Non-invasive discrimination between diabetic states (HBA1C<8% and HBA1C>10%) using photoplethysmography

Abstract

Diabetes mellitus is a group of metabolic diseases associated with the production and/or reaction of insulin leading to hyperglycemia. Glycated hemoglobin (HbA1c) level is generally measured for hyperglycemia. The risk of developing complications depends on both the duration of diabetes and hyperglycemia. A trend of increasing arterial stiffness has been identified in type 2 diabetes. Photoplethysmographic (PPG) pulse wave provides a ‘window’ into the properties of small arteries whereas stiffening of these arteries will alter the PPG waveform. In this research, the potential of PPG in discriminating between type 2 diabetic patients at risk of having HbA1c level > 10% has been investigated. To this end, PPG signals recorded from diabetic patients with different levels of HbA1c (HbA1c level < 8% and HbA1c level > 10%) were acquired from the index finger of the right arm of 101 subjects (53 subjects with HbA1c level < 8% and 48 subjects with HbA1c level > 10%) at a sampling rate of 275 Hz. The area under the curve of PPG (auc-PPG) was proposed in analyzing the PPG pulse contour. Results of t-test analysis show that auc-PPG is significantly larger in diabetic patients with HbA1c level < 8% than in those with HbA1c level > 10% (p-value <0.001). Repeated measurement of PPG using paired t-test on 30 diabetic patients with HbA1c level < 8% and 26 diabetic patients with HbA1c level > 10% (total 56 subjects) show that there is no significant difference in the mean value of auc-PPG between the first measurement and repeated measurement for both groups. Finally, a logistic regression model for estimating the risk of having HbA1c level > 10% among diabetic patients was estimated using data from 51 female diabetic patients. The model shows that the auc-PPG is an independent predictor for estimating the risk of having HbA1c level > 10% (p-value = 0.005) among female diabetic patients.