GH-METHODS

Math-Physical Medicine

NO. 140

Using math-physical medicine methodology and bioengineering techniques to control diabetes

Corresponding Author: Gerald C. Hsu, eclaireMD Foundation, USA.

Introduction
In this paper, the author describes his GH-Method: math-physical medicine along with bioengineering techniques to establish his prediction models for controlling type 2 diabetes (T2D) and other chronic diseases by using a lifestyle management approach.

Methods
The author briefly describes various academic methods and practical techniques he uses to develop his diabetes research methodology, glucose prediction models, and extension into other chronic diseases in the endocrinology area.  Since endocrinology involves many internal organs in a “black box” manner, his research must rely on observing detailed physical phenomena, while collecting relevant big data, creating realistic hypothesis, and then verifying the validity of his theory via either mathematical equations if possible, or at least via statistical or computational models, including artificial intelligence.

  • (A) Category selection and definition:  He selected four outputs of medical conditions or diseases, i.e. weight (obesity), glucose (diabetes), blood pressure (hypertension), and lipid (Hyperlipidemia); and six inputs of lifestyle, i.e. food, water intake, exercise, sleep, stress, and daily routine life pattern.
  • (B) Elements selection and definition:  He selected a total of 500 elements for the above-mentioned 10 categories, for example, 9 for sleep, 40 for stress, 15 for daily routines, over 100 for food, etc.
  • (C) Data collection and storage:  He developed an APP software product to be used on a smart phone or PC to collect relevant data and to be stored on a local device, but backed up onto a cloud server.  Approximately 2-million data points regarding his health and lifestyle have been collected and stored on the server since 2010.
  • (D) Data format consistency:  He defined and then cleaned the collected big data according to his consistent defined format.
  • (E) Metabolism mathematical modeling:  He applied topology, finite element method, and nonlinear algebra concepts to develop a metabolic mathematical model with an 11-page long equation.  HbA1C (hemoglobin A1C) prior to 2015, an extension of metabolism output has included a 15% “safety margin”, an idea came from his nuclear power plant design experiences.
  • (F) Postprandial plasma glucose (PPG) prediction:  He used optical physics to predict post-meal PPG value directly from ~6,000 meal photos with ~8 million nutritional data in food nutrition database.  The predicted PPG values used physical phenomena observation, hypothesis creation, and then mathematical verifications which bypassed the long-period of study on molecular structure and their internal chemical interactions among food nutritional ingredients.
  • (G) Empirical equation of exercise on PPG:  He conducted various experiments on types and intensities of exercise, which were suitable for majority of T2D patients to develop a sophisticated conversion equation and a simplified conversion formula between PPG and exercise.
  • (H) Time-series analysis of statistics:  He used time-series method to obtain correlation coefficients (R) between glucose and its key influential factors such as fasting plasma glucose (FPG) vs. body weight, body weight vs. food portion, and PPG vs. food, exercise, and others.
  • (I) Spatial analysis of statistics:  He used spatial analysis to identify specific data patterns and their moving trends.  For example, he applied triangular dual-biomarkers relationship spatial analysis diagrams to identify the differences between one patient taking medication and another patient without medication.
  • (J) Big data analytics:  He collected approximately 2-million data over nine years from one T2D patient (author) and conducted various data analysis to derive relevant and effective mathematical equations or simple formulas in order to be used by other T2D patients to predict their biomedical outcomes.
  • (K) Signal processing:  He identified five influential factors of FPG and 19 influential factors of PPG via signal processing of wave theory and signal waveform decomposition technique.
  • (L) Wave theory:  He studied glucose waveform’s key characteristics, e.g. wavelength, frequency, period, and amplitude to establish different PPG waveforms, including Himalaya, Grand Canyon, twin peaks, and a completely different FPG waveform.  He used wave theory and data segmentation analysis to investigate specific waveform characteristics.
  • (M) Energy theory:  He calculated the associated energy levels of specific glucose waveforms to determine the damage caused on internal organs due to these associated excessive “left-over” energies from high glucoses.
  • (N) Frequency domain analysis:  He used the Fourier transform technique to convert a time domain representation into a frequency domain presentation and then investigated the energy associated with low-frequency glucoses vs. high-frequency glucoses.
  • (O) Candlestick (K-line) chart:  He used this financial analysis tool from Wall Street’s stock prices to study open, close, minimum, maximum, and average glucoses over a period of time and their moving trends or relationships.
  • (P) Elasticity and dynamic plastic behaviors of structure element:  He investigated and build up an artery rupture model (about 20% of situations) and micro-vessel leakage model using structural mechanics knowledge.  This helped his understanding of CVD, stroke, renal, retinal, bladder, and foot ulcer to calculate risk probability of having these diabetes complications.
  • (Q) Fluid dynamics:  He investigated and build up an artery blockage model (~80% of situations) of CVD and stroke.  Lipid has no effect on micro-vessels.
  • (R) Risk probability of having artery and micro-vessel issues:  CVD and stroke’s artery blockage (fluid dynamics) and rupture (structure mechanics) as well as renal, retinal, bladder, foot ulcer, nerve system’s micro-vessel’s leakage and damage (structure mechanics) have a common root cause of high glucose (diabetes) in combination with high blood pressure.  High lipid is the “accomplice” creating diabetes complications while high glucose is the “culprit”.
  • (S) Segmentation analysis:  He used segmentation-analysis technique in many situations, including the following five prominent types:

(1) Using glucose levels to investigate high glucose (>180 mg/dL or >140 mg/dL) impact on T2D complications and life-threatening insulin shock (<70 mg/dL).
(2) Using segmented nations, locations, and meal types data to learn where and which kind of meals are better for glucose control.
(3) Using both low and high carbs/sugar amounts of meal contents and time instants (between 0-minute and 180-minutes) segmented data to identify communications between the brain and other vital internal organs via nervous system.
(4) Using annualized segmented glucose data (both FPG, pre-meals and pre-bed glucose, PPG baseline glucose) to investigate annual “self-recovery” rate of his own pancreatic beta cell health state of insulin production capability.
(5) Using a segmented period of severe plantar fasciitis condition to investigate its direct relationship with glucose level via exercise reduction.

  • (T) Glucometer device reliability analysis:  He used glucose details and A1C pattern characteristics to investigate the product reliability and measurement data accuracy of a specific popular brand of 14-data continuous glucose monitoring (CGM) device which may cause a “life-threatening” critical situation for certain diabetes patients if they are linking their CGM data with their insulin automatic pump device.
  • (U) Progressive behavior modification of psychology model:  He used this mentality and psychology modification model to link T2D patients’ physiological behavior profile with their lifestyle data and diabetes physiological data together.
  • (V) Artificial Intelligence (AI):  He applied AI in many aspects and cases of his diabetes research work.  He developed an AI Glucometer for other T2D patients to use on their smart phones or PCs in order to control their diabetes conditions effectively.

Results
In this section, the author only displays 11 representative figures to show part of his diabetes research results using the above-mentioned 22 GH-Method: math-physical medicine methodology and bioengineering techniques.  Each individual figure tile indicates certain major methods he utilized.  In reality, his research work usually involves multiple methods or techniques or combination thereof which are listed in the section of Methods.  This article will not discuss in detail on how he conducted his research, including physical phenomenon interpretation, theoretical hypothesis, mathematical proof, and derivation of equations or formulas.  These figures simply illustrate a part of his research results, his knowledge, and experiences of diabetes control.

Conclusion
This article summarizes some of his math-physical medicine research methodologies and bioengineering techniques from both an engineering and computer science perspective.  His research results prove the power and effectiveness of his GH-Method: math-physical medicine and bioengineering techniques on his chronic diseases biomedical research work.

Figure 1: Math-physical medicine vs. Bio-chemical medicine
Figure 2: FPG vs. Weight
Figure 3: PPG (time-series and spatial-analysis)
Figure 4: Signal processing of PPG waveform to decompose into 19 sub-waveforms (major factors are carbs/sugar and walking)
Figure 5: Candlestick (K-Line) Chart of glucose variances
Figure 6: Metabolism Index (M1 through M10) and HbA1C
Figure 7: Segmentation analysis of low-cards food vs. high-carbs food to study communications between brain and internal organs
Figure 8: Three PPG waveforms and their associated energies
Figure 9: Conversion of time domain curve into frequency domain curve to study high-frequency glucoses vs. low frequency glucoses amplitudes and associated energy
Figure 10: Risk probability of having CVD, stroke, and renal complications using structure mechanics and fluid dynamics
Figure 11: Linkage between behavior psychology and diabetes physiology
Figure 12: Artificial intelligence applications, including an AI Glucometer which achieves ~99% of PPG prediction accuracy