和西方食物在烹飪過程中不放糖(除甜點外)。 日本、韓國、中國長江以南和東南亞地區(從越南一直延伸到新加坡)在烹飪過程中都會添加糖和鹽。
(6)我在分析某一知名品牌的連鎖餐廳時發現了一個有趣的現象。通常情況下,我不在任何一家連鎖餐廳吃午餐或晚餐,但是早餐例外,因為它們為了薄利多銷,就必須把早餐的份量做小,再加上我再把它們所提供的含碳水化合物的面包類減半進食,這樣我就可以控制好我的早餐後血糖值了。我在同一品牌的美國連鎖餐廳用餐的平均血糖值為122.9mg/dL,在日本則為117.4mg/dL(日本是一個例外,因為他們非常注重並且服從總公司頒布的烹飪標準作業程序(SOP),但是到了臺灣則變成125.3mg/dL,中國則變成126.2mg/dL。我的觀察結果是:在中國和臺灣雖然是同一個品牌,但是,它們到了當地以後,食材多少都會發生變化。而且,我懷疑其采購方式和烹飪的標準作業程序(SOP)都有可能不完全與總部的要求一致。
(7)2013至2014年間,我學習食材營養成分過程中,曾得出了一個錯誤的觀點,那就是: 可以”盡量多吃蔬菜”。但是在2015年,當我收集了六百萬個不同食材的各種營養數據並加以研讀後,我發現各種蔬菜的營養成分其實都有差異。有一天忽然間我的靈感閃現,我通過檢查不同顏色的蔬菜,來辨別它的碳水化合物和糖分的含量,得出了一個便捷的蔬菜與血糖關系的預估方式,那就是使用蔬菜的顏色來分類。請參見圖5-7:蔬菜中碳水化合物和糖含量匯總。我目前的結論是:假如我吃很多蔬菜,餐後血糖水平(PPG)仍有可能會升高。因此我必須注意選擇蔬菜的顏色以及食用的總量,這樣才能獲得更準確的血糖預測值。
(8)當我想吃零食、甜點或水果時,我會選擇恰當的時間,並且限量,以控制血糖和體重。最佳的方法就是在兩餐之間少量食用,例如:上午10點或下午3點。為了要控制體重,睡前盡量少吃。水果對身體健康有益,但要避免高糖份的水果(如:菠蘿,香蕉等),而且一定要限量。有了這種控制機制,血糖就會健康無比。

圖5-7:蔬菜中碳水化合物和糖含量匯總

我也仔細觀察並分析記錄下“極端”情況。 特別是血糖超過200 mg/dL 時。請參見圖5-8 是我17次的用餐情況記錄,2015年5月1日至2016年10月20日血糖值超過200 mg/dL。
值得一提的是,有3個主要因素會讓我的餐後血糖水平(PPG)升高,第一是在個體獨立餐廳內食用東亞食物,第二是在美國連鎖餐廳吃午餐或晚餐,第三是在飛機或遊輪上進餐。如果我能事先獲知這些食物的營養成分,然後使用正確的工具來預測餐後血糖值,再結合自身的意志力與持久力,那麽在這種地方就餐仍然是會安全的。

圖5-8:2015年5月1日至2016年10月20日的17次用餐記錄顯示餐後血糖水平超過200 mg/dL

如下圖所示:圖5-9:分析血糖值大於140的原因,顯而易見高碳水化合物、高糖食物、亞洲食物等三項原因是將血糖值增加了約58%(> 140 mg/dL)。 另壹個發現是,仍有10%的未知因素,無法解釋為何會引發高血糖。

圖5-9:分析血糖值大於140的原因

經由大數據的研究可以很容易地發現碳水化合物、糖分丶和血糖之間的關系。使用以下的幾個方法,可以很容易地通過控制食物的質量和數量來預測餐後血糖值。
(1)我用食品包裝袋上所提供的營養成分和配料成分的克數除以20來估算。例如:碳水化合物16克,16÷20 = 0.8,並將0.8輸入到carbs欄內。同樣地,糖分也是這樣來估算。
(2) 在家烹飪時,我會以手掌的面積和厚度,或是拳頭的體積大小,來當作是100%的量用來估算判斷澱粉分和糖的量。但依我過去幾年(得糖尿病的15年後)的估算、判斷、比較,我最近注意到,我必須以我手掌或拳頭大小的三分之二或者二分之一來當作是100% 的量來估算判斷,我猜測可能是因為受到糖尿病的長期影響,自身內部器官對於碳水化合物(澱粉)和糖分的耐受力已大為降低。在收集更多關於這種現象的大數據之後,我或許需要構建另外一個層次的人工智能來解決這種身體內部器官的質性變化。
(3)我的應用工具可從食材庫中搜索每項食材的營養成分,然後將它們相加,以獲得碳水化合物和糖的總消耗量。
(4)大多數水果都含碳水化合物和糖分,但有一些水果,如:香蕉、菠蘿和葡萄它們的碳水化合物和糖分含量更高。
(5)在用餐時,請一定不要食用點心類甜食,因為它們含有高碳水化合物、高糖、高鹽和高油脂,這些對身體的健康都是很不利的。多吃綠葉蔬菜,非綠色蔬菜盡量少吃,如:茄子,糖含量高的甜菜、胡蘿蔔、玉米、洋蔥和番茄。
對糖尿病患者而言,最重要的原則是要在一天中保持“均衡”的血糖值,也就是說:要隨時降低高血糖,及隨時提高低血糖(避免低血糖休克)。當妳能保持住目標體重,營養攝入均衡,自身的糖尿病和其他慢性疾病就會得以控制。

日期:2016年10月28日 13:00
第6節:血糖和其他因素(運動、壓力、旅行、溫度)
血糖和運動
除食物和飲食外,運動也是至關重要的,它有助於降低血糖。我的APP應用軟件中涵蓋了各式各樣不同類型的運動。但是據我自己多年來的經驗,對老年人而言,保持勻速健走是最好的鍛練方法。我的健走均速是2.5英哩/小時,約步行6000步/小時,或100步/分鐘。
2012年,我平均每天健走8000步,即3.3英哩。那段時間,因為體重是194磅,因超重而導致膝蓋負擔太重,我不能走太久。到2016年,我逐步增加到每天17200步,或每天7.2英哩,到了那個時候,我的體重已經降到172磅,對我而言真是輕而易舉。請參見圖6-1,6-2和6-3。但是,如果持續走路太多,有時候會給膝部或者足底造成負擔,所以最好再進行其他類型的運動,譬如太極拳,來輔助每天的運動量。
2015年初,我發現如果將健走分成三部分,即每餐後2小時內鍛煉,餐後血糖會明顯降低。通過長時間分析我的血糖數據,我還了解到,當平均血糖約140 mg/dL 時,如餐後健走1000步,血糖會降低7至10個點。因此,當我的平均血糖值下降到約120 mg/dL 時,餐後每1000步可將血糖降低4至6個點。這種差異是源於我的數學模型內部所出的初始狀況的假設。
圖6-1:健走記錄(2012-2016年)

圖6-2:晚上健走記錄(2012-2014年)

圖6-3:三餐後健走記錄(2015-2016年)

血糖和壓力
壓力會導致很多健康問題。當人們經歷長期、持續性的壓力時,它會嚴重影響身心健康。在我長達三十年的職業生涯中,一直生活在高壓狀態下。正因為此原因,我分別在五個不同的時間和場合,經歷過多次嚴重的心絞痛,並始入了第二型糖尿病,這最終也導致了我患上了慢性腳趾損傷、膀胱損傷和腎臟損傷。 然而在我從高壓的企業負責人的位置退下來以後,我喜歡上了平靜、無擾的生活方式(2014年除外)。 2014那一年我分別經歷了三次生活事件的壓力(3-6月、9-10月、11-12月)。 請參見圖6-4:2014年3月至12月的壓力期與2015年1月至2016年10月的平和期比較圖。
圖6-4:壓力期(2014年3月至12月)與平和期(2015年1月至2016年10月)比較圖

下圖6-5、6-6、6-7、6-8和6-9,可見壓力、血壓、血糖值和A1C水平之間的明顯相關性。
由於接連發生第二和第三次壓力事件,圖表可見壓力評分、高血壓和血糖與3-6月和9-12月的兩個時間跨度的一致性。因A1C需要3-4個月的平均血糖值,所以A1C值的峰值約在3個月後體現出來。

圖6-5:2014年壓力評分

圖6-6:壓力期血壓升高

圖6-7:高血壓和壓力期的日高血糖

圖6-8:高A1C峰值在3個月後的高血糖出現

圖6-9:高壓力分數和較高的90天平均血糖對比圖

2015年我受了兩次外傷。第一次發生在6月23日,我跌倒在一個路邊的斜坡上,結果面部受傷,進了急診室。發生此事故後,三天之內的血糖值分別為152 mg/dL、208 mg/dL和154 mg/dL,第4天才恢復平穩約120 mg/dL。第二次事故發生在12月4日,我在一個建築工地上受了腿傷,再次進了急診室,在這次事故後,三天的血糖值分別為145 mg/dL、175 mg/dL、165 mg/dL,也是到第4天才恢復平穩約120 mg/dL。由於這兩次突發事故,造成我的血糖水平暫時性升高。兩次事故都是經過了4天,創傷狀態平穩後,我的血糖才逐漸降低恢復平穩狀態。2016年4月8日,我在樓梯上踏空,萬幸身體沒有受傷,但是就在我食用完低碳水化合物和低糖分的早餐兩個小時後,餐後血糖值竟為148 mg/dL。我那天不僅是和平時吃的是一樣的早餐,餐後還健走了4000步。雖然這是小事一樁,但很明顯,壓力的確會影響血糖的水平。

血糖和旅行
在我過往忙碌的企業生涯中,曾走遍世界各地。此處的分析我做了一些簡化,只編譯了2012至2016年的旅行記錄。大於3.5個小時的飛行時間(再加上三到四個小時的進出機場)被我定義為長途旅行,它會影響兩頓餐食的血糖。小於3.5個小時的飛行時間(再加上三到四個小時的進出機場)被定義為短途旅行,它只會對一頓餐食的血糖造成影響。過去5年中,平均每兩周(更準確地說了每12.9天),我就要旅行一次。從我乘機旅行期間的健康狀況分析來看,我註意到旅行期間血糖和新陳代謝的變化都很明顯。它們受到的影響主要是來自於兩個因素,第一是大多數的航空公司與相關部門所提供的食物對糖尿病患者来说都是不友善的,第二是在飛機上或者在機場裏的運動空間非常有限。當我找出了這些原因以後,旅行時段內為了保持良好的血糖值,我會非常小心地選擇對血糖安全性高的食物,而且盡可能用餐之後在機場內的通道上健走。因此,2015至2016年期間,凡空中旅行期間我的血糖和代謝指數都得到大大的改善,幾乎接近正常水平120 mg/dL。請參見6-10圖分析結果。

圖6-10:血糖、新陳代謝和航空旅行的關聯性

血糖和天氣
我在美國成長,也在不同的州內生活了多年,我所居住的大部地區都是無汚染、天氣好、氣候溫和(氣溫15-25攝氏度)。 2016年的上半年,我在東亞地區逗留了6個月,跨越了冬季、春季和初夏。雖然我穿梭於不同的城市,但是我嚴格律己,盡量保持正常的生活方式,對食物、運動、壓力、睡眠、飲水量和其他等項都進行監督。但我注意到,2至6月份,當亞洲的溫度越來越高時,血糖值也會隨之而上升。我無法解釋原因,但我很想知道,炎熱的天氣是否會影響身體的新陳代謝。圖6-11 可見一個短期間(約四個半月)的數據觀察。我之所以寫這個課題,是為了邀請其他研究人員的參與,以得到對這個課題廣泛的關注和研究。

圖6-11:血糖和氣溫的關聯性

日期:2016年10月29日 15:00
第7節:高脂血症和高血壓
2000年開始我就一直收集自己的體檢數據,並將它們輸入到自己的工具中。圖7-1、7-2、7-3和7-4 是繪制的脂質圖形。圖中可見,2000至2012年我患有高脂血症。2012年我主要專注於控制自身的糖尿病,但是我的總方針是利用預防醫學的理念來管理自我,使用最佳的生活方式管理自我,用更好、更有效的方法來控制各種慢性病。2014年我開發了一款以數學模擬代謝模型為生活方式管理的有效應用工具。 因此,最終不僅我的血糖得以控制,我的血脂質數據也變得健康了。 新陳代謝我們將在下一章節進一步探討。

圖7-1:甘油三酯(2000-2016年)

圖7-2:高密度脂蛋白-膽固醇(2000-2016年)

圖7-3:低密度脂蛋白-膽固醇(2000-2016年)

圖7-4:總膽固醇(2000-2016年)

脂類與食物的“質”有著密不可分的關系。 為了飲食健康,我列出了一份食物質量提醒和記錄清單以供參考。 請參見圖7-5:食物質量提醒和記錄。如果您能完全遵守此表內的規定,會得到0.5分。 如果您完全違反規定,則會得到1.5分。 請參閱圖7-6:2014年中到2016年10月20日的食物質量評分。我的“食物質量”滿意度為96% , 得分為0.54,滿意度=(1.5- 0.54)/(1.5-0.5)。

圖7-5:食物質量提醒和記錄

圖7-6:食物質量評分

我的血壓數據如圖7-7、7-8和7-9所示。 正如我前一個章節里提到的,自2014年3月到12月,本人經歷了三次連續性的生活事件壓力期,如下圖所示,您可以看到它們對血壓的影響。 我已經把“攝入低鹽量”作為“食品質量”的要求之一,並在自己的日常飲食中嚴格律己。 從圖7-10所示高血壓病因的分析中,數據清晰顯示了生活上的事件壓力與高血壓有著密不可分的關系,此外,還有海外旅行、時差、飯後運動時間、極端天氣條件等。

圖7-7:最高日收縮壓(SBP)和日舒張壓(DBP)

圖7-8:平均日收縮壓(SBP)和日舒張壓(DBP)

圖7-9:平均日心率

圖7-10:高血壓病因分析

日期:2016年11月1日 11:00
第8節:新陳代謝指數(MI)和 綜合身體健康狀態單位(GHSU)
2014年全年,我進行了研究並開發了以整體健康和慢性疾病為主題的產品。起初我嘗試從定義“新陳代謝”入手,但是失敗了。 譬如:在韋氏詞典中定義“新陳代謝 = 生物體內物質和能量的自我更新過程(細胞或生物體中)”。我也曾嘗試咨詢了好幾位醫師,也無法得到一個準確清楚的定義。總之,這個名詞對許多人來說,只有一個模糊的定性描述(Qualitative Description),而缺乏一個明確的定量描述(Quantitative Description)。後來,我嘗試系統化的收集數據,再加以科學化的整理分析,中間還導出了一套AI監管的數學控制方程式(Governing Equation)去重新定義“新陳代謝”這個名詞。
2014年我花了一整年的時間來研究這個領域,其間我還自創了一個新的術語,它被稱為“新陳代謝指數”(MI: Metabolism Index)。它是基於四大類的人體健康日常輸出數據,和六大類的人體健康日常輸入數據,而這些數據都與慢性病有著密不可分的關系。四類日常輸出數據包括:體重、血糖、血壓和血脂。六類日常輸入包括:運動、飲水、睡眠、壓力、食物及日常生活作息。由於輸入、輸出和生物醫學系統之間的互動是動態的,也就是說它們是會隨音時間而變化的,所以我把“時間”作為第十一大類。
更細一點的定義:每個大分類都是由許多細微的元素所組成。譬如:睡眠有9個元素 (睡前思緒萬千、睡眠時間、醒來次數、晨間清新感覺、醒後頭疼、夢境多寡、環境舒適度、睡眠期間生病或身不舒適、失眠的程度),壓力有33個元素(並非所有因素都適合每個人),而食物類則有約100個元素。再加上系統運算推出與人工智能產生的新元素,最終會有大約五百個因素需加以組織、記錄和分析。當然,每日有效地解決這個問題,對人腦而言,絕對是一個巨大的負擔。在數學理論上,最大的挑戰是如何解決這11個不同分類的大範圍和五百個細微元素之間的互相連動性的問題(Inter-Connectivity)。因此,我利用結構工程的有限元素概念 (Finite Element Method)和動態塑性工程概念 (Dynamic Plastic Engineering Modeling)來模擬這個生物醫學系統 (Bio-Medical system)。我建立了以人工智能為基礎的多組具有各種邊界條件(Boundary Conditions)的數學控制方程式 (Governing Equation)。通過這個過程的努力,剩下的要解決的問題就是要應用計算機科學去開發實用的產品,特別是使用自動化計算和人工智能,這也就是大數據分析和雲運算的用武之地了。
綜合身體健康狀態單位(GHSU: General Health Status Unit)是最近90天內代新陳謝指數(MI)的動態平均值。 最初,我采用了醫學界的定義規範,也就是數值越低越好,我定義了新陳代謝指數在0.5(最佳條件)到1.5(最壞條件)的範圍內。 當新陳代謝指數和綜合身體健康狀態單位都低於1.0時,表示健康狀態良好。 但如果該值超過1.0,可能預示健康和生活方式上出問題了。就我自身而言,我在每個大分類範圍中確定一組最佳元素,並且先確定自身要求達到的最佳健康狀況(Targeted Health Status): 體重為170磅、 血糖為120 mg/dL、收縮壓和舒張壓為120/80、 甘油三酯/高密度脂蛋白-膽固醇/低密度脂蛋白-膽固醇/總膽固醇為150/40/130/200。新陳代謝指數和綜合身體健康狀態單位的“均衡”水平實際上是在73.5%,也就是說:高於73.5%代表不健康,低於73.5%代表健康。 再提示一次,我使用一般醫療實際的做法,也就是用較低的數值來代表更好的健康。
在2016年10月20日這天,我的新陳代謝指數和綜合身體健康狀態單位分別為58.45%和57.1%,這表明在過去的3個月裏我很健康。 我在過去2年內的各種醫院的實驗室測試結果,也證實了我身體非常健康。 通過對生活方式管理,應用量化醫學來控制慢性病是一項有效的措施,我們也可以稱之為“預防醫學“的一部分。 請參閱8-1和8-2圖中2012-2016年間及2015年4月11日至2016年10月20日期間我的綜合身體健康狀態單位和新陳代謝指數。

圖8-1:代謝指數和綜合健康狀況單位(2012-2016年))

圖8-2:代謝指數和綜合健康狀況單位(2015年4月11日-2016年10月20日)

做完新陳代謝指數和綜合身體健康狀態單位的基本概念介紹之後,讓我們研究一些主要分類的評分數值。前面章節中我們已看到許多收集的數據匯總,如:體重、腰圍、血糖、血壓、血脂、食物、運動和壓力。剩余的缺失部分也很重要,但與健康輸出數據沒有直接關聯。本節我還得重申一下食物數據。圖8-3列出了源自我的數學計算模型及其轉化的“滿意度”得出的匯總分類分數,而“滿意度”一詞一看就一目了然。
除了2014年間開發的新陳代謝模型,還有另外兩個重大突破,壹個是2015年4月11日開發的體重預測軟件和2015年6月1日開發的血糖預測軟件。這三個模型為我提供了巨大的幫助和準確的指導,以幫助我控制糖尿病。本節的數據和圖形顯示中,我選擇了2015年4月11日至2016年10月20日之前的資料作為標準區間進行比較。

圖8-3:代謝指數轉換成滿意度級別換算表表

我的飲水分數是0.74,其滿意度為95%(100%定義為每天飲用6瓶或3000毫升水)。 在此期間,我每天平均喝5.7瓶或2850毫升的水。

圖8-4:飲水

2014年我經歷了三次生活壓力事件。 然而近期內我完全沒有任何生活上的壓力。 因此,我的壓力得分為0.51,其滿意度為99%。

圖8-5:壓力評分

睡眠分類有9個因素。 其中,睡眠時間和睡眠中斷是我研究中最重要的兩個元素。 我的總睡眠分數是0.74,滿意度是86%,總體還算可以。

圖8-6:睡眠評分

此期間我平均每晚睡7小時15分鐘,睡眠充足。
圖8-7:睡眠時間

對於大多數中老年男性來說,前列腺增大導致夜間排尿是影響睡眠質量的最主要原因。以我為例,我的夜尿頻繁則是另外一個問題,我的泌尿科醫生告訴我,我的膀胱由於長期受糖尿病的影響已遭損傷。所以在2012至2014年間,我每晚平均起夜4次。然而到了2015至2016年間,這個問題得以改善,每晚平均起夜1.8次(不到2次),這個數據多少顯示了是通過調整自我生活方式改善的結果,已經使我的膀胱功能得到了某種程度上的修復。

圖8-8:起夜引起睡眠障礙

食物分數僅是數量分數和質量分數的平均值。我的分數是0.73,滿意度為77%,就我個人而言,這個得分已經相當不錯。

圖8-9:食物評分

我要重申一下食物數量和質量的分數在控制慢性疾病中的作用。食物數量是控制體重的關鍵,這樣才能控制多種慢性疾病。我的分數是0.91,或91%為正常食物消耗量(份量),這樣才能維持體重在172磅。我下一個目標是將體重減到168-169磅,因此要降低食量在80%左右。

圖8-10:食物質量評分

另一方面,我的食物質量分數為0.54,其滿意度為96%。如果您每天遵循這系統內的20條規定,就可得到滿分100%。依我的經驗,此分數可有助於降低血脂及血壓,並且記錄在總體身體健康狀態單位的數據上。由於基因,也可以說是遺傳,我天生就是低血壓,但是在創業上市的過程中,也經歷過許多重壓事件,導致了我患上了“暫時性”的高血壓。但是我拒絕服用任何治療高血壓的藥物,而是通過改變自我生活方式來糾正這個健康問題。
圖8-11:食物質量評分

最後,讓我們看壹下我的日常生活模式的規律性的得分是0.74,滿意度為95%。這說明我在過去的7年裏,壹直遵循著一套規律而健康的日常生活模式。該類別每天共檢查14個元素。 世界各地都有實際的證據表明簡單健康而又規律的生活方式是長壽的重要關鍵之一。從商場退休以後,我最終找到了自己新的追求和愛好(用我學的自然科學去研究預防醫學及病患的社會心理人格學),同時也過著簡單快樂而又有規律的生活。

圖8-12:日常生活習慣方式評分

日期:2016年11月1日 14:00
第9節:結論
2010年8月至2016年底,我置身於研究“採用量化醫學控制第二型糖尿病”這個課題。4年來,自我研讀並且深入理解了六種慢性疾病和食物營養。 此外,我還用了3整年的功夫來研究和開發3個主要以應用數學、計算機科學和工程模型來模擬人體器官的生物醫學系統。 這幾年來,我專門為患者開發了一套應用軟件,他們可在 iPhone/iPad或電腦(PC)上使用。 2014至2016年在我的雲服務器上已記錄和存儲了超過一百萬筆我個人的疾病、健康和生活方式的“純或凈”數據(Clean Data)。
雖然我的研究成果與其他醫學界的研究異曲同工,但我仍然希望能夠從我個人的量化數據中得出相同的有效結論,因為它是使用大數據、雲運算和近代科技導出的相同結論,所以我希望我個人的大數據可以為其他患者提供更多的參考價值和結論的可信度。
如前所述,我提供了過去5年的個人健康數據。但我相信,我的發現和工具非常適用於血糖值在 90-400 mg/dL範圍內的眾多第二型糖尿病(T2D)患者, 因為我個人過去的血糖值就是在這個範圍內上下浮動。我的雲服務器中盡管還有其他患者的數據可供參考,但是我還沒有太多的閑暇時間來分析這些數據。接下來我會將此項目作為下一個課題,留到第二階段繼續深入研究。
2010年8月醫生告訴我糖化血紅蛋白(A1C)和尿蛋白檢驗值(ACR)都處於高危狀態,需要注射胰島素以及做腎透析,我聽後非常害怕,不知如何是好。回顧自己的健康管理的狀況一直都不好,尋醫求助得到的只是更多樣化和更大量的藥物,而這些藥物的副作用也無法完全得知。我終於意識到: 我只能尋求自身的幫助,凡事還是要靠自己。當然那些需要藥物、手術或緊急護理的病例,還是需要在專業醫生的指導下治療,因為治療醫學還有它的必要性與一定性。患慢性疾病並非一夜之間的事情,也不可能在一夜之間就治愈。但是通過“預防醫學”來改變生活方式是顯而易見的有效控制方法。自從我被診斷患有糖尿病之後,我就知道這種病是不可能徹底治愈的,但我能做的是盡我所能去控制它,使它不再惡化。依我個人經驗,我發現大多數糖尿病患者有3個根本問題:
(1)疾病知識匱乏
(2)缺乏有效的應用工具
(3)缺乏意志力和堅持
控制慢性疾病必須從改變生活方式做起。現在我已掌握足夠的糖尿病知識。通過研究,我也開發了實用性工具,並做好每日自身的控制工作。但讓我困惑的是,我無法影響並改變身邊其他人的生活方式和行為,我相信醫生們也有同樣的困惑。目前我正在研究“社會物理學”這個課題,也就是:使用自然科學,包括:數學、物理、計算機科學和各種工程方法來解決和改變人類的個性行為與社會大眾的心理互動行為,它不僅包括了病患本人的個性,也包含了與他人互動的行為影響力。我計劃通過“社會物理學”並結合自身的理工學識與經驗,將其納入到今後第二期的研究項目課題中。我了解這將是另外一個漫長而又艱辛的研究過程,而我成立eclaireMD醫學基金會的目標就是為了解決某些疾病以及其相關的醫療護理問題,並且希望能夠通過非營利性的公益活動來幫助全世界的其他慢性病的廣大患者。

日期:2016年11月1日 15:00
第10節:鳴謝
我要誠摯地感謝以下人員:
首先,我要感謝Norman Jones教授,他是我生命中的貴人。他不僅給我提供了在麻省理工學院博士班深造的機會,而且還培養了我如何解決問題,如何進行科學研究的修養。
我還要感謝 James Andrews 教授。我在愛荷華大學碩士班學習失敗時,他給予了我極大的鼓勵、幫助和支持。他信任我,並為我安排入讀工程科的基礎本科,同時還在計算機科學的眾多課程中,做了充足的準備工作,他還帶領我,第一次進入了生物力學的研究領域。
這兩位偉大的教授給予了我很大的幫助和鼓舞,就是因為他們兩位恩師造就了今天的我,所以我才有能力來回饋社會,才有能力伸出援手去幫助他人。
Jamie M. Nuwer醫學博士來自史坦福大學和加州大學洛杉磯分校醫學院,她是一名聰明、年輕的女醫師,對她的病人們非常有同情心,而且熱情。她在門羅公園的史坦福大學醫療中心工作時給予了我很多的照顧,對於她的鼓勵和支持我深表謝意。
Neal Okamura 博士來自加州的聖拉蒙地區醫院,過去曾經是我的主治醫生,自1992年至2012年,這20年來我深受他的細心照顧。2010年他再次對我的糖尿病嚴重程度給予警告,正因如此,觸發了我開發這個項目的想法與決心。
Jeffrey Guardino 和 Kristine Sherman,這兩位來自加州門羅公園的史坦福大學醫療中心的醫師,自2012年以來一直是我的主治醫生。經過多次拜訪、交談中,我們詳細談到了如何使用我開發的工具來改善自身的健康狀況。也就是2015年的時候, Guardino 博士鼓勵我寫出這篇論文以供病患和其他的醫生們參考閱讀。
Lynn Bui 博士來自加州大學柏克莱分校和洛杉磯分校醫學院的癌症腫瘤專家,自2010年啟動這個項目以來,她一直是一位值得信賴的朋友和醫學顧問。
旅行中我在臺灣遇到了羅嘉雷(Jia-Lei Loo)和曾啓禎(Chi-Jen Tseng)兩位醫學博士。他們不僅是我的醫療顧問,還是我的網友,我們經常在線上聊很多關於健康的話題,在此我也非常感謝他們的支持和鼓勵。尤其是羅博士在過去的5年裏,不斷地為我解說我的年度健康檢查數據,他也是第一個見證了我的努力與進步的人。
我也對來自加州門羅公園的史坦福大學醫療中心的James Ratcliff 和 JoEllen VanZander兩位醫師深表謝意,感謝他們一直以來的關心和鼓勵。
我還要感謝 Steven Bhimji 和 Patricia Hsiao 兩位醫學博士,感謝這兩位醫生在早期開發這個產品時候給予我的醫學知識與協助。
感謝 Gay Winterringer,她是一位營養學和食品學博士,感謝她給予的食物營養的知識和令人深刻的大量食品營養數據。
Dennis Heller 是我以前在半導體企業的同事,我們曾是在商場上並肩作戰的戰友,離開半導體產業以後我們成為好朋友。在此我要感謝他在整個項目期間的付出和支持。
我更要感謝我在麻省理工學院讀書時認識的老朋友村木豐彥(Toyohiko Muraki)工學博士與教授,他對新陳代謝模型的開發投入了很多精力,並采用多維非線性工程建模技術,對我的研究工作進行了大量的跟進和後續討論工作。
另外,對美國註冊護士 Janet Kwan 的貢獻我也深表謝意,她自2013年初加入該項目以來一直全心全意的支持和付出。通過和她的交流探討,也讓我學到了很多關於糖尿病的實際護理常識。
最後,我還要感謝我的妻子:莉莉,雖然她不是一個至關重要的科技研發人物,但是她也是一名糖尿病的長期患者,並且參與到了我的開發項目和實驗當中來。她每天和我分享我的工作心得,並給予我熱情的、無怨無悔的支持,還提供了與我截然不同的數據以供參考,讓我得以撰寫我的第二篇醫學論文,我從她那裏也深受啟發與鼓舞。

Read More

Quantitative Medicine

Posted by on Dec 15, 2016 in News | 0 comments

Title: Using Quantitative Medicine (a branch of Translational Medicine) to control Type 2 Diabetes

 

By: Gerald C. Hsu, eclaireMD Foundation

Date: 10/20/2016 – 3/8/2017

 

Table of Contents

Abstract
Section 1: Introduction (2 figures)
Section 2: Methods (3 figures)
Section 3: Weight and Diabetes (8 figures)
Section 4: Glucose (8 figures)
Section 5: Glucose and Food & Exercise (17 figures)
Section 6: Glucose and Others (8 figures)
Section 7: Hyperlipidemia and Hypertension (10 figures)
Section 8: Metabolism Index (MI) and General Health
Status Unit (GHSU) (12 figures)
Section 9: Conclusions (0 figure)
Section 10: Acknowledgement (0 figure)

Total 86 pages and 68 figures

BACKGROUND

The author has had long-term chronic diseases and suffered from type 2 diabetes, hyperlipidemia, and hypertension for a period of 20 years.  His primary health data in 2010 are listed as follows:

Weight:  205 lbs.

Waistline:  44 inches

One-time snap check of PPG:  350 mg/dL

90-days of averaged glucose:  280 mg/dL

A1C:  10.0%

ACR:  116 mg/mmol

Triglycerides:  1161 mg/dL

AIMS

The author spent 7 years (2010-2016) conducting research to find an effective way to control his severe diabetic condition.  As a result, he developed three mathematical models and various tools to control his chronic diseases, with an emphasis on type 2 diabetes.  During this process, he collected approximately one million “clean” data regarding his health condition.  He applied multiple disciplines, including advanced mathematics, computer science (e.g. database, big data analytic, cloud computing, and mobile technology), nonlinear and dynamic digitized engineering modeling, and artificial intelligence or “AI” (automation and machine learning) to simulate the human organic metabolic system.

METHOD

The author created two prediction models of weight and glucose values in order to provide patients with an early warning to alter their lifestyle.  The glucose prediction model includes factors such as diabetes medication consumed, quality of food and meal, type of dining location (home or restaurant), exercise amount, stress and trauma, current residence location and weather condition, traveling category and frequency, decreasing internal organ function, and time delay impact on glucose measurement.  The weight prediction model includes factors such as quantity of food and meal, type of exercise, change in calories, sleep impact, water consumption, and other prominent factors.

The most significant achievement is that he successfully eliminated the use of all his diabetic medications within the past 2 years (2015-2016), while maintaining his A1C level within the range of 6.2% to 6.6%.  In addition, his hypertension and hyperlipidemia came under control.  His recent summarized health data is listed below:

Weight:  172 lbs.

Waistline:  32 inches

Averaged 90-days glucose:  115 mg/dL

A1C:  6.3 – 6.5% (without any medication)

ACR:  12.6 mg/mmol

RESULTS AND DISCUSSION

His entire research and development efforts have been based on lifestyle management as a part of preventive medicine, using collection, processing, and analysis of quantitative medical and health data.  The results are displayed in more than 60 figures and diagrams which have hundreds to thousands of data within each illustration.  This paper has indicated many conclusive correlations among 11 categories. These categories include four health outputs (weight and waistline, glucose, blood pressure, lipid), six health inputs (food and meal, exercise, stress, sleep, water drinking, life pattern regularity), and time effect.  All of these 11 categories are composed of approximately 500 elements. All of them are carefully monitored by special-designed computer software via smartphone or personal computer.  About 95% of these elements are managed by AI in which 20 to 25 elements are required by the patient’s daily input.  The metabolism indices for the past 5 years are highly consistent with the author’s health state found from various lab test results.  The accuracy of weight prediction and glucose prediction has reached 99.9% and 99.0%, respectively.

The author’s findings of correlations and conclusions are highly consistent with the commonly available understandings within the medical community.  There is no personal prejudice inside this study as it is based on experimental facts.  Therefore, the author hopes that the same existing medical conclusions can be further backed up and proven by a scientific big data and analytics approach.

The phase 2 of this project will include the following:

(1) Collecting and presenting data from mass population of worldwide patients with type 2 diabetes and;

(2) Researching and developing more effective ways to influence or alter patients’ behavior to adapt a better lifestyle management

 

Section 1:   Introduction

The 7-year research data contained in this paper are based on the work conducted by a long term diabetic patient, who is also the paper’s author.  From here on, I will use the first person to describe my conditions.

I am an entrepreneur and, starting in 1995, was the CEO of a high-tech publicly traded company. I had lived a continuous high-stress life due to the nature of my work.  During a routine blood test in 1998, my glucose level measured 350 mg/dL.  Between 1998 and 2000, I suffered several bouts of severe low blood sugar known as insulin shock.  Attached below is a table which summarizes my health examination data from 2000 to 2010.  Note that my highest triglycerides level was 1,161 mg/dL, A1C 10.0%, and ACR 116 mg/mmol.  On August 3, 2010, my physician advised that I needed to begin using insulin immediately and would most likely end up on dialysis within 3 years.  In the following 3 months, I moved to a new city with less stressors, shut down all of my business enterprises, and started the journey to save my life.  After more than 6 years of doing research on six types of chronic diseases, I have completely changed my lifestyle and have my diseases under control.  During this time, I lost 26 lbs. (going from 198 lbs. to 172 lbs.), and reduced my waistline by 12 inches (going from 44 inches to 32 inches).  On September 1, 2016, my health examination data indicated the following much improved results: ACR 12.6 mg/mmol, A1C 6.6% (WITHOUT taking any diabetic medications for over one year), triglycerides level 67 mg/dL, HDL 48 mg/dL, LDL 103 mg/dL, total cholesterol 156 mg/dL, and BMI 25.

Figure 1-1:

Comparison of health data

Figure 1.1

 

 

 

 

 

 

My overall diabetes conditions during the past 15 years can be seen in Figure 1-2: A1C over 15 years (2000-2016).  Prior to 2012, my lab tested A1C values were out of control, “all over the map”.  However, after 2012, my lab tested A1C values were around 6.6 and they have been enveloped and confined by my mathematical simulated A1C curve.  It should also be noted that, during this same period, I have gradually and then completely eliminated all kinds of diabetes medications.

Figure 1-2:

A1C over 15 years (2000-2016)

Figure 1.2

 

 

 

 

 

 

 

Section 2:   Methods

During the late part of 2010, I came to realize that I had been totally ignorant in the area of chronic diseases, even though I was very well educated in other areas (having studied 7 different fields at various colleges over 17 years).  Therefore, I decided to dedicate my efforts on acquiring the needed “knowledge” to control and improve my health conditions. During the initial 2-year period (2010-2011), I studied internal medicine, with a special interest in 6 chronic diseases which were diabetes, hypertension, hyperlipidemia, heart disease, stroke, and obesity.  During the next 2 years (2012-2013), I focused on food science and nutrition.

After finishing those 4 years of self-studying and preparation, I was ready to fully address my health problems.  I thought about starting with traditional medical research methodology, i.e. basic research starting from the cellular level.  However, I did not have sufficient financial resources and professional knowledge to go down that route.  I also believed the most difficult hurdle was my age (I was 63 years old in 2010).  Therefore, I took stock of my strengths, which were mathematics, computer science, and various engineering disciplines.  I have never received any formal training in the biomedical area; therefore, I used a nonlinear dynamic engineering model and Finite Element concept of structural engineering for inorganic materials to simulate the human body’s organic metabolism system.  And then, I applied advanced mathematics to develop this model’s governing equation.  I defined 4 inter-connected body output categories of weight, glucose, lipid, blood pressure; and 6 variable but also inter-connected body input categories of food, exercise, stress, sleep, water hydration, and life pattern or regularity for longevity.  These 10 categories contain several hundred detailed elements.  For example, just “Stress” category contains 33 different stressors for both “normal” person and “abnormal” person (e.g. person who suffers personality disorders).  I also included “Time” as my 11th category since human body conditions evolve over time, i.e. “Dynamic”.  The human body’s organic characteristics must be dealt with using Artificial Intelligence (AI), through trial-and-error and other techniques.  In the modeling process, I excluded all environmental factors such as pollution, radiation, toxic chemicals, poison, hormonal therapy, viral infection, and others due to their complexity and the difficulty of data collection.  These factors are important for cancer (cancer is also one kind of chronic diseases).  It should be noted that my research is focused on preventive medicine, therefore, drugs for treatment medicine is only included as a part of the glucose prediction tool.

Given the average 3-4 month lifespan of blood cells, which carry glucose and lipids throughout human body, I defined the data collected during the first 3-month period as the initial conditions for solving these mathematical governing equations.  Therefore, it is important for any patient to use my tool to collect his/her initial 3-months data as complete as possible.  After applying these initial conditions on the mathematical system, I can then “solve the equation” (in a mathematical sense), and afterwards, system starts to learn by itself through AI.  In 2014 and 2015, I began building this “organic” bio-medical math model and two practical prediction models for both body weight (after one night of sleep) and glucose levels (2 hours after a meal) by using AI.  On January 1, 2012, I started collecting my own body health data and used that data to continuously test and improve my mathematical system. To date, I have collected and processed near 1 million “clean” data on myself.  Without including AI in system capabilities, the human mind would not be able to deal with such a large and complicated database.  As shown in Figure 2-1 and 2-2, I have reached a 99.9% accuracy on body weight prediction (4/11/2015-10/21/2016) and a 99.0% accuracy on glucose prediction (6/1/2015-10/21/2016).

Using my iPhone APP “Tool”, I can easily manage the massive health data over the cloud, predict my vital signs such as weight and glucose, and be able to monitor my overall health status via the Metabolism Index (MI) and General Health Status Unit (GHSU), which is defined as a three-month running average of the MI value.  From the chart, by mid-2014 (both MI and GHSU had dropped below the dividing health level of 73.5% for my case), it was clear that my overall health conditions had improved significantly (as of now, my MI and GHSU are at around 57%) through a better lifestyle management.

Figure 2-1:

Predicted and actual body weight

(4/11/2015 – 10/21/2016)

Figure 2.1

 

 

 

 

 

 

 

Figure 2-2:

Predicted and actual daily averaged glucose

(6/1/2015 – 10/21/2016)

Figure 2.2

 

 

 

 

 

 

Figure 2-3:

Metabolism Index (MI) and

General Health Status Unit (GHSU)

Figure 2.3

 

 

 

 

 

 

 

Section 3:   Weight & Diabetes

It is well known how difficult it is to reduce body weight and maintain for an extended period of time.  In 2000, my average weight was 198 lbs., in 2010 it was 194 lbs., and in 2016 it was 172 lbs.  From 2013 to 2014, my weight fluctuated around 180 lbs. due to my busy travel schedule and not managing my lifestyle and health well.  After developing the metabolism model in 2014, I started to use this tool to manage my overall lifestyle along with reducing the amount of long-distance flying trips.  Combined with my newly developed weight prediction model on April 11, 2015, I have achieved significant weight reduction.  See Figure 3-1 of weight reduction.

Figure 3-1

Weight (2012-2016)

Figure 3.1

 

 

 

 

 

 

 

Reducing my waistline was a much tougher problem to address than weight reduction.  From 2000 to 2014, my waistline ranged from 42 to 44 inches.  Only until mid-2015, when I started to watch my overall metabolism and use the weight prediction tool, I finally achieved my waistline reduction goal of 32 inches.  See Figure 3-2 for waistline reduction.

Figure 3-2

Waistline (2012-2016)

Figure 3.2

 

 

 

 

 

 

 

There are many factors that affect body weight; however, diet and exercise are the two primary components that people can control.  While types and quality of food have a strong correlation with types of chronic diseases (to be discussed in future sections), the quantity of food has a strong correlation with weight as indicated in Figures 3-3 & 3-4.  For example, from June 2015 to October 2016, my weight has been around 172 lbs. while my food and meal quantity has been around 91% of my normal portions.  But, if I want to further reduce my average weight below 170 lbs., I must cut down my food & meal quantity to around 80% of my normal portions.

Figure 3-3

Food & Meal Quantity (6/1/2015-10/20/2016)

Figure 3.3

 

 

 

 

 

 

 

Figure 3-4

Body Weight in the morning (6/1/2015-10/20/2016)

Figure 3.4

 

 

 

 

 

 

 

Now, let us examine the correlation between weight and glucose.  Other than genetic factors, being overweight or obese is the main fundamental cause of Type 2 Diabetes (T2D). During my analysis of weight vs. glucose, when I plot out my weight >180 lbs. and daily average glucose >160 mg/dL, I have identified a strong correlation between these two factors.  See Figure 3-5 and 3-6 below.  My tool’s functionality of predicting post-meal glucose levels before consuming food is very crucial for controlling my PPG (Postprandial Plasma Glucose).

Figure 3-5

Weight over 180 lbs.

Figure 3.5

 

 

 

 

 

 

 

Figure 3-6

Daily averaged glucose over 160 mg/dL

Figure 3.6

 

 

 

 

 

 

 

A person’s weight is constantly changing throughout the day and evening due to food and exercise as the most important factors.  During the day time, food consumption increases weight and exercise burns off calories.  Throughout the night, the processes of vaporization, urination, and bowel movement affect body weight reduction. During the past year, my average weight gain between bedtime and the morning of the same day is 2.8 lbs.  My weight reduction between bedtime and the next morning is also 2.8 lbs.  That is why I can maintain a constant weight at 172 lbs. (+/- 5 lbs.) over the past year.  My tool’s functionality of predicting the next morning’s weight is a crucial component of controlling my weight; therefore, my weight control is a crucial part of controlling my diabetes condition.  Figures 3-7 & 3-8 show my two kinds of weight changes.

Figure 3-7

Weight gain between bedtime and morning of the same day

Figure 3.7

 

 

 

 

 

 

 

Figure 3-8

Weight reduction between bedtime and next morning

Figure 3.8

 

 

 

 

 

 

 

Section 4:   Glucose

On August 3, 2010, my lab test results showed my A1C as 10.0%, ACR as 116 mg/mmol, and triglyceride level as 1,161 mg/dL.  After receiving a warning from my physician, I decided to change my overall lifestyle.  I began collecting my health and lifestyle data on January 1, 2012.  To date, approximately 5-years’ worth of complete data has been collected.  In my cloud database, I’ve stored about one million data points regarding my body and lifestyle, which includes original inputted data, system calculated data, and processed data.  In Figure 4-1: daily glucose and 90-days average glucose, my glucose levels varied between 86 mg/dL and 227 mg/dL with an average value of 129 mg/dL.  In Figure 4-2: my lab-test A1C values were between 6.3% and 7.1% with an average value of 6.6%. My mathematical simulated A1C values (between 6.3% and 8.1% with an average value of 7.1%) had an accuracy of 92.4%.

The laboratory A1C test result reflects the average glucose values for the past 90 days. The eclaireMD Wellness APP calculates a mathematically simulated A1C value based on the user’s daily glucose input data. However, this mathematically simulated A1C value has a built-in artificial intelligence to automatically customize the calculation according to the changes of the user’s biomedical system’s parameters.

Frequent calibration by inputting laboratory test results can improve the accuracy of the mathematically simulated A1C results.  My mathematical A1C curve has completely “enveloped” the actual lab-tested A1C data from 2012 to 2016. This means that my simulated A1C can provide me with an upper-bound warning before I get tested for the laboratory A1C.

Figure 4-3 lists all of my lab-test A1C values. Although I have taken into account both the lifespan of red blood cells and different linear and nonlinear mathematical models, glucose still has a somewhat unpredictable output value due to a highly nonlinear biomedical body system that changes with time. There are many elements that affect glucose values and I will display more of my research results in following discussions regarding these elements.

Although during 2012 through 2016, the overall A1C curve remains at a somewhat steady state, i.e. around 6.6, the most important factor of controlling my diabetes condition is that I have  been decreasing the dosage of my medication over two years’ time frame and finally completely removed all of my diabetes medications about one year ago.  Between 2012 and 2013, I was taking three different diabetes medications such as Januvia 100 mg, Actoplus Met 15 mg/850 mg, and Metformin 2000 mg.  I started to decrease the number of drugs and also reduce dosage amounts from the beginning of 2014.  By the end of 2015, I had completely eliminated taking any diabetes medication.

During this period, I witnessed different degrees of “withdrawal symptoms.”  Within one month of removing or reducing medication, my glucose chart fluctuated greatly, with many ups and downs, without any clear and reasonable explanation.

Figure 4-1:

Daily Glucose and 90-days Averaged Glucose (2012-2016)

Figure 4.1

 

 

 

 

 

 

 

Figure 4-2:

Mathematical Simulated A1C and Lab-tested A1C Comparison

Figure 4.2

 

 

 

 

 

 

 

Figure 4-3:

List of My Past Lab-tested A1C Values and Corresponding Mathematically Simulated A1C Values

Figure 4.3

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

During the course of data collection and analysis, I have noticed some deviation that exists between tested data and simulated data.  Of course, my first attempt was trying to identify most of the important elements which affect glucose and A1C.  Some of them will be addressed in the following discussions.  Around April 2015, I became intrigued by the difference between “Dawn Phenomenon” (higher fasting glucose) and Fasting Glucose (lower fasting glucose).  Without possessing a complete knowledge of how the liver and pancreas function in terms of creating and controlling glucose, I decided to study fasting by using a big data analysis approach.  Thus far, I have collected more than 400 morning (pre-breakfast) fasting plasma glucose (FPG) data.  Since the prediction of FPG is very different from the prediction of PPG (postprandial plasma glucose), and the linear weight of FPG is about 25% of my glucose data collection per day, I decided to use my 90-day average daily glucose as the initial condition for FPG.  Figure 4-4: FPG Study Results reveals my preliminary finding based on 360-days’ worth of data.  It indicates that, although Daily FPG goes up and down and is difficult to predict, after a long period of time, the averaged FPG settles around the 90-days averaged glucose value. The deviation between predicted and actual is only 1.6% and my predictions reach approximately 98.4% accuracy.

Figure 4-4:

FPG Study Results

Figure 4.4

 

 

 

 

 

 

I have been a long-term diabetic patient for almost 20 years.  I realized that having knowledge, a reliable tool, and will power are the three main ways to control this disease.  During the past 2 years, I have applied my acquired medical knowledge and developed my APP tools to help control both my weight and glucose.  The main tools are my weight predictor developed on April 11, 2015 and my glucose predictor developed on June 1, 2015.  It should be noted that the metabolism model (MI and GHSU) I developed in 2014 is the foundation for improvements in my overall health.  When I could predict my weight and glucose beforehand, it became much easier for me to adjust the amount and quality of my food, along with the frequency and intensity of my exercise.  In other words, if I can wisely adjust my input parameters, my output values will most likely be automatically adjusted for the better.  It should also be noted that the body is organic (nonlinear) and dynamic (changes with time), so we have to constantly monitor for signs of change.  Two fundamental rules had to be followed in terms of using these two prediction tools.  First, I had to follow the prediction model’s suggestions regarding input value as precisely as possible.  Second, after measuring my weight or glucose, I was not allowed to go back to readjust my original input values in order to change the prediction’s accuracy (unless there were some new findings or facts that I had just learned or realized from applying this prediction experience).  Some degree of artificial intelligence (AI) has been built into the system as well, but the entire bio-medical system needed to be continuously observed and modified along the way.  That is why I created a non-profit medical research foundation to continuously work in depth on this subject even after my death.

On July 1, 2016, I entered the following note in my diary:

“Based on the period of 6/1/2015 to 6/30/2016, 13 months of study on actual data analysis, my Chronic Tool has reached 99.2% accuracy rate on my A1C prediction and 98.4% accuracy rate on my 90-days averaged predicted glucose value.  Therefore, starting today 7/1/2016, I will not use the traditional blood test (test strip method) entirely to measure my glucose.  I will mainly depend on my mathematical model to predict my glucose in order to control my diabetes.  I will then go to a hospital or clinic to measure my A1C after 10/1/2016 to make sure that my diabetes condition is still under control.  During the upcoming 3 months, I must be extremely careful in monitoring my diet and meal contents and do my post-meal exercise diligently.  In addition, there are several primary factors to be monitored as well.  If this test is finally successful, my invented mathematical biomedical model on metabolism to control diabetes can reduce both cost and pain for worldwide diabetes patients.”

Since that day, I decided to decrease the amount of testing using traditional glucose meter (finger piercing and test strip method, or glucose meter/ glucometer).  However, I still needed to continue testing my fasting glucose by using the glucometer because of my ongoing research of A1C variation due to weighting factors of FPG vs. PPG.  Whenever I ate my meals at restaurants where no reliable nutrition information was readily available, or cooked a new dish at home using new food materials, I would measure my post meal glucose by the finger piercing method.  For the period of July 1, 2016 to October 23, 2016, I have 115 days’ worth of 460 glucose data, 115 FPG data (25% of total), and another 115 PPG data (33% of total PPG and 25% of total data) as actual measured glucose data.  The remaining 230 data (50% of total data) were relied on my predicted glucose values.  The tentative conclusion from this experiment is that I could eliminate the finger-piercing method and still get a 99.6% accuracy on my glucose results.  In order to validate my findings, I plan to continue this experiment for at least another year.  This is an important temporary finding that, if this conclusion is true and my prediction method is proven to be reliable, then most Type 2 diabetes (T2D) patients whose average daily glucose level fall between 100 mg/dL and 400 mg/dL will be able to use my prediction tool to control their diabetes.  It was my original goal of conducting this kind of reliability research to remove the burden and cost associated with finger piercing and test strips.  Hopefully, by using an easy and useful tool as an option, we can further reduce patients’ reluctance on their efforts of controlling diabetes.  Please see Figure 4-5: 50% of glucose data based on predictions (7/1/2016-10/23/2016).

In Eastern Asia, Europe, and USA, approximately 9-10% of their populations have diabetes.  Latin America has about 8.5%, while Africa has about 5% of their populations with diabetes. Particularly in developing nations, they do not have sufficient resources to do frequent glucose testing, let alone A1C testing.  This easily accessible prediction tool will be able to help them start controlling their diabetes.  On the other hand, Apple devices are too expensive for many to be readily accessible.  Another fact is that most elderly populations are more likely to be diabetic candidates and they may not be as in tune with modern technology and APP’s. Therefore, my charitable medical research foundation, eclaireMD, has also developed a web and cloud-based tool which is aimed for PC users.  These version of my tools are much easier to access by patients in developing countries and elderly populations.

My weight and glucose prediction tool can be downloaded for free from Apple’s, APP Store by searching for “Wellness” product under the keyword of “eclaireMD.”

Figure 4-5:

Preliminary Data Analysis to determine the reliability and accuracy of glucose prediction when patient dropping off finger-piercing and testing-strip method. These preliminary data have 50% of glucose values based on predictions (7/1/2016-10/23/2016)

Figure 4.5

 

 

 

 

 

 

 

 

Section 5:   Glucose & Food

In order to study the relationship between glucose and food, I have developed an APP known as SmartPhoto for the iPhone.  Within SmartPhoto, I constructed a relational database structure to attach with each picture stored in the iPhone album.  The data structure has 5 levels:

1. Group:  (USA, Japan, France, etc.)

2. Category: (home cooking, chain restaurant, individual restaurant, airline, cruise, etc.)

3. File: (Denny’s, McDonald’s, Greek Restaurant, Asian Food, etc.)

4. Name: (restaurant name, dish name, menu item, etc.)

5. Content: (anything you want to keep a record, e.g. Nutrition ingredients, etc.).

Once food photos with its data structure are stored in SmartPhoto, they can be sorted and searched any way a person chooses.

Please see Figure 5-1: SmartPhoto Samples of food & meal with glucose level attached with each photo.  From May 1, 2015 through October 20, 2016, I collected a total of 1,591 pictures of food and meal with an average glucose level (PPG) of 121.8 mg/dL.  During the same period, my daily averaged glucose level (including FPG) is 121.41 mg/dL.

Figure 5-1:

SmartPhoto Sample Pictures of Food & Meal

Figure 5.1

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

From the years of 2012 through 2014, during my glucose analysis, I came to a tentative conclusion that my high glucose periods (close to 140 mg/dL) were contributed to traveling overseas.  Please see Figure 5-2: Glucose results from 2012 to 2014.  I found that the majority dishes of Eastern Asia (excluding Northern China), Hawaii, and Tahiti contain high contents of sugar in their cooking process.  Rice, flour, and/or taro are the main sources of carbohydrates. However, during my recent extended stay in various eastern Asian countries and Hawaii over 8 months has introduced another prominent fact.  My average glucose level dropped below 120 mg/dL – a drop of 20 points from previous periods.  Please see Figure 5-3: Glucose results from 2015 through 2016.

After careful analysis of this average glucose decline, I discovered the following four reasons:

(1) I followed my rule of choosing food material and picking menu items more cautiously when I use the glucose prediction capability of my tool;

(2) I spread my daily walking exercise to three post-meal time frames, averaging 4,000 steps after each meal, instead of concentrating on one walk in the evening (this will be discussed in future sections);

(3) I watched my food and meal intake and walking exercise more carefully on traveling days.  For example, after eating a meal at the airport restaurant or airline lounge, I make the effort to walk 3,000 to 4,000 steps in the aisles connecting the boarding gates;

(4) My SmartPhoto tool’s analysis capability also provides me many insights regarding dining locations, food menus, and cooking material selection.

Figure 5-2:

Glucose during period of 2012-2014

Figure 5.2

 

 

 

 

 

 

Figure 5-3:

Glucose during period of 2015-2016

Figure 5.3

 

 

 

 

 

 

 

From examining the big picture data in SmartPhoto, I tabulated the results in Figure 5-4: Summary Table of Averaged Glucose at Different Eating Locations.  There are a total of 1,591 food and meal pictures with an average of PPG value of 121.8 mg/dL.  During the same period from May 1, 2015 to October 20, 2016, my average daily glucose from my APP tool, including FPG, is 121.41 mg/dL – this is another supporting point of why I decided to use my average daily glucose value as the initial predicted FPG value – while my 90-day average glucose is 123.75 mg/dL as shown in Figure 5-5: Glucose during SmartPhoto Period from May 1, 2015 to October 20, 2016.

Figure 5-4: Summary Table of Averaged Glucose and Different Eating Places

Figure 5.4

 

 

 

 

 

 

 

 

 

Figure 5-5:

Glucose during SmartPhoto Period (05/01/2015-10/20/2016)

Figure 5.5

 

 

 

 

 

 

 

My preliminary explanation and interpretation of causes for these summarized results are as follows:

(1) The average glucose values in all the studied nations are similar, measuring between 119.9 and 125.6 mg/dL.  From 2015 to 2016, I followed strict rules for food and meal intake along with the similar ratio between eating at home and eating outside in every country.

(2) Home cooking equates to a 115.3 mg/dL of glucose value, eating in chain restaurants (where nutritional ingredients information is published) equates to 125.2 mg/dL, eating in individual restaurants (where nutrition information is unavailable) equates to 132.3 mg/dL, eating at an airport, in an airline lounge, and in-flight meals equates to 134.0 mg/dL, and eating ready-cooked food from supermarkets equates to 140.6 mg/dL.

(3) Airline-related food produces high glucose points due to the fact that there are limited options on food items and limited space for post-meal exercising.

(4) After studying the nutrition of major food items, I have tried not to eat processed foods.  However, when I have limited options, I can still eat them provided that I read the nutrition facts on the labels carefully (especially the information regarding carbohydrates and sugar).

(5) Further detailed analysis regarding individual restaurants received the following average glucose values:

USA: 129.8 mg/dL

Japan: 139.6 mg/dL

Taiwan: 136.7 mg/dL

Other Nations: 130.8 mg/dL

Figure 5-6: Measured Average Glucose for Different Eating Places

Figure 5.6

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

In general, American and Western food do not include sugar in the cooking process (except in desserts).  Japanese, Korean, southern Chinese, and Southeast Asian cultures add both sugar and salt into dishes during the cooking process.

(6) I discovered one interesting observation from analyzing a particular popular brand of chain restaurants.  Usually, I avoid eating lunch or dinner at any chain restaurant.  However, breakfast is an exception since the portions are usually smaller due to economic reasons.  As a result, the portion of carbohydrates and sugar are also greatly reduced in certain chain restaurants’ breakfast foods.  This same particular brand of American chain restaurant has an average glucose value of 122.9 mg/dL, while Japan has 117.4 mg/dL, Taiwan has 125.3 mg/dL, and China has 126.2 mg/dL.  My observation is that this particular chain restaurant in Taiwan and China add some local flavors to the menu items; furthermore, I suspect that its standard operating procedures (SOP) of procurement and cooking may not completely comply with its headquarter’s requirements.

(7) From 2013 to 2014, while I was studying food and nutrition, I drew an incorrect conclusion that I could eat as many vegetables as I wanted.  Later in 2015, after I compiled several million points of data on food nutrition, I discovered the differences among various vegetables.  One way I distinguish between how different vegetables affect my glucose is by color.  Please see Figure 5-7: Summary Contents of Carbs and Sugars in Vegetables.  I came to the conclusion that if I eat large quantities of vegetables, my PPG can increase to a higher value.  I must pay attention to the color of vegetables when I eat them in order to get a more accurate glucose prediction.

(8) When I have a craving for snacks, desserts, and/or fruits, I can definitely consume them, however, I must limit the quantity in order to control both my glucose value and weight.  The best practice for me is to eat limited amount of them between meals, for example at 10 am or 3 pm.  I avoid giving in to my cravings before bedtime to assist with my weight control.  Fruits are important for overall body health, however, it is important to avoid eating high-sugar content fruits (such as pineapples, bananas, etc.) and also limit the quantity consumed.  With this control mechanism, I can maintain a healthy level of glucose.

Figure 5-7:  Summary Contents of Carbs and Sugars in Vegetables

Figure 5.7

 

 

 

 

 

It would be interesting to analyze the “extreme” cases in my records, e.g. studying glucose over 200 mg/dL.  Figure 5-8 displays all of my 17 meals which contributed to glucose over 200 mg/dL from May 1, 2015 to October 20, 2016.

It should be noted that the 3 major sources contributing to my extremely high PPGs are eating at individual restaurants offering East Asian food, American chain restaurants, and meals on airlines and cruises.  I can still eat at these locations provided I have knowledge of the food nutrition, use the right tool to predict post-meal glucose value, and have sufficient willpower to resist giving into cravings at the wrong time of day.

Figure 5-8:

17 meals contributed to PPG over 200 (5/1/2015-10/20/2016)

Figure 5.8

 

 

 

 

 

 

 

 

 

Furthermore, as indicated in the following Figure 5-9: Analysis of Causes for Glucose Values Greater Than 140, it is clear that high carbs & sugar food and Asian food have contributed about 58% of higher glucose values (>140mg/dL) causes.  Another interesting fact is that about 10% of unknown reasons occurred, which means I could not explain the actual causes of those high glucose values.

Figure 5-9:

Analysis of Causes for Glucose Values Greater Than 140

Figure 5.9

 

 

 

 

 

 

 

Research has shown that carbohydrates and sugars directly affect glucose levels.  By using the following rules, I can estimate my glucose level before I consume my meal by controlling the food quality and quantity.

(1) I can find the ingredients on the Nutrition Facts label on the food packaging.  I use the amount provided in terms of grams divided by 20 to get the portion estimate.  For example, carbohydrate has 16 grams, then calculate it as 16/20=0.8. The value of 0.8 is entered into the carbs input box of the tool.  I also calculate the sugar amount by using the same method.

(2) When I cook at home, I need to estimate the percentage based on using my open-hand area for estimation or my fist size for volume estimation as 100%.  However, based on my observation for the past few years of portion estimation, I have noticed recently that I need to reduce my 100% estimation to 2/3 of my hand or fist size.  My guess is that my body’s toleration of carbohydrates and sugar has been reduced due to  the effects of diabetes.  After collecting more data regarding this phenomenon, I may need to build another layer of artificial intelligence to address this organic change.

(3) My tool can also search each item of my food components from the food bank, and then add them up to get the total consumption of both carbohydrates and sugar.

(4) Most fruits have both carbohydrates and sugar, but some fruits such as bananas, pineapples, and grapes have higher carbohydrates and sugar content.

(5) It is highly recommended not to eat any desserts, since they contains high carbohydrates, sugar, salt, and fat, which are not healthy for you.  Try to eat plenty of green leafy vegetables; but avoid or reduce non-green vegetables such as beets, carrots, corn, onions, and tomatoes which have higher sugar content.

The most important principle for diabetic patients is to “even out” their glucose wave during the entire day, i.e. push the high tide downward (reduce hyperglycemia) and lift the low tide up (avoid shock from low glucose) like an ocean wave.  Once you are able to maintain your target weight and have a balanced nutrition, your diabetes and other chronic disease conditions should be under control.

Section 6:   Glucose & Others (Exercise, Stress, Travel, Temperature)

Glucose and Exercise

Other than food and meals, exercise was another important factor that contributed to my glucose reduction.  In my APP, I include many different types of exercises.  In my many years of experience, I believe that walking at average speed is the best kind of exercise for many senior citizens.  My average walking speed is 2.5 miles per hour, about 6,000 steps per hour, or 100 steps per minute.

In 2012, I walked an average of 8,000 steps, or 3.3 miles, per day.  During that time, it was difficult for me to walk too long because I was overweight and had weak knees.  By 2016, I gradually increased to 17,200 steps per day, or 7.2 miles per day without any difficulty.  Please see Figures 6-1, 6-2, and 6-3.

In the beginning of 2015, I discovered my PPG would significantly decrease if I spread out my daily walking exercise into 3 segments, i.e. exercising within 2 hours after finishing each meal.  By examining my glucose data for extended periods of time, I also learned that when my average glucose was around 140 mg/dL. With every 1,000 steps taken after a meal, I could reduce my glucose level 7 to 10 points.  However, when my average glucose value dropped to around 120 mg/dL, I could reduce my glucose level 4 to 6 points with every 1,000 steps after a meal.  This difference is due to the assumptions I made in my mathematical models.

Figure 6-1:

Walking Exercise (2012-2016)

Figure 6.1

 

 

 

 

 

 

 

 

 

 

Figure 6-2:

Waking Exercise Concentrating in the Evening (2012-2014)

Figure 6.2

 

 

 

 

 

 

 

 

Figure 6-3:

Walking Exercise Spreading over After 3 Meals (2015-2016)

Figure 6.3

 

 

 

 

 

 

 

Glucose and Stress

Stress causes many health problems.  When people go through a long and continuous stressful lifestyle pattern, it can severely affect their health.  During my demanding thirty-year career, I endured a constant stressful lifestyle. This led me to have severe chest pains on 5 different occasions, and severe Type 2 diabetes, which resulted in chronic toe injuries, bladder damage, and kidney damage.  However, after my retirement, I’ve enjoyed a peaceful, non-eventful lifestyle (except for the year 2014).  During that year, I went through 3 episodes of higher than normal stressful events from March through June, then from September through October, and again from November through December.  Please see Figures 6-4: Comparison of Stressful Periods from March to December 2014 and Peaceful Period from January 2015 to October 2016.

Figure 6-4:

Stress Scores Comparison of Stressful period (3/2014-12/2014) and Peaceful Period (1/2015-10/2016)

Figure 6.4

 

 

 

 

 

 

 

From the following Figures 6-5, 6-6, 6-7, 6-8, and 6-9, we can observe the clear correlation among stress, blood pressure readings, glucose values, and A1C levels.

Since the second and third stressful events occurred back to back, the charts reflect the high stress score, hypertension, and glucose to align with two time spans from March through June and again from September through December. However, the A1C value peaks approximately 3 months later than these time spans because A1C takes 3 to 4 months’ worth of average glucose values.

Figure 6-5:

Stress Score During 2014

Figure 6.5

 

 

 

 

 

 

 

Figure 6-6:

Higher Blood Pressures During Stressful Periods

Figure 6.6

 

 

 

 

 

 

 

 

Figure 6-7:

Higher Daily Glucose During Higher Blood Pressure and Stressful Periods

Figure 6.7

 

 

 

 

 

 

 

Figure 6-8:

Higher A1C peaks Appear around 3 months Later of High Glucose

Figure 6.8

 

 

 

 

 

 

 

Figure 6-9:

Putting Higher Stress Scores and Higher 90-days Averaged Glucose Together

Figure 6.9

 

 

 

 

 

 

 

Furthermore, I suffered two separate physical traumas in 2015.  The first incident occurred on June 23rd, where I fell on a sloped walkway resulting in a face injury and an emergency room visit.  My recorded glucose values for the following three days after the accident were 152 mg/dL, 208 mg/dL, and 154 mg/dL, but then on the 4th day, the value dropped to normal level around 120 mg/dL.  The second incident occurred on December 4th, where I sustained a leg injury on a construction site and went to the emergency room again.  My recorded glucose values for the following three days after the accident were 145 mg/dL, 175 mg/dL, 165 mg/dL and then dropped to normal level around 120 mg/dL on the 4th day.  As a result of these two stressful incidents, my glucose level increased temporarily. Both cases took 4 days to allow the traumatic impact on my glucose to diminish.  On April 8, 2016, I fell after missing a step on a stairway, but I did not sustain any physical injury.  However, two hours after eating my normal low-carbohydrate & low-sugar breakfast, my PPG had spiked at 148 mg/dL. Not only had I ate the same breakfast as I always did, but I had also walked 4,000 steps afterwards.  Although this was a minor incident, it demonstrated that stress can affect the glucose level.

Glucose and Travel

Throughout my life as a businessman, I have traveled extensively worldwide.  To make it simple for my glucose discussion, I have only compiled my traveling record from 2012 to 2016.  I define long trips as air travel time with more than 3 hours (along with +/-2 hours going in and out of the airport) which can affect about 2 meals.  I define short air travel trips as flying time with less than 3 hours (along with +/-2 hours going in and out of the airport) can affect about 1 meal.  During the past 5 years, on average, I have flown every two weeks, or more precisely, every 12.9 days.  From my analysis of my health status on flying days, I noticed that both my glucose and metabolism were affected noticeably by the traveling.  The two main reasons my glucose and metabolism were affected are due to airline food not being the best option for diabetics, and limited space for exercise. Once I figured out the main causes, I managed my air travel meals very carefully by watching what I can eat safely and to walk as much as I can after my meal in the crowded airport space.  Therefore, during 2015 to 2016, both my glucose and metabolism index during air travel days have greatly improved, almost reaching my normal level of 120 mg/dL.  These analysis results are in Figure 6-10.

Figure 6-10:

Correlation Among Glucose, Metabolism and Air Travel

Figure 6.10

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Glucose and Weather

I spent 40 years living in different states in the U.S. with less pollution, great weather, and mild climate (with temperatures ranging from 15 degree to 25 degree Celsius).  During the first half of 2016, I stayed in East Asia continuously for more than 6 months, throughout winter, spring, and early summer.  Although I was traveling to different cities, I disciplined myself to maintain a routine lifestyle, which includes monitoring food, exercise, stress, sleep, water intake, other routines, etc.  However, I noticed my glucose level continued an upward trend during February through June, when the temperature was getting hotter in Asia. I could not explain why but I wondered if the hot weather conditions affected my metabolism.  Figure 6-11 provides a preliminary and short period (about 4.5 months) of data observation.   I wrote this information here to invite other researchers’ attention and input on this topic.

Figure 6-11

Correlation Between Glucose and Atmosphere Temperature

Figure 6.11

 

 

 

 

 

 

 

Section 7:  Hyperlipidemia and Hypertension

I have compiled my physical examination data and entered them into my tool since the year 2000.  The plotted lipid graphics are reflected in Figure 7-1, 7-2, 7-3, and 7-4.  Results have shown that I suffered from hyperlipidemia from 2000 to 2012.  Since 2012, although my focus was to control my diabetes, my overall strategy was to utilize preventive medicine via a better and effective lifestyle management.  The mathematical simulated metabolism model developed in 2014 provided an effective tool for lifestyle management.  As a result, while my glucose values were under control, it assisted me in changing my unhealthy lipid data into a healthy state.  Further discussions about metabolism will be found in the next section.

Figure 7-1:

Triglycerides (2000-2016)

Figure 7.1

 

 

 

 

 

 

 

Figure 7-2:

HDL-C (2000-2016)

Figure 7.2

 

 

 

 

 

 

 

Figure 7-3:

LDL-C (2000-2016)

Figure 7.3

 

 

 

 

 

 

 

Figure 7-4:

Total Cholesterol (2000-2016)

Figure 7.4

 

 

 

 

 

 

 

Lipids have a close relationship with the “quality” of food.  In order for me to consume a better quality of food, I included a list to serve as both reminder and record for the food quality.  Please see Figure 7-5: Reminder and Record of Quality for Food & Meal.  In this table, if you follow all the rules, then you will get a score of 0.5. If you violate all the rules, then you will get a score of 1.5.  Please see Figure 7-6: Score of Quality for Food & Meal from mid-2014 to October, 20 2016.  My “Quality for Food & Meal” satisfaction level is 96% – reflecting a score of 0.54, satisfaction level = (1.5-0.54)/(1.5-0.5).

Figure 7-5:

Reminder and Record of Quality of Food & Meal

Figure 7.5

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 7-6:

Score of Quality of Food & Meal

Figure 7.6

 

 

 

 

 

 

 

My blood pressure data is shown in Figures 7-7, 7-8, and 7-9. As I mentioned in an earlier section, from March through December of 2014, I had 3 consecutive stressful periods and their impact on my blood pressure can be seen on these diagrams.  I have already included “limit salt intake” as one of the requirements in “Quality for Food & Meal,” which I firmly follow in my diet.  From Figure 7-10: Analysis of Causes for High Blood Pressure, my data showed that major stressful events are highly correlated to hypertension, which is then followed by overseas traveling, jet-lag, post exercise timing, extreme weather condition, etc.

Figure 7-7:

Highest Daily SBP & DBP

Figure 7.7

 

 

 

 

 

 

 

Figure 7-8:

Averaged Daily SBP & DBP

Figure 7.8

 

 

 

 

 

 

Figure 7-9:

Averaged Daily Heart Rate

Figure 7.9

 

 

 

 

 

 

 

Figure 7-10:

Analysis of Causes for High Blood Pressure

Figure 7.10

 

 

 

 

 

 

 

 

Section 8:  Metabolism Index (MI) and General Health Status Unit (GHSU)

For the entire year of 2014, I have conducted research and development on the subject of overall health and chronic diseases.  At the beginning, I tried to find a good definition of “Metabolism” but failed.  For example, the Webster Dictionary defines it as “metabolism = the organic processes (in a cell or organism) that are necessary for life.”  Finally, I tried to define metabolism in a quantitative way.

I created a new term called Metabolism Index (Ml). It is based on four categories of human body health’s daily output data and six categories of human body health’s daily input data related to chronic diseases.  The four categories of daily output include body weight, blood sugar, blood pressure, and lipid. The six categories of daily input include exercise, water drinking, sleep, stress, food and meals, and daily routines.  Since input, output, and the biomedical system are dynamic, i.e. they are changing with time; I included “Time” as the eleventh category.

Within each category, there are many more elements.  For example, there are 8 elements in sleep, 33 in stress (not all elements are suitable for everyone), and approximately 100 for food and meal, etc.  At the end, there are several hundred of elements that need to be addressed, recorded, and analyzed.  Of course, it is a huge burden to figure them out effectively on a daily basis.  The biggest challenge is how to solve the inter-connectivity issues among 11 different categories and hundreds of elements.  As a result, I utilized the finite element concept and dynamic plastic behavior of structural engineering to model this system.  I was able to build a set of mathematical governing equations with various boundary conditions.  With these efforts, the remaining problem to solve was to apply computer science, especially computational automation, and artificial intelligence.  This is where big data analysis and analytic come into play.

The General Health Status Unit (or GHSU) is the moving average of Metabolism Index (MI) over the most current 90 days.  Originally, I defined MI to fall within the range of 0.5 (best condition) to 1.5 (the worst condition).  When both MI and GHSU are under 1.0, it means that your health is generally good.  However, if these values are over 1.0, you may have some health issues or related lifestyle problems.  For myself, I finalized an optimal set of elements within each category and also defined my desired healthy level status: 170 lbs. for body weight; 120 mg/dL for glucose; 120/80 for SBP/DBP; and 150/40/130/200 for triglycerides/HDL-C/LDL-C/total cholesterol.  The “break-even” level for both MI and GHSU is actually 73.5%, i.e. above 73.5% is unhealthy whereas below 73.5% is healthy.  Please note that I have adopted the general medical practice of the lower value to represent as better or healthy.

As of October 20, 2016, my MI and GHSU are at 58.45% and 57.1% respectively, which indicate that I am healthy.  My physicians have also confirmed that my general conditions are very healthy based on various laboratory test results of the past 2 years. This is an actual application of how to control chronic diseases via applying quantitative medicine on lifestyle management, i.e. a branch of preventative medicine.  Please see Figures 8-1 and 8-2 regarding my MI and GHSU for a period of 2012 through 2016 and another period of April 11, 2015 to October 20, 2016.

Figure 8-1:

MI & GHSU (2012-2016)

Figure 8.1

 

 

 

 

 

 

 

Figure 8-2:

MI & GHSU (4/11/2015-10/20/2016)

Figure 8.2

 

 

 

 

 

 

 

With the introduction of basic concepts regarding MI and GHSU, let us examine scores of some major categories.  In previous sections, we have already seen many figures of collected data summary, such as weight, waistline, glucose, blood pressure, lipids, food & meal, exercise, and stress.  The remaining missing categories are also important but not so directly linked with the health output data.  I will repeat food & meal data and figures in this section.  Figure 8-3 lists the summary category scores derived from my mathematical computational model and their transformed “satisfaction level” which is a self-explanatory phrase.

Besides the Metabolism Model that was developed in 2014, two other major breakthroughs were produced: the Weight Prediction on April 11, 2015 and Glucose Prediction on June 1, 2015. These three models provided tremendous help and accurate guidance to help control my diabetes.  Therefore, in this section’s data and figure display, I select the period from April 11, 2015 to October 20, 2016 as the standard common period for comparison.

Figure 8-3:

Conversion Table of MI Category Scores to Satisfaction Levels

Figure 8.3

 

 

 

 

 

 

 

 

My drinking water score is 0.74 and its satisfaction level is 95% (100% is defined as drinking 6 bottles or 3,000 ml of water per day). During this period, I have been drinking 5.7 bottles or 2,850 ml of water on average per day.

Figure 8-4:

Water Score

Figure 8.4

 

 

 

 

 

 

 

Three major stressful events happened to me in 2014; however, during this period, I did not encounter stressful situations. Therefore, my stress score is 0.51 and its satisfaction level is 99%.

Figure 8-5:

Stress Score

Figure 8.5

 

 

 

 

 

 

 

Sleep category has 8 elements. Among them, sleep hour and sleep interruption due to waking up are the most important two elements for my case.  My total sleep score is 0.74 and its satisfaction level is 86%, not bad at all.

Figure 8-6:

Sleep Score

Figure 8.6

 

 

 

 

 

 

 

 

During this period, I slept 7 hours and 15 minutes per night on average, which is quite sufficient.

Figure 8-7:

Sleep Hours

Figure 8.7

 

 

 

 

 

 

 

For most male senior citizens, night time urination due to prostate enlargement is the most disturbing factor affecting sleep.  In my case, I was told by my physician that my bladder was damaged due to the long term effects and severity of diabetes.  During the years from 2012 to 2014, I used to wake up 4 times at night to use the bathroom. However, during the period from 2015 to 2016, I only wake up 1.8 times (less than 2) per night on average.

Figure 8-8:

Sleep Disturbance due to Wake Up

Figure 8.8

 

 

 

 

 

 

 

 

My food and meal score is simply the average of both quantity score and quality score.  It is 0.73 and its satisfaction level is 77%, which is a decent score.

Figure 8-9:

Food & Meal Score

Figure 8.9

 

 

 

 

 

 

 

I will repeat both the food quantity and food quality scores in order to emphasize their roles in controlling chronic diseases.  Food & Meal Quantity control is important for weight control, and in turn to control multiple chronic diseases.  My score is 0.91, or 91% of my normal food consumption (portion size) which allows me to maintain my weight at 172 lbs. I have started another push to drop my weight down to 168-169 lbs. by reducing my portion size around 80%.

Figure 8-10:

Food & Meal Quantity Score

Figure 8.10

 

 

 

 

 

 

 

My food quality score is 0.54 and its satisfaction level is 96%.  You can get the 100% score if you follow the ready-defined 20 rules precisely every day.  From my experience, this score helps me to lower my blood lipid to reflect the data in the healthy level status. Genetically, I was born with low blood pressure, but previously as a businessperson I encountered many stressful events that caused me to have “temporarily” hypertension.  I refuse to take medication for my “high” blood pressure, so instead I changed my lifestyle in order to correct this health problem.

Figure 8-11:

Food & Meal Quality Score

Figure 8.11

 

 

 

 

 

 

 

Finally, let us examine my daily routine life pattern score of 0.74 and its satisfaction level 95%.  This means that I follow a regular routine in my daily life pattern.  This category has a total of 14 elements to be checked on a daily basis.  Evidence has shown that a simple and regular life pattern contributes a lot to life longevity.  I finally was able to live this kind of life after I retired and to find new interests to pursue, while maintaining a simple but routine life.

Figure 8-12:

Daily Routine Score

Figure 8.12

 

 

 

 

 

 

 

 

Section 9:   Conclusions

This project “Using quantitative medicine to control diabetes” started in August of 2010 through the end of 2016.  For 4 years, I studied several chronic diseases and food nutrition in depth.  In addition, I invested 3 years of research and develop these 3 major mathematical and prediction models to simulate the human body’s biomedical system.  Along the way, I created an application software for patients to use on their iPhone/iPad or PC.  I was able to produce and store more than one million of “clean” data regarding my health and lifestyle in the cloud server from 2014 to 2016.

Although my findings were no different from other research and discussions that existed in the medical community, I hope the same conclusions are drawn from my personal quantitative data; therefore, it can provide confirmation and credibility for other patients to follow.

As I mentioned previously, I am presenting my personal health data from the past 5 years.  However, I am confident that my findings are highly applicable to many other Type 2 Diabetes (T2D) patients with cases of having glucose values in the range of 90 to 400 mg/dL.  Although there are other patients’ data available in the cloud server through the use of my tool, but I have not spent enough time to analyze their data yet.  This will be Phase 2 of this project.

In August of 2010, when my physician informed me that both my A1C and ACR values were dangerously high, I was extremely scared and did not know what to do next.  I thought about my health condition long and hard and then I realized that there is only one person who can help me.  That person is MYSELF.  For those cases that need medication, operation, or urgent care, a trained medical doctor is definitely needed for treatment and guidance.  However, in the case of chronic diseases, I did not get this disease overnight thus it cannot be cured overnight.  The clear way to overcome these conditions is through “preventative medicine,” which requires a lifestyle change.  Since I was diagnosed with diabetes, there is no way I could cure my disease completely, but I can do my best to control it from getting worse.  From my personal journey, I found 3 fundamental problems associated with most diabetic patients:

(1) Lack of disease knowledge;

(2) Lack of a useful tool;

(3) Lack of willpower and persistence.

That is why it is difficult to change our lifestyle in order to control our chronic diseases.  By now, I have acquired sufficient knowledge regarding diabetes.  Through my research, I have also developed practical tools to control it on a daily basis. However, I am still puzzled about how to influence others to change their lifestyles and behaviors.  I am currently studying this problem via “Social Physics”, i.e. using natural science, including mathematics, physics, computer science, and various engineering methods to address and alter human beings’ social behaviors.  I know this is another long and tough journey.  I founded eclaireMD Foundation with a goal to address certain diseases along with medical problems and to conduct nonprofit activities to help other patients worldwide.  My plan is to incorporate what I learn through “Social Physics” and incorporate it in future phases of this project.

 

Section 10:   Acknowledgement

I would like to thank the following people:

First and foremost, I wish to express my appreciation to Professor Norman Jones, who is a very important person in my life.  Not only did he give me the opportunity to study at MIT, but he also trained me extensively on how to solve problems and conduct scientific research.

I would also like to thank Professor James Andrews.  He helped and supported me tremendously when I failed academically at the University of Iowa.  He believed in me and prepared me the undergraduate requirements on how to build my engineering foundation.

These two great professors helped me, so that I could one day help others.

Jamie M. Nuwer, MD in Palo Alto, California, is a bright, young physician who has compassion in helping her patients.  She also took me under her care while she worked at the Stanford Network in Menlo Park.  Her encouragement and support are deeply appreciated.

Neal Okamura, MD in San Ramon, California, was my primary physician for 20 years, from 1992 to 2012.  He was the one who warned me in 2010 regarding the severity of my diabetic condition, which triggered me to launch this project.

Jeffrey Guardino, MD and Kristine Sherman, MD of the Stanford Network in Menlo Park, California, have been my primary physicians since 2012.  During my many visits with them, we spoke in detail about how I used my tool to improve my health.  It was Dr. Guardino who encouraged me in 2015 to write this paper for others to read.

Lynne Bui, MD in San Jose, California, has been a trusted friend and medical advisor since 2010 when I started this project.

During my journey, I met two medical doctors in Taiwan, Jia-Lei Loo, MD and Chi-Jen Tseng, MD.  Not only are they my medical advisors, but they are also my social network friends.  We constantly chat about many health-related issues online.   I am grateful to them for their support and encouragement.

My deep appreciation goes to both James Ratcliff, MD and JoEllen VanZander, MD of the Stanford Network in Menlo Park, California, for their care and encouragement.

I also would like to express my appreciation to both Steven Bhimji, MD, and Patricia Hsiao, MD, for their early participation of the “knowledge” development for this project.

I want to thank Gay Winterringer, who has a PhD in nutrition and food science, for her impressive input and knowledge.

I have known Dennis Heller for many years while working together and we became good friends after my retirement. I would like thank him for his time and feedback throughout the project.

I would like to extend my appreciation to my dear old friend from the MIT days, Dr. Toyohiko Muraki, for his input on my development of the metabolism model, using multi-dimensional nonlinear engineering modeling technique, and many follow up discussions regarding my research work.

My deep appreciation also goes to Janet Kwan, RN for her devoted contribution and whole-hearted support to this project since she joined in early 2013.  I have discussed and exchanged numerous ideas regarding diabetes disease and its care with her.

Last, but not least, I would like to thank my wonderful wife, Li Li, who is a diabetic patient herself and has participated in my program.  Her participation and daily support of my efforts provided additional data points different from mine, resulting in many insightful revelations.

 

Read More

Bunions (Q & A)

Posted by on May 20, 2015 in News | 0 comments

FootBUNIONS (Q & A)

What is a bunion?
Bunion is an abnormal bony bump at the base of the big toe. Medically, it is referred to as hallux valgus deformity. This enlarged and misaligned joint causes the big toe to point toward the second toe instead of staying straight forward in a normal position. Sometimes, it may even overlap the second toe. The pressure from the big toe may cause the little toes to be misaligned as well. The skin covering the joint may look red and inflamed.

Can you have bunion on the other toes?
Yes, you can have bunion on the base of your little toe (fifth toe). It is called bunionette or tailor’s bunion.

Are bunions painful?
Yes, bunions can be very painful especially with walking. Larger bunion may cause difficulty with walking.

What are the complications of bunion?
Bunion causes your big toe to look distorted and misshapen cosmetically. In addition, the fluid-filled sac around the joint can get inflamed and a condition called bursitis may develop. Over a long period of time, arthritis may develop on the affected joint.

Do bunions affect both genders?
Yes, bunions affect both men and women. But it is estimated that bunions are ten times more common in women than men.

Do bunions run in the families?
Yes, bunions run in the families. If you have a defect in the structure of your foot, you are more likely to develop bunions.

What are the causes of bunions?
Bunions develop from wearing a tight-fitting, narrowed or pointy toe shoes or high heels. It could also be genetic or due to abnormal structure in your foot that you are born with.

What are the signs and symptoms of bunions?
Common symptoms include pain, swelling, redness and tenderness around the joint on the big toe. Common signs are having abnormal bony bump on the side of the big toe and big toe points toward the little toes. The pressure from the big toe may cause the crowding of the little toes and develop a condition called hammertoe. The skin on the bottom of your foot may get thickened and calluses may develop on the skin along the big toe. If bunion is large and severe, you may have pain with walking.

How are bunions diagnosed?
There are no specific tests to diagnose bunions. The doctor will make the diagnosis based on the examination of your foot. He or she will look for redness, swelling, inflammation and the skin changes on your foot and big toe. The doctor will assess the range of motion of the joint on the big toe. Some doctors may order X-ray of the foot to see if you have deformity or abnormal structure of the bone. X-ray can also detect arthritis of the foot.

How are bunions treated?
Treatment depends on severity of pain and effects on your daily life. If you have mild pain that comes on and off, the doctor may try conservative treatments first. He or she will recommend that you wear comfortable shoes with wide toes or sandals. You can find over-the-counter arch support, pads for your soles and spacers between the toes. There are some orthopedic shoes or devices you can get with a prescription.

As for pain, you can take acetaminophen (Tylenol) or ibuprofen (Motrin, Advil), naproxen (Aleve, Naprosyn) or NSAIDs (Aspirin, Ecotrin). You may apply ice pack to the affected area to help reduce inflammation and swelling. If you have a more severe pain, your doctor can give you corticosteroid injection to the area.

If you do not find relief with conservative treatment and pain medication, surgery is an option. There are different surgical procedures available and the doctor will discuss the right one for your case. You can go home on the same day for bunion surgery. Bunionectomy is one of the surgical procedures in which abnormal bony growth is removed and the big toe is realigned back to the normal position. The doctors will instruct you to wear proper shoes after the procedure to ensure a recovery and prevent new bunions from forming. Surgical option is available for bunionette (bunion on fifth toe) as well.

Surgery is not an option for teenagers because the bones are not fully grown yet. Bunions in teenagers are often treated with conservative treatments.

Contributed by Patricia Hsiao M.D.
Sources: aaos.org, mayoclinic.com, medicinenet.com, ncbi.nlm.nih.gov

Read More