ECG Classification

The aim of this project was to use machine learning to create a classification algorithm to distinguish between normal heart rhythms and premature ventricular contractions (PVCs). Different algorithms were researched and artificial neural networks (ANN) were found to give results with the highest accuracy. An MLP neural network was chosen for this project as it is one of the simplest ANN while also giving excellent results. The signal was filtered using a wavelet transform method before a total of 9 different features were extracted for input into the MLP. At first, the algorithm extracted normal and PVC beats only, before it was altered to include all types of beats, with non-normal and non-PVC beats all grouped into a third classification – ‘other’.