Identification of cardiac autonomic neuropathy patients using cardioid based graph for ECG biometric

In this paper, the application of data mining applied on Cardioid based person identification mechanism using electrocardiogram (ECG) is presented. A total of 50 subjects with Cardiac Autonomic Neuropathy (CAN) were obtained from participants with diabetes from the Charles Sturt Diabetes Complication Screening Initiative (DiScRi). The patients can be categorized into two types of CAN which are early CAN and definite/severe CAN. Euclidean distances obtained as a result of the formation of the Cardioid based graph were used as extracted features. These distances were then applied in Multilayer Perceptron to confirm the identity of individuals. Our experimentation results suggest that person identification is possible by obtaining classification accuracies of 99.6% for patients with early CAN, 99.1% for patients with severe/definite CAN and 99.3% for all the CAN patients. These results indicate that ECG biometric is possible and QRS complex is not severely affected by CAN with the ability to identify and differentiate individuals.