Exhaled gas biomarkers: a non-invasive approach for distinguishing diabetes and its complications

Abstract

Exhaled gas detection offers a safe, convenient, and non-invasive clinical diagnostic method for preventing the progression of diabetes to complications. In this study, gas chromatography-mass spectrometry (GC-MS) analysis and statistical methods were employed to identify four volatile organic compounds (VOCs) that exhibit significant differences between patients with Type 2 Diabetes Mellitus (T2DM) and those with Diabetic Complications (DC). Compared with those in DC patients, the concentrations of isoprene, acetone, and isopropanol were found to be higher in T2DM patients, whereas the concentrations of tetradecane were lower. Based on the sets of these four VOCs, a voting classifier was constructed using three machine learning methods–Support Vector Machine (SVM), Random Forest (RF) and K-Nearest Neighbors (KNN). The accuracy, sensitivity, specificity, F1 score, and AUC value of the voting classifier are 90.8%, 92.1%, 89.5%, 0.909, and 0.988, respectively, in distinguishing between T2DM and DC. This diagnostic method of exhaled gas detection provides an important foundation for preventing DC and monitoring disease progression of DM.