Analysis of ECG Segments for Non-Invasive Blood Glucose Monitoring

Continuous blood glucose (BG) monitoring is necessary to avoid the deadly health complications from diabetes mellitus. The conventional method of measuring and monitoring BG is by pricking the finger which causes pain and discomfort to patients. To tackle this issue, there are research focusing on physiological signals, such as an electrocardiogram (ECG) to create a model capable of continuous glucose measurement. However, there are ECG segments that have not been considered that have the possibility of improving the performance for non-invasive BG monitoring. In this paper, we perform an oral glucose tolerance test (OGTT) on thirteen adults while continuously recording the ECG signal. A control experiment was also performed without the consumption of glucose. We captured continuous ECG signals and extracted 9 ECG segments. Boxplot and correlation coefficient analysis was performed on the extracted segments to observe the changes for BG. The result reveals a consistent pattern among QT, ST, QTC segments from each participant. HR and RR-I segments have dominant inverse behavior with a 92% correlation with the QT segment. While PRQ and QRS segments can also be included due to 85% and 77% correlation respectively with QTC segments. Whereas R-H and P-H segments have weak results with most of their values below 50%.