Math-Physical Medicine

NO. 249

Predicting Finger PPG by using Sensor PPG waveform and data via regression analysis with three different methods (GH-Method: math-physical medicine)

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

This paper describes the research results of the predicted finger-piercing postprandial plasma glucose (PPG) value using a Sensor PPG waveform and its data collected from the continuous glucose monitoring (CGM) device.  The author utilized a regression analysis based on three distinctive methods: Method A on matched timing, Method B on matched PPG, and Method C on the lowest Sensor PPG.  In other words, he tried to identify one particular sensor PPG value on the CGM sensor 3-hour waveform, which is “nearly-equal” to the measured finger PPG value.

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 CGM device on his upper arm and checked his glucose measurements ~80 times each day.  After the first bite of his meal, he measures his interstitial glucose level every 15 minutes for a total of 3-hours or 180 minutes.

He has maintained these dual glucose testing for 718 days (5/5/2018 – 4/23/2020) for the purpose of in-depth glucose research and a special comparison of these “almost-matched” PPG values based on three different models.

In this analysis, he has combined three PPG values of breakfast, lunch, and dinner into a daily average PPG and further utilized the following three different models:

  • Model A (matched timing based)
    Using regression analysis to search for the matching time instants and then calculate their associated PPG values.
  • Model B (matched glucose based)
    Using regression analysis to search for the near-equal glucoses and then calculate their associated time instants.
  • Model C (lowest glucose based)
    Using the lowest PPG value within 180 minutes after the first bite of food.

Other than using the regression analysis concept and method, the majority of his tasks are actually simple and straightforward mathematics along with statistics work plus some software modifications for big data analytics.  His primary concerns are always his data accuracy and data integrity because he cannot allow to create an inaccurate Finger PPG prediction value from a collected Sensor PPG dataset and waveform.

For the period of 2014-2017, based on the Finger PPG data, he has already developed numerous conversion formulas, calculation equations, and outcome refinements for some important diabetes variables, such as the PPG relationship with diet, exercise, weather temperature, glucose prediction, HbA1C prediction, and much more.  Therefore, he must maintain the validity and accuracy he has achieved regarding these equations and formulas.

The author has discovered that both finger and sensor measurements have some inherent issues related to the devices’ reliability and their data accuracy.  As seen in Figure 1, there are no strong correlation (from R = 32% for 90-days moving average curve to 46% for daily data curve) between finger PPG and sensor PPG due to their inherited device reliability and accuracy issues.

Figure 1: Correlation Coefficient R of measured daily glucoses between Finger PPG and Sensor PPG (daily data with R of 46% and 90-days moving average with R of 32%)

Figures 2 shows the comparison between Finger PPG and calculated PPG values based on three chosen models of the CGM Sensor PPG respectively, in terms of their correlation coefficients (R) and average PPG values. Figure 3 combines these curves associated with three respective models to compare against the finger PPG with data table of accuracy % and correlation coefficients.  

From Figures 2, 3, and 4, it is obvious that Model 3 (the lowest PPG value) offers the best fit in terms of the combination of both average accuracy (~100%) and acceptable correlation (R = 49%).  

Figure 2: Average PPG and R between Finger PPG and three respective models (5/5/18 - 4/23/20)
Figure 3: Final comparison and choice among 3 models, Model C is the best choice (curve graph and data table)
Figure 4: Comparison among Finger PPG, Sensor PPG, and 3 models (both daily curve and 99-days moving average curve)

After a lengthy and complicated analysis work utilizing both timing value based (Model A) and glucose value based (Model B), the author found that the lowest CGM sensor PPG value (Model C) would provide the best choice for matching with finger PPG value.

The author could then proceed with his software enhancement work of replacing finger PPG measurement data entry by using the computer calculated input results from the measured sensor PPG waveform data directly.  This approach will not only save the author lots of data-entry time, but also maintain his computer APP usage and its data accuracy’s confidence in terms of continuously applying his already developed mathematical equations and formulas to further predict other important biomedical variables, such as the predicted HbA1C value.


  1. Gerald C. Hsu, eclaireMD Foundation, USA. April 2020.  ”Predicting Finger PPG by using Sensor PPG waveform and data via regression analysis with two different methods, matching time and matching glucose (GH-Method: math-physical medicine).”