Noninvasive HbA_1c Detection based on Digital Volume Pulse Signal using Machine Learning

Diabetes mellitus is a metabolic condition characterized by a prolonged presence of sugar molecules in the bloodstream, which leads to hemoglobin glycation. Glycated hemoglobin (HbA<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1c</inf>) is quantified as a proportion of total hemoglobin cells. The traditional approach necessitates invasive blood testing procedures in specialized laboratories. The study suggests a noninvasive approach for measuring %HbA<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1c</inf> levels using a digital volume pulse (DVP) signal. The DVP sensor captures signals at three specific wavelengths: green, red, and infrared (IR). Once the data has been preprocessed and relevant features have been extracted, a random forest regressor is used to predict the percentage of HbA<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1c</inf>. The collection of DVP data from 9 subjects validates the study. The estimated values are compared to the actual %HbA<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1c</inf> values. The results demonstrate a close agreement between the estimated values and actual values. The results indicate a strong correlation between the reflected DVP signal and the different wavelengths, with Pearson’s correlation coefficient (r) values of 0.924, 0.975, 0.931, and 0.986. Furthermore, the error analysis employs mean squared error and root mean squared error as the evaluative metrics.