Arrhythmia Classification Using Alexnet Model Based on Orthogonal Leads and Different Time Segments

In this paper, cardiovascular disorders were classified using the pretrained model Alexnet. The Physikalisch-Technische Bundesanstalt (PTB) database was used with a sample size of 50 recordings out of 549. Five classes were included: healthy control, myocardial infarction, cardiomyopathy, bundle branch block, and dysrhythmia. First, standard preprocessing was applied to the orthogonal leads, and then the heart vector magnitude was calculated. After that, the signals were divided into segments of different lengths. Finally, the segments were transformed to scalograms using the continuous wavelet transform (CWT). The 1-sec, 2-sec, and 3-sec scalograms were 380, 190, and 120 units, respectively. The accuracy, precision, recall, and f1-score of the validation were calculated after training the model on the obtained scalograms. The best performance was achieved using the scalograms of 3-sec length (0.98, 0.9818, 0.98, and 0.9809, respectively). These results have successfully validated that using the Alexnet model on scalograms achieves excellent classification accuracy for the selected classes.