Citation

Sayed Ismail, Sharifah Noor Masidayu and Abdul Razak, Siti Fatimah and Ab. Aziz, Nor Azlina (2026) Explainable Arrhythmia Detection from ECG Signals Using CNN and GRAD-CAM. In: 14th International Conference on Information and Communication Technology, ICoICT 2026, 30 July 2026 - 31 July 2026, Bandung.

Text 14.pdf - Published Version Restricted to Repository staff only Download (1MB)

Abstract

Arrhythmia is a major global health threat that acts as a critical indicator for underlying cardiovascular disease, which remains the leading cause of death among Malaysians. Alarmingly, this condition no longer affects only the elderly population; it has increasingly contributed to mortality among individuals aged 41 to 59, with approximately 5,380 deaths reported in Malaysia in 2024 alone. While some arrhythmias are minor, others are life-threatening and can increase the risk of mortality and morbidity. Existing studies have significantly contributed to improving diagnostic processes; however, most research faces difficulties to resolve high computational cost of deep models and the "black box" problem in clinical settings. A shallow CNN with Grad-CAM solves this gap by providing a lightweight, transparent decision-support system. This study proposed an arrhythmia detection using a 3-layer CNN with Grad-CAM for interpretability purposes. The training model acquired 76.69% accuracy, while the testing model acquired 57.83% accuracy. This disparity was noted due to the imbalanced distribution of ECG samples in both the training and testing sets. Despite this, the availability of visual explanation from GRAD-CAM successfully highlights the ECG regions that contribute the most to the abnormal condition detection. The findings of this study are expected to create a "white-box" support tool and contribute toward building more trustworthy and interpretable machine learning-based diagnostic support systems, thereby increasing confidence and acceptance among medical practitioners in clinical decisionmaking.

Item Type:	Conference or Workshop Item (Paper)
Uncontrolled Keywords:	Electrocardiogram, arrhythmia, convolutional neural network, grad-cam
Subjects:	Q Science > QA Mathematics > QA71-90 Instruments and machines > QA75.5-76.95 Electronic computers. Computer science
Divisions:	Faculty of Engineering and Technology (FET) Faculty of Information Science and Technology (FIST)
Depositing User:	Ms Rosnani Abd Wahab
Date Deposited:	02 Oct 2026 04:36
Last Modified:	02 Oct 2026 04:36
URII:	http://shdl.mmu.edu.my/id/eprint/16834

Downloads

Edit (login required)