ECG Arrhythmia Classification Using Non-Linear Features and Convolutional Neural Networks

Heart arrhythmia classification algorithms are an important tool for continuous monitoring of patients at risk. By analyzing 12 ECG-lead signals, these algorithms can help us to diagnose different cardiac diseases. Thus, our team, CardioLux proposes a novel approach to denoise ECG signals and classify the nine cardiac arrhythmias using Convolutional Neural Networks (CNN) trained with more than 6700 ECG recordings as defined in the Physionet Challenge 2020. First, a noise removal process is initially applied with Savitzky-Golay smoothing filters. Secondly, we extracted 300 features, clustered in time, frequency, and time-frequency groups, including linear and non-linear characteristics. Thirdly, 27 features were carefully selected to train our model using our feature-selection procedure. Finally, we implemented CNN to reduce noise-aware signals and bias during our training model. The proposed methodology developed so far was tested with 10-fold cross-validation on the training set and yielded a challenge score of 0.22. Overall, the feature extraction and selection stage can help improve the performance of the heart arrhythmia training model by selecting the best characteristics. Our model keeps a high level of interpretability, demonstrating a high range of possibilities that can be configured using CNN.