Screening of Ischemic Heart Disease based on PPG Signals using Machine Learning Techniques.

The increasing rate of cardiac ailments has led to the rise in the scrutinization of ones cardiac health. The prevalent techniques for detecting heart diseases are costly and require expert supervision as well as modern equipment. Thus there is a need for an alternative low cost and easily available technique. Finger-tip photoplethysmography (PPG) signals can be used for identifying Ischemic Heart Disease (IHD). This technique of screening the disease will be very helpful to the inhabitants of remote, underdeveloped and unprivileged areas. Time-domain analysis of the signal was done for extracting different features. Segregation of diseased and healthy subjects was performed using Decision Trees, Discriminant Analysis, Logistic Regression, Support Vector Machine, KNN, and Boosted trees. Ten different performance metrics was studied using the confusion matrix. After analysis, the accuracy, sensitivity, specificity, and precision of 0.94, 0.95, 0.95 and 0.97 respectively was obtained using Boosted tress classifier. ROC and AUC were calculated to establish the robustness of the classification methods for determining IHD patients.