Fractional Quokka swarm optimization enabled Hierarchical Convolutional Neural Network for IoT Driven heart disease detection and monitoring.

Heart disease remains one of the leading causes of mortality globally, resulting in millions of deaths annually and imposing a significant burden on healthcare infrastructure and clinical management systems. This alarming prevalence underscores the critical need for highly accurate, timely, and efficient detection methods to enable early intervention and reduce mortality rates. Moreover, classical detection models had limited adaptability to heterogeneous clinical data and an inadequate capability to model complex, non-linear relationships inherent in physiological signals and patient records. To solve such complexity, the heart disease detection is done using a novel model named Fractional Quokka swarm optimization-enabled Hierarchical Convolutional Neural Networks (FQSO_HCNN) in this research. The system model for Internet of Medical Things (IoMT) is established, and each patient is registered in IoT devices. Subsequently, the collected data is transmitted to a secure cloud server, where it is stored and utilized for performing heart disease detection. Initially, the heart disease data is preprocessed by dual normalization, which enhances data quality and consistency. Afterwards, feature fusion is done based on Jensen difference with MobileNet, which allows the integration of handcrafted and deep features to capture both global patterns and local variations in the data. This fusion improves the feature discriminability, providing a richer input representation for the next stage. To address class imbalance, adaptive synthetic sampling (ADASYN) is applied, which generates more informative minority samples. Finally, the Hierarchical Convolutional Neural Network (HCNN) is used for detecting heart disease, leveraging its layered architecture to model complex relationships in the data. Also, the hyperparameters are optimized using the proposed Fractional Quokka Swarm Optimization (FQSO), which combines Quokka Swarm Optimization (QSO) with Fractional Calculus (FC) to balance exploration and exploitation. The public benchmarking Heart Disease Database is used for evaluating the proposed model. Furthermore, FQSO_HCNN achieved Matthews Correlation Coefficient (MCC), False Omission Rate (FOR), True Positive Rate (TPR), and accuracy of 0.946, 0.051, 95.78 %, and 96.52 %, respectively.