Cardiac Arrhythmia Detection Using CNN

Abstract

The diagnosis of cardiac arrhythmias can be a tedious process when done by hand and could benefit greatly from computer automation. To this end, an algorithm was developed to distinguish between normal heart beats and abnormal arrhythmic beats in an PCG Recording. First an algorithm was developed to find the location of QRS complexes in the PCG Recording. Principal component analysis was performed using the area around the QRS complex. 20 of the resulting principal components were used to train a simple linear classifier to distinguish between normal and abnormal beats. The classification performed reasonably well with a sensitivity of 85.4% and specificity of 91.7%. More sophisticated signal processing and classification techniques could be applied to improve these numbers, but the algorithm is a good starting point.

Keywords: CNN, PCG Recording, MIT-BIH Arrhythmia Dataset, PYTHON