A Comparative Classifier-Based Approach of an AI-Driven Clinical Decision Support System for Accurate Diagnosis of Ischemic Heart Disease

Background and Aims: Ischemic Heart Disease (IHD) is a leading cause of mortality worldwide and is often misdiagnosed due to overlapping symptoms with other cardiovascular conditions. This study aimed to develop an Artificial Intelligence (AI)-driven Clinical Decision Support System (CDSS) to aid in the accurate diagnosis of IHD using machine learning classifiers. Methods: A retrospective dataset comprising ۸۰۰ clinical records was used, with ۷۱۲ cases retained after preprocessing. The data included ۱۶ attributes such as age, sex, blood pressure, cholesterol, diabetes status, and personal habits. Attribute selection was informed by expert consultation and benchmark UCI datasets. Using Weka ۳.۷.۰, ۵۹ AI classification models were initially evaluated across multiple classifier families including Bayesian, tree-based, neural networks, rule-based, lazy learners, and support vector machines. Model performance was assessed using ۱۰-fold cross-validation with sensitivity, specificity, accuracy, precision, F-measure, kappa statistics, and ROC area as key metrics. The top-performing algorithms were integrated into the CDSS prototype and evaluated against physician diagnoses and gold-standard ECG-based diagnoses. Results: The KSTAR classifier demonstrated the highest performance with an accuracy of ۷۹.۳۲%, sensitivity of ۸۹%, specificity of ۷۹%, and ROC of ۰.۹۹۵. Other models such as IBk, RandomTree, and MLP also exhibited high diagnostic reliability. The CDSS showed substantial agreement with physician impressions (k = ۰.۴۵, p < ۰.۰۰۱), indicating its potential utility in real-world clinical settings. Conclusion: AI-based CDSS can significantly enhance diagnostic accuracy for IHD and support clinical decision-making, particularly in resource-limited settings.

کلیدواژه ها:

Clinical Decision Support Systems ، Ischemic Heart