Integrating multi-modal data for enhanced heart disease prediction: an HPC-based clinical decision support system

Introduction Heart disease remains one of the leading causes of mortality worldwide, necessitating accurate and early prediction systems. Traditional methods often fail to effectively integrate diverse clinical data sources, creating a need for advanced, scalable solutions.Objectives This study aims to develop a High-Performance Computing (HPC)-based Clinical Decision Support System (CDSS) for improved heart disease prediction using multi-modal healthcare data.Methodology The proposed system utilizes the Symile MIMIC dataset, which includes Electronic Health Records (EHR), time-series ECG signals, and medical imaging data. Categorical variables are encoded using label encoding techniques. Feature extraction is performed using specialized models: AlexNet for medical images, Gated Recurrent Units (GRUs) for ECG time-series data, and a Feedforward Neural Network (FNN) for EHR data. The extracted features are integrated through a weighted feature fusion approach to enhance predictive performance. The model is implemented in Python and evaluated using standard performance metrics.Results Experimental findings demonstrate high predictive performance, achieving accuracy values of 0.987 and 0.988 for 70/30 and 80/20 training-testing splits, respectively, along with strong sensitivity, specificity, and precision.Practical Implications The proposed system offers a reliable and scalable solution for clinical decision-making, enabling early detection and improved management of heart disease in real-world healthcare settings.