Design of an iterative method for enhanced early prediction of acute coronary syndrome using XAI analysis

Abstract

<p>The escalating prevalence and acute manifestations of Acute Coronary Syndrome (ACS) necessitate advanced early detection mechanisms. Traditional methodologies exhibit limitations in predictive accuracy, sensitivity, and timeliness, thus hindering effective intervention and patient care management. This study introduces a comprehensive machine learning-based approach to surmount these constraints, thereby enhancing early ACS prediction capabilities for different scenarios. Addressing data integrity, the methodology encompasses rigorous data preprocessing techniques, including advanced missing value imputation and outlier detection, to ensure dataset reliability. Feature selection is meticulously conducted through a recursive feature elimination and correlation analysis, thereby distilling critical predictive indicators from extensive clinical datasets. The study harnesses diverse algorithms—Support Vector Machines, Logistic Regression, Gradient Boosting Machines, and Deep Forest—tailored for nuanced ACS detection, balancing simplicity with computational depth to optimize performance metrics. The proposed model exhibits a superior predictive proficiency, as evidenced by significant improvements in precision, accuracy, recall, and reduced prediction delay compared to the existing approaches. The Logistic Regression coefficients and the SHapley Additive exPlanations (SHAP) values provide interpretative insights into the risk factor significance, facilitating personalized patient risk assessments. Furthermore, the study pioneers a clinically applicable risk scoring system, which is thoroughly evaluated through sensitivity, specificity, and positive predictive value metrics. Implications of this research extend beyond theoretical advancement, offering tangible enhancements in ACS predictive analytics. The enhanced model promises improved patient outcomes through timely and accurate ACS detection, thus optimizing healthcare resource allocation. Future research directions are identified, which advocate for the exploration of novel risk factors and the application of cutting-edge machine learning techniques to foster inclusivity and adaptability in diverse healthcare settings.</p>