การพัฒนาระบบประมาณระดับกลูโคสในเลือดแบบไม่รุกล้ำ

Abstract

Researchers have been working on designing and developing a device and method for non-invasive blood glucose measurement that uses optical detection. The device operates by emitting light energy at near-infrared radiation and near-red wavelengths through areas of the body capable of detecting the contraction of PPG blood vessels, such as the fingertips. This non-invasive approach to monitoring blood glucose levels is crucial for individuals at risk of or suffering from diabetes or obesity, as it allows them to track their blood glucose levels and receive appropriate treatment to maintain equilibrium. This contrasts with current invasive measurement methods, such as self-monitoring blood glucose, which requires a small needle to puncture the fingertip and collect a blood sample. For individuals who need to measure their blood glucose levels daily, repeated punctures can result in bodily injuries and larger wounds. The researchers’ technique estimates blood glucose levels by analyzing characteristic signals obtained from the contractile signal collection in a time-series manner. They found a correlation between blood glucose data and 1 0 features, including heart rate signals, ac and de components, perfusion index, and the ratio of the perfusion index between infrared and red-light signals. The researchers created a blood glucose estimation model using a Polynomial Regression model, based on the most correlated characteristics of the ac component and perfusion index of the red-light signal. They tested the 1 st, 2nd, and 3rd derivatives in 2 participants over a period of 1 0 days, with each participant providing data for 5 days. The blood glucose test was divided into two parts: the first part after glucose intake and theง second part after eating, while controlling the subject’s diet. The first-order differential regression model was found to be the most accurate, with an accuracy of 99.78% for the red ac component model and 99.77% for the red-light perfusion index model.