Machine Learning-Based Diagnosis of Coronary Heart Disease with Clinical and Biomarker Data

Coronary heart disease (CHD) is a common condition associated with high disability and mortality. Although coronary angiography (CAG) remains the gold standard for diagnosis, its accessibility is limited in medically underserved regions or for patients who are ineligible for the procedure. The aim of this study was to develop a prediction model using machine learning algorithms that integrate clinical data and plasma biomarkers for timely and noninvasive CHD diagnosis.

A total of 191 patients who underwent CAG at the First Affiliated Hospital of Henan University between March and October 2023, comprising 151 CHD patients and 40 non-CHD controls, were included. Clinical data and plasma samples were collected; the levels of eight biomarkers (LPA, MCP-1, TNF-α, IL-1β, NRG4, S100A8/A9, HMGB1, and cfDNA) were measured by ELISAs and fluorescence spectrophotometry. The dataset was split into training, test, and validation sets (8:1:1) and standardized. After the SMOTE was applied to the training set, key features were selected via recursive feature elimination with cross-validation (RFECV). Six machine learning algorithms were used to construct the models, and their performance was evaluated on the test set and via 10-fold cross-validation. SHAP was used to interpret the optimal model.

Eight important features were identified. Among the six models, the KNN model performed best on the test set, achieving an AUROC of 80.7%, accuracy of 82.7%, F1‑score of 89.5%, precision of 86.4%, and sensitivity of 92.7%.

A machine learning model integrating clinical and plasma biomarker data has potential for predicting CHD in patients who present with chest discomfort, offering a potential diagnostic aid when CAG is not readily accessible, pending validation in larger, multicenter studies. However, these findings are based on a single-center cohort of 191 patients and require external validation in larger, multicenter populations.