Cardiovascular disease detection: A hybrid machine learning-AI framework for personalized diagnosis and risk assessment

• Published in: PloS one Vol. 20; no. 10; p. e0335421
• Main Authors: Tawfeek, Medhat A, Ibrahim Alrashdi, Alruwaili, Madallah, Allahem, Hisham
• Format: Journal Article
• Language: English
• Published: San Francisco Public Library of Science 01.10.2025
• Subjects: Accuracy; Algorithms; Cardiovascular disease; Cardiovascular diseases; Customization; Decision making; Deep learning; Diagnosis; Disease detection; Electronic health records; Electronic medical records; Genomes; Global health; Health care; Health services; Heart diseases; Heart failure; Learning algorithms; Likelihood ratio; Machine learning; Mathematical models; Medical diagnosis; Medical history; Medical imaging; Mortality; Neural networks; Particle swarm optimization; Patients; Polymorphism; Quality of life; Risk assessment; Risk groups; Support vector machines; Wearable computers
• ISSN: 1932-6203; 1932-6203
• DOI: 10.1371/journal.pone.0335421

Cardiovascular disease (CVD) is considered the number one killer disease in the world, underlining the importance of the application of more accurate diagnostic and therapeutic tools. Traditional screening procedures usually do not provide identification and guidance based on individual peculiarities that might result in less than beneficial results. This study seeks to create a hybrid computational framework that synergistically integrates a Support Vector Machine (SVM) classifier, a Particle Swarm Optimization (PSO) algorithm for hyperparameter tuning, and an AI-based interpretation module (SHapley Additive exPlanations, SHAP) to enable early diagnosis and risk assessment beyond various profiling of patients. A mathematical model was developed to provide the framework to deal with the diagnostic complexity of cardiovascular disease. Machine learning (ML) and AI techniques are then used to improve clinical decision-making. The proposed framework employs a variety of forms of patient data, namely electronic health records, medical images, and genomic data, to construct patient models. It utilizes advanced algorithms to enable accurate disease prognosis, identify high-risk individuals for early intervention, and facilitate personalized treatment strategies. This approach will help to eliminate the expense of ineffective therapies, shorten delays in care, and eventually improve patient outcomes and quality of life. Preliminary results on the MIMIC-III clinical database (v1.4) showed that the proposed framework performs better than previous methods by achieving higher accuracy 98.4%, precision 97.5%, recall 96.4%, F1 score 96.9%, and AUC-ROC 97.35%. Moreover, the sensitivity 96.4%, specificity 98.7%, and a low negative likelihood ratio (0.036) of the proposed framework demonstrate its ability and power in identifying high- and low-risk patients. The hybrid ML-AI framework provides an improved way for early detection of cardiovascular disease, which helps in personalizing treatments for patients. It also enables healthcare delivery through its combined predictive power to improve healthcare service.