Year: 2026
Type: book-chapter
Abstract: Accurate and automated detection of cardiac arrhythmias Using electrocardiogram (ECG) data is a significant research challenge.Cardiac arrhythmias have been a major cause of cardiovascular mortality.The work proposes a unique end-to-end deep learning (DL) system for automated arrhythmia classification on the MIT-BIH Arrhythmia Database.The proposed model uses a Learnable Discrete Wavelet Transform (LDWT) layer whose filter coefficients are jointly optimized with the downstream classifier via backpropagation, in contrast to previous approaches that rely on fixed-basis Continuous Wavelet Transform (CWT) scalograms as precomputed, non-trainable features.A hybrid classifier consisting of a one-dimensional Convolutional Neural Network (1D-CNN) for local morphological feature extraction, a Bidirectional Long Short-Term Memory (BiLSTM) network for temporal rhythm modeling, and a self-attention layer to weigh discriminative time steps is fed multi-scale sub-band coefficients generated by the LDWT, concatenated with FFT magnitude features.The AAMI EC57 inter-patient protocol is used to train and assess the entire pipeline, which divides beats into 5 standard categories: Normal (N), Supraventricular ectopic (S), Ventricular ectopic (V), Fusion (F), and Unknown/paced (Q).The novelty of the proposed work lies in the model's ability to independently identify arrhythmia-discriminative frequency bands due to the differentiable wavelet foundation.
Authors: Sumita Lamba, Satender Kumar, Manoj Diwakar
Institutions: Quantum University, Graphic Era University
Language: English
Topic: ECG Monitoring and Analysis
Subfield: Cardiology and Cardiovascular Medicine
Field: Medicine
Domain: Health Sciences
Open Access status: hybrid
OpenAlex ID: W7220839397