Automated Cardiac Health Diagnosis: A Time-Domain Approach

Cardiological problems are one of the leading causes of human fatality. Electrocardiogram is a major noninvasive tool for monitoring heart conditions. The human vision is not suitable to identify the minute changes in Electrocardiogram wave amplitude and time intervals; hence an automatic diagnostic tool is necessary for precise abnormality detection. This paper presents a classification method to classify seven heartbeat conditions-normal and six classes of abnormalities. The algorithm implements a time domain approach to obtain the statistical features from the Electrocardiogram beats extracted from the arrhythmia database. This objective of this work is to find the suitability of time domain features to arrhythmia classification with machine learning. The statistical features are extracted from raw ECG signal, the time derivative, time integral and 5-point first derivative stencil of the ECG data. The cardiac abnormality classification is implemented with Support Vector Machine. The attained classification accuracy is upto 93% for chosen input feature pairs for binary Support Vector Machine.