Comparison between mathematical methods to estimate blood glucose levels from ECG signals

Abstract

Diabetes mellitus, known as diabetes, is a chronic disease that affects the control of blood glucose concentration levels, it is a disease that mostly affects adults (type 2 diabetes), but it can also occur in children (type 1 or childhood diabetes), as well as in pregnant women (gestational diabetes). Diabetes is one of the diseases with the highest prevalence and high mortality worldwide. Diabetes has no cure, but continuous monitoring to maintain blood glucose levels in normal ranges reduces the possibility of suffering from gastrointestinal problems, vision loss, limb amputations (such as diabetic foot) and damage to vital organs such as the heart and kidneys, among other associated complications. This article compares the results in glucose estimation by using a linear, quadratic and cubic regression considering the electrical characteristics generated in the cardiac conduction (HR, HRV, T-wave peak, and QT interval) recorded on a single-lead electrocardiogram (VII), used as a non-invasive blood glucose estimation model. The best estimate was obtained using a cubic regression. The validation was performed using the Clarke grid having 77.78 % of data in the A zone and 22.22 % in the B zone and a Pearson correlation value of 0.94103 in the cubic regression.

Keywords

• Diabetes
• Electrocardiogram
• Blood glucose concentration
• Linear regression
• Quadratic regression
• Cubic regression