International Journal of Scientific Research and Engineering Development

This paper presents an innovative framework for predicting high-risk cardiac arrhythmia using an Optimized Deep Active Learning (ODAL) model. The proposed approach integrates the strengths of deep learning, fuzzy logic, and active learning to overcome key limitations of conventional arrhythmia detection systems, including low generalization capability and high dependency on large annotated datasets.
The ODAL framework introduces an uncertainty-based active sampling mechanism that identifies and selects the most informative electrocardiogram (ECG) data instances for model training. This process minimizes annotation costs while ensuring robust learning from limited labeled data. Furthermore, fuzzy optimization is employed to fine-tune model parameters and handle the inherent uncertainty in physiological signals, enhancing prediction stability and interpretability.
Experimental evaluations conducted on real-world ECG datasets demonstrate the model’s superior performance. The ODAL model achieved an F1-score of 90% for non-sinus rhythm detection and an overall classification accuracy of 86%, outperforming several existing deep learning-based approaches in both precision and diagnostic reliability.
By combining deep learning intelligence with active and fuzzy learning principles, the proposed ODAL framework delivers an intelligent, automated, and reliable solution for early arrhythmia diagnosis. This system effectively assists healthcare professionals in clinical decision-making, promoting preventive care, timely intervention, and improved patient outcomes.