Linear Regression Analysis Results of the CGM Sensor PPG Comparison Between Predicted PPG Data Using the Candlestick Model and the LEGT Model Against the Measured Sensor PPG Data During a 2-year COVID-19 Quarantine Period for a Type 2 Diabetes Patient Based on GH-Method: MathPhysical Medicine (No. 541)

Abstract

Since 5/5/2018, the author has been applying a continuous glucose monitoring (CGM) sensor device on his upper arm that collected and recorded the complete glucose data continuously at 15-minute time intervals on his iPhone. He accumulated 96 glucoses per day over the past ~3.5 years. As a result, over these 1,272 days, he has compiled a total of 122,112 glucose data and stored them in his database where postprandial plasma glucose (PPG) occupies 45,792 data size and 37.5% of the total glucose database. During the 2020-2021 COVID-19 quarantine period, he maintained a strict daily routine, without any travel, allowing him to reach an overall healthy lifestyle. Therefore, all of the 19 influential factors of PPG are mainly control by two primary factors: carbs/sugar intake amount (average at 13.1 gram, low-carb diet) and postmeal walking exercise (average of 4,300 steps). These lifestyle improvements helped reduce his PPG waveform amplitudes, including the four associated PPG data of candlestick (aka K-line) model: opening, maximum, minimum, and closing. Based on the simplified and healthy lifestyle, he can then easily utilize his developed candlestick (aka K-line) model to develop another set of predicted PPG values in addition to the LEGT model results shown in paper No. 540. In previous research reports, he applied physics concepts and theories, engineering models and equations, mathematical concepts and formulas, computer big data and artificial intelligence (AI) techniques, as well as some statistical approaches. However, the majority of published medical papers he read are mainly based on statistics. As a result, in this article, he selected one of the basic statistical tools, linear regression analysis, to study the comparison between his predicted PPG using K-line model and CGM sensor measured PPG. In conclusion, the linear regression analysis results using the K-line model provide more accurate predicted PPG results than using the LEGT model. Actually, both of the maximum PPG and minimum PPG of K-line model are generated through LEGT equation, but the starting PPG of K-line is a measured sensor value instead of the calculation via 0.97*FPG. Moreover, the K-line model contains one extra data point for the closing PPG value. This is why the K-line model has extracted from 4 out of 13 measured sensor PPG data which results in a higher prediction accuracy than the LEGT prediction calculated from carbs/sugar intake grams and post-meal walking steps.