Smart healthcare: A novel deep learning based OptGPDCNN framework for heart disease prediction on the IoT platform

• Published in: Biomedical signal processing and control Vol. 112; p. 108594
• Main Authors: Rao, Gorapalli Srinivasa, Muneeswari, G.
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.02.2026
• Subjects: Cardiovascular disease (CVD); Enhanced Ali Baba Forty Thieves algorithm (EAFT); Improved CapsNet; Modified Chameleon swarm algorithm (MCSA); Optimized Global Pooling Dilated Convolutional Neural Network (OptGPDCNN); Improved CapsNet; Enhanced Ali Baba Forty Thieves algorithm (EAFT); Modified Chameleon swarm algorithm (MCSA); Optimized Global Pooling Dilated Convolutional Neural Network (OptGPDCNN); Cardiovascular disease (CVD)
• ISSN: 1746-8094
• DOI: 10.1016/j.bspc.2025.108594

Cardiovascular disease (CVD) remains a leading cause of mortality worldwide, presenting ongoing challenges to global health systems. Despite substantial advancements in research, accurate prediction and diagnosis of heart disease are hindered by factors such as inconsistent data quality, lack of comprehensive medical datasets, and suboptimal use of patient-specific features. To overcome these issues, this study proposes a novel deep learning-based diagnostic framework, the Optimized Global Pooling Dilated Convolutional Neural Network (OptGPDCNN), tailored for accurate and real-time heart disease detection. Data inconsistencies are addressed using robust preprocessing techniques, including missing value imputation, min–max normalization for feature scaling, and categorical encoding to ensure dataset consistency and quality. The proposed framework incorporates an Improved Capsule Network (Improved CapsNet) for advanced feature extraction, enabling effective representation of complex clinical patterns. To enhance model interpretability and reduce feature redundancy, the Enhanced Ali Baba Forty Thieves (EAFT) algorithm is employed for feature selection. Hyperparameter optimization is achieved through a Modified Chameleon Swarm Algorithm (MCSA), ensuring optimal model performance. Moreover, a Self-Paced Ensemble with Auxiliary Classifier GAN (SPE-ACGAN) is integrated to address class imbalance, thereby improving recognition of minority cases and overall generalization. Experimental results on four benchmark CVD datasets show that the proposed framework achieves a prediction accuracy of 99.56 %, F1-score of 99.20 %, sensitivity of 99.17 %, and precision of 99.24 %, outperforming existing state-of-the-art methods. These results demonstrate the effectiveness and reliability of the proposed framework for clinical heart disease diagnosis.