کنفرانس مهندسی برق ایران

صفحه اصلی / سی و چهارمین کنفرانس بین المللی مهندسی برق

A Two-Stage Hierarchical Deep Learning Model for Inter-Patient Arrhythmia Classification

نویسندگان :

Mohammad Ahmadabadi¹ Ali Sadr²

1- دانشگاه علم و صنعت ایران 2- دانشگاه علم و صنعت ایران

کلمات کلیدی :

Arrhythmia،Electrocardiogram،Inter-patient،Deep learning،Bagging

چکیده :

The significant inter-individual variability and class imbalance in electrocardiogram (ECG) data pose major challenges for reliable inter-patient arrhythmia classification. This study proposes a hierarchical two-stage deep learning framework to discriminate among three heartbeat types: Normal (N), Supraventricular Ectopic Beat (SVEB), and Ventricular Ectopic Beat (VEB). In the first stage, a simplified ResNet architecture effectively separates VEB from non-VEB beats by leveraging their distinct morphology. In the more challenging second stage, we distinguish SVEB from Normal beats by combining a set of informative RR-interval-based temporal features with a verification network based on a CNN-Residual Bidirectional LSTM (CNN-ResBiLSTM) model. To address severe class imbalance in both stages, a tailored bagging strategy constructs balanced training subsets, and final decisions are made via robust majority voting. The proposed method is evaluated on the MIT-BIH Arrhythmia Database under the inter-patient paradigm. It achieves an overall accuracy of 96.40%, with well-balanced sensitivities of 89.68% for SVEB and 96.92% for VEB, demonstrating superior performance, particularly in detecting SVEB beats, compared to existing state-of-the-art approaches.