Cardiac Diagnosis based ECG Images Classification System using Convolution Neural Network

Electrocardiogram (ECG) is a widely-used for the diagnosis of heart disease and the cardiac arrhythmia classification for an efficient clinical approach. The ECG has the random signal nature. Thus, acquiring the accurate diagnosis of ECG becomes a challenging task. To analyze the ECG signal in the time domain, many machine learning approaches are used to analysis in ECG signals. However, these methods take the long calculation time because of its' preprocessing is complex and need to use the features extraction and selection methods to get the accurate heart diseases. ECG signals are one-dimensional signals to be processed while CNNs are better applied to multiple patterns of biomedical image recognition applications. Convolutional neural networks (CNNs) can use the arrhythmia images of ECG to find the cardiac diagnosis via automatic feature extraction of CNN. In this study, the morphology of ECG images is focused into a cardiac arrhythmia classification by processing the CNNs. The proposed CNN model demonstrates the results of accuracy that are efficient for irregular heartbeats or arrhythmias detection like atrial fibrillation and flutter. The proposed model is trained and tested on the arrhythmia database getting from the ECG library, the model achieved higher performance than the other machine learning method (SVM) for ten arrhythmia classification for cardiac diagnosis (the training and validation accuracy are 98.6% and 92% than 75% of SVM).