Abstract

Aim: Hypertension is a widespread cardiovascular disease and a major risk factor for heart conditions. This study aimed to identify vascular, pathological, demographic, and radioanatomical differences between hypertensive patients undergoing conventional coronary angiography and healthy individuals, using Machine Learning (ML) algorithms.
Materials and Methods: A retrospective analysis was conducted on coronary angiography images from 109 hypertensive patients (Group 1) and 96 healthy individuals (Group 2), aged 20–80 years. Diameters of the proximal, middle, crux, distal, posterolateral (PL), and posterior descending (PD) branches of the right coronary artery (RCA) were measured with right anterior oblique (RAO) cranial projection. Measurements of the left coronary artery (LMCA), proximal anterior interventricular artery (LAD), and proximal circumflex artery (CX) were obtained using left anterior oblique (LAO) caudal projection. Selected biochemical parameters were also retrieved from the hospital archive.
Results: ML algorithms achieved hypertension prediction accuracies ranging from 0.73 to 0.85. SHAP analysis of the Random Forest (RF) model indicated that the PL branch of the RCA had the greatest radioanatomical impact, while estimated glomerular filtration rate (eGFR) was the most significant biochemical predictor.
Conclusion: The findings demonstrate that hypertension can be predicted with high accuracy by combining coronary artery diameter measurements with biochemical parameters through ML algorithms. This integrative approach highlights the potential of modern computational methods in cardiovascular risk assessment.

Key words: Hypertension, conventional angiography, coronary artery diameter, machine learning algorithms