Cardiac, diabetic and normal subjects classification using decision tree and result confirmation through orthostatic stress index

Cardiac and diabetic diseases are common and troublesome complications, leading to increased morbidity and mortality rates that cause a large economic burden. Researchers have explored cardiac or diabetic diseases for analysis separately; however, combined study of both diseases is missing. A retrospective Heart Rate Variability (HRV) analysis on ECG data was performed, analyzing the data of 78 subjects, which includes 26 cardiac (suffering from congestive heart failure), 17 diabetic and 35 normal (healthy subject or control) subjects. These subjects were labeled as cardiac, diabetic and normal based on HRV parameters. A classification and regression tree (CART) was employed to develop an automatic classifier to infer implications of disease in the early stages. Furthermore, to cross verify the above results, an orthostatic stress index was calculated from HRV parameters of ECG data recorded in the sitting and supine position. Cardiac subjects have the lowest HRV values among the three classes, whereas healthy subjects have the highest. The results are consistent with OSI values. The results are consistent with the consensus shown by previous studies, namely that depressed HRV is a useful tool for risk assessment in patients, and implying that cardiac subjects are at highest risk. The proposed classification model has achieved an accuracy of 93.58%, and a precision and recall of 94%.