Math-Physical Medicine

NO. 276

Characteristic pattern study of the glucose waveforms using GH-Method: mathphysical medicine

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

This paper describes the research results of the waveform characteristic patterns of daily finger and sensor glucoses, postprandial plasma glucoses (PPG), and fasting plasma glucoses (FPG) over a period of 2+ years.

Since 1/1/2012, the author has measured his glucose values using the finger-piercing method: once for FPG and three times for PPG each day. On 5/5/2018, he applied a continuous glucose monitoring (CGM) device on his upper arm and checked his sensor glucoses at 76.89 times each day and, by 2/19/2020, measurements were taken every 15-minute interval. He has maintained these dual glucose testing methods for 770 days from 5/5/2018 to 6/13/2020. Currently, he uses this database to conduct in-depth research on certain glucose characteristic patterns. In total, he has already collected 62,285 glucose data to be utilized for this particular study.

He applies both time-series analysis, X or Y versus time, which is similar to EKG charts, along with spatial analysis in a two-dimensional X and Y space, without “time” factor, to analyze his collected big glucose data.

n time-series analysis, when the correlation coefficient (“R”) is greater than 50% (strong), then it is considered as “highly correlat-ed”. When R is between 30% and 50%, it is deemed “somewhat correlated”. When R is less than 30% (weak), then it is considered as “non-correlated”. It should be noted that the correlation coeffi-cient can only be calculated for two sets of data. By using time-se-ries analysis, the author presents his results in both daily discrete data chart and 90-days moving average data chart. The reason for including the 90-days moving average data is based on the general understanding that the HbA1C value is an average glucose for the past 90 days. Please note that these two correlation coefficients are slightly different between the daily discrete data and 90-days mov-ing average data. This is due to the minor differences existing be-tween collected glucose data and calculated moving average data.

In spatial analysis, if the “data cloud” is concentrated within a long and narrow band (similar to the shape of a cucumber or a football) and skewed with an angle where the slope is greater than zero, which means the existence of correlation, then these two sets of data are correlated. On the other hand, if the angle of the plotted data cloud is either flat or vertical, then they have an exceptionally low value of R and considered as non-correlated.

Regarding the work process of identifying correlation from spa-tial analysis, he first must identify the data ranges, which are the minimum and maximum of both X dimension and Y dimension. He then applies a visual estimation and trial-and-error process to figure out the best-fitted “slope” of a skewed “data cloud”. This approach is much simpler and faster than derivation of an equation since he only needs an approximate guess from the diagram. The last step is to establish two skewed boxes as shown in his spatial analysis diagrams. The orange box indicates that data within +/- 10 % of the skewed green line of slope in the center, while the yellow box indicates that data within +/- 20 % of the center green line of slope. Each box has its different area percentage of total data con-tained in the colored box.

Another purpose of this study is to demonstrate the effectiveness of these two statistical tools, time-series analysis and spatial anal-ysis, on conducting certain type of medical research work.

Figure 1 shows the summarized time-series analysis and spatial analysis for all three glucoses of both discrete and 90-days moving average values, including daily glucose, PPG, and FPG. Figures 2, 3, 4 are 3 respective diagrams of daily glucose, PPG, and FPG. Figure 5, 6, 7 are 3 detailed processed data of spatial analysis of daily glucose, PPG, and FPG. Figure 8 places three spatial analy-sis data cloud together. Figure 9 shows details of PPG waveforms which include 0-minute, 60-minutes, 90-minutes, 105-minutes, 120-minutes, 180-minutes, average and peak PPG values.

Significant Conclusions from Figure 1 Through 9 Are Listed:

  1. Sensor glucose is 13%-14% higher than finger glucose, sensor PPG is 17%-18% higher than finger PPG, both sensor FPG and finger FPG are almost identical.
  2. From time-series analysis results, these three glucose sets, daily glucose, PPG, and FPG, have six correlation coefficients which are between 42% and 56%. This means that they are correlated, but not strongly correlated. His belief is that both testing methods, including finger piercing and CGM sensor, have their different product reliability issues. At times, the margin of error could reach to 25% or higher. He has written a few articles regarding this concern.
  3. The three spatial analysis diagrams in Figure 8 show that all of these three data clouds are skewed with some angles. This means that the finger data and sensor data are correlated.
  4. These eight curves in Figure 9 demonstrate a high similarity existing among these eight waveform patterns. This is an in-teresting finding because the entire 180-minutes PPG wave-form pattern has been already determined and even somewhat fixed when we eat certain amount of carbs/sugar, maintain a specific exercise level, and meet specified secondary require-ments.
Figure 1: Summarized data table
Figure 2: Time-series and spatial analysis results of daily glucose
Figure 3: Time-series and spatial analysis results of PPG
Figure 4: Time-series and spatial analysis results of FPG
Figure 5: Spatial analysis detailed data of daily glucose
Figure 6: Spatial analysis detailed data of daily PPG
Figure 7: Spatial analysis detailed data of daily FPG
Figure 8: Spatial analysis results comparison of three glucoses
Figure 9: 8 time-series PPG curves

Regardless of the reliability issues from the glucose testing devices, the test results of glucoses from either finger-piercing or sensor collection would indicate a reasonable high correlation. This statement has been proven by the author using his >62,000 glucose data and two reliable statistical tools [1-6].


  1. Hsu, Gerald C (2020) Predicting Finger PPG by using Sensor PPG waveform and data via regression analysis with three dif-ferent methods using GH-Method: math-physical medicine, EC Diabetes and Metabolic Research 4: 21-24.
  2. Hsu, Gerald C (2020) Applying segmentation pattern analysis to investigate postprandial plasma glucose characteristics and behaviors of the carbs/sugar intake amounts in different eating places using GH Method: math-physical medicine (No. 150).
  3. Hsu, Gerald C (2020) Applying segmentation pattern analysis to investigate postprandial plasma glucose characteristics and behaviors of the carbs/sugar intake amounts in different na-tions using GH Method: math-physical medicine, Internation-al Journal of Nanotechnology and Nanomedicine, Opaston-line 5: 17-20.
  4. Hsu Gerald C (2020) A case study of the impact on glucose, particularly postprandial plasma glucose based on the 14-day sensor device reliability using GH-Method: math-physical medicine, Journal of Bioscience & Biomedical Engineering, Unisciencepub 1:1-4.
  5. Hsu Gerald C (2020) Comparison of two glucose measure-ment results and their influence on excessive energy’s impact on risk probability of cardiovascular disease and stroke using wave characteristic analysis using GH-Method: math-physi-cal medicine (No. 48).
  6. Hsu Gerald C (2020) Type 2 Diabetes Nursing Guidelines based on Wave and Energy Theories Associated with Three Distinctive PPG Waveforms Developed via GH-Method: math-physical medicine, Journal of Nursing & Healthcare, opastonline 5: 1-2.

Copyright: ©2020 Gerald C Hsu,. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.