Skip to main content

DBRH-5: Legendre Polynomial-Based Attention Mechanism in Dual-Branch ResNet Hybrid-5 Towards Arrhythmia Diagnosis.

Dr ARHIN, Joseph Roger
Lecturer
  0503681064
  jrarhin@uew.edu.gh

Authors
Arhin, J. R., Zhang, X., Coker, K., Sam, F., Apawu, J., & Ribeiro, N. N. D.
Publication Year
2024
Article Title
DBRH-5: Legendre Polynomial-Based Attention Mechanism in Dual-Branch ResNet Hybrid-5 Towards Arrhythmia Diagnosis.
Conference Title
International Conference on Computational Science and Computational Intelligence
Publisher
Springer Nature
Place
Las Vegas, USA
Abstract

This study presents a deep learning (DL) model, Dual-branch ResNet Hybrid-5 (DBRH-5) with a novel Legendre Polynomial-based Attention (LegPoBA) mechanism, for accurate ECG arrhythmia classification. Leveraging the orthogonal properties of Legendre polynomials, LegPoBA enhances DBRH-5’s ability to capture intricate patterns in detecting subtle variations in ECG signals. Extensive experiments revealed the remarkable success of DBRH-5 over existing state-of-the-art models, attaining impressive scores. A hypothesis test further validates the statistical significance of the improvements. LegPoBA increased the model’s performance by 4.35% accuracy and reduced the loss rates by less than 1%. The DBRH-5 with LegPoBA model provides a reliable tool for automated arrhythmia diagnosis, significantly improving AI-driven healthcare.

© 2019 University of Education, Winneba