Detect if a patient has Atrial Fibrilation using ECG signals using different Machine Learning classifiers
Using dataset from the Physionet Challenge 2017 (https://www.physionet.org/challenge/2017/) we classify an ECG signal to one of 4 categories:
- normal sinus rhythm
- atrial fibrillation (AF)
- an alternative rhythm
- or is too noisy to be classified
At first we preprocess an ECG signal to extract the required features for the classification process, we accomplish this by using a python package(neurokit). After extracting the required features we find the importance of each feature using Random Forest and we use classification methods from the sklearn package to classify a signal. The classifiers we used are Random Forest,SVM, KNN, Naive Bayes using Grid Search for parameters tuning. We estimated the number of features we need for better accuracy in the classification.