Attentional Seq2Seq Encoder-Decoder and Next Generation Reservoir Computing for Heartbeat Classification

Cardiac dysfunctions are a global concern due to the high number of deaths it causes worldwide, as noted by the World Health Organization. The electrocardiogram signal is renowned for its simplicity, safety, and ability to detect heart conditions and identify cardiovascular diseases. Early detection of arrhythmias and anomalies by automatic classifiers can significantly support physicians' work. Here, we aim to develop a heartbeat classifier by applying a cutting-edge technique, the next-generation reservoir computing, in conjunction with an attentional seq2seq encoder-decoder classifier. The classifier performance was evaluated using the MIT-BIH database, which adheres to the guidelines provided by the Association for the Advancement of Medical Instrumentation (AAMI), focusing only on the inter-patient framework. For the challenging <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$S$</tex> category, the classification performance in sensitivity and positivity are 79.80% and 91.53% for inter-patient cases, respectively. In terms of overall accuracy, we get 99.29%. This cutting-edge technique provides computing efficiency and could replace the data-hungry convolutional neural network.