a single-lead explainable-AI guided multiresolution network with train-only quantitative features for trustworthy ECG arrhythmia classification

View the PDF file for the paper entitled ExGNET: an acceptable multi -guideline connection network in order to clarify it with quantitative features of the train only to classify ECG cardiac arrhythmias, written by Tushar Talukder Showrav and 2 other authors

PDF HTML (experimental) view

a summary:deep learning has greatly pushed the performance of ECG, yet its clinical adoption remains impeded by the challenges on the ability to interpret and publish on the edge restrictions restricting resources. To bridge this gap, we suggest an EXGNET, a new and reliable ECG cardiac network specifically designed for single signals, specially designed to achieve a balance between high accuracy, interpretation, and compatibility with the edge. EXGNET integrates Xai’s supervision during training through the cross -bound, which directs the model to the ECG clinical areas, similar to the concentration of the cardiologist. The motivation behind this supervision is done through the ground that was automatically created, derived through an innovative approach based on the fluctuation of the heart rate, without the need for handicrafts. To enhance the classification accuracy without prejudice the simplicity of publishing, we merge the quantitative ECG features during training. This enriches the model with multi -field knowledge but is excluded during inferring, while maintaining the lightweight model to spread the edge. In addition, we offer an innovative size mass to capture short and long -term signal features efficiently while maintaining mathematical efficiency. A strict evaluation of Chapman and Ninabo data collections educates ExGNet superiority, which achieves an average accuracy of five times by 98.762 % and 96.932 %, and F1 grades by 97.910 % and 95.527 %, respectively. Comprehensive eradication studies and evaluation of quantitative and qualitative interpretation confirm that Xai’s guidelines are central, which clearly enhances the concentration of the model and reliability. In general, EXGNET sets a new standard by combining high -performance cardiac arrhythmia classification with interpretation, and paves the way for more ECG portable health control systems.