CARDIAC ANOMALIES DETECTION FROM ECG

Abstract

Cardiac diseases are one of the leading cause of death across the world. Identification of many heart diseases involving irregular heartbeat relies upon the reading of an ECG. An ECG is the electrical signal of the heart that is taken using electrodes put on the skin. Ideally 12-lead ECG on the patients and on the chest. The magnitude of the heart is then measured from the twelve different angles and is recorded over a period of time. In this project we propose efficient technique to automatically extract the ECG features.

In this work we are able to identify at least 9 diseases (ischemia peaked, ischemia2, ischemia3, ischemia4, stemi1, stemi2, stemi3, stemi4 and stemi5) from data collected by ECG in raw form or in biometric form. The efficiency of the results is as high as 70% for regular (non-noisy) data.

All the ECG databases used here are from the Pysionet’s [1] online databases.