Intelligent Decision Support System Facilitating Early Detection of Cardiovascular Disease

BackgroundCardiovascular disease (CVD) remains the primary cause of mortality worldwide, with higher fatality rates in India. Multi-modal diagnostics integrating electrocardiogram (ECG) analysis, cardiac biomarkers, and region-specific insights can enhance early detection and clinical triage. MethodsIn this cross-sectional study, ECGs along with clinico-epidemiological data were collected from two regions-North and South Indian cohorts. ECGs were clinically annotated and were analyzed using Earth Movers Distance (EMD) and Support Vector Machine (SVM). A 2 ml blood sample from both cohorts was also collected for molecular analysis. Age and sex matched serum samples were curated to detect four novel cardiac biomarkers. ROC curve analysis was used to assess the relationship between the biomarker index and the probability of ECG abnormality. GraphPad 10.5 and Stata version 17.0 was used for statistical analysis. ResultsA total of 774 clinically annotated ECGs were collected (498 from North and 276 from South Indian cohort). For the molecular analysis, 34 serum samples were curated the North Indian cohort and 54 from the South Indian cohort. The proposed SVM algorithm reported combined accuracy of 90% for non-biomarker analyzed ECGs and 94% accuracy for biomarker analyzed ECGs. South Indian cohort with abnormal ECGs reported significantly higher GDF-15 level (1145.8 {+/-} 476.7 pg/ml) (p<0.001) as compared to North Indian cohort. GDF-15 also exhibited potential prediction ability for detecting abnormal ECG findings (AUC-0.8853). ConclusionOur study demonstrates GDF-15 as a potential biomarker for detecting region-specific abnormal ECG patterns. The proposed multimodal platforms high diagnostic accuracy reinforces the efficacy of the artificial intelligence-driven approach in early screening and triaging CVD cases across a diverse Indian population. FundingThis study was funded by the Indian Council of Medical Research (ICMR) under its Ad-hoc Research Grant Scheme (Proposal ID: 2020-3500).