Deep Learning Based Hybrid Residual Attention and Echo State Network for High-Accuracy Heart Disease Prediction

Background Early and accurate prediction of ischemic heart disease (IHD) is essential for reducing mortality and enabling timely intervention. Misdiagnosis can lead to severe health outcomes, emphasizing the need for robust and intelligent predictive models. Deep learning approaches have shown strong potential in identifying hidden patterns in medical data and aiding clinical decision-making. Methods This study proposes a novel Hybrid Residual Attention with Echo State Network (HRAESN) model that integrates Attention Residual Learning (ARL) with Echo State Networks (ESN) to enhance feature extraction and temporal data learning. The hybrid model is designed to refine feature attention through residual learning while leveraging ESN for efficient time-series prediction. Two publicly available benchmark datasets were used for evaluation: the Kaggle Cardiovascular Disease dataset comprising 70,000 instances and the UCI Heart Disease dataset containing 303 instances. Missing values in both datasets were handled using a multiple imputation technique tailored for ischemic heart disease. Model performance was assessed using standard classification metrics, including accuracy, sensitivity, specificity, precision, recall, and F-measure. Results The proposed HRAESN model demonstrated superior classification performance compared to traditional machine learning and deep learning approaches. It achieved an accuracy of 98.4% on the Kaggle dataset and 97.7% on the UCI dataset. Additionally, the model showed high sensitivity and specificity, indicating strong diagnostic capability and reliability in identifying both diseased and non-diseased cases. Conclusions The HRAESN model effectively combines the strengths of residual attention mechanisms and echo state networks, resulting in improved accuracy and stability for ischemic heart disease prediction. Its strong performance on benchmark datasets confirms its potential as a valuable clinical decision support tool for early detection of IHD. Future work may focus on optimizing model complexity and integrating real-time medical IoT data to enhance practical deployment in healthcare systems.