Evaluation of radioanatomical and biochemical features of patients with hyperlipidaemia by conventional angiography

Purpose: The aim of this study is to evaluate the cardiovascular and biochemical changes caused by hyperlipidemia using machine learning (ML) algorithms. Materials and Methods: This retrospective study was performed on conventional coronary angiography images and biochemical findings of 70 patients with hyperlipidemia (Group 1) and 150 individuals without significant coronary artery stenosis (Group 2). Conventional angiography images were retrospectively reviewed, and diameter measurements of the proximal, middle, crux, distal, posterolateral, and posterior descending branches of the right coronary artery were obtained from right anterior oblique cranial projections. The diameters of the left main coronary artery, the proximal anterior interventricular artery, and the proximal circumflex artery were measured from left anterior oblique caudal projections.Results: ML algorithms predicted hyperlipidemia with an accuracy of 0.89-0.98. SHAP analysis identified the initial diameter of the left main coronary artery as the most influential radioanatomical feature and triglyceride level as the most influential biochemical feature contributing to the machine learning model predictions.Conclusions: High rates of hyperlipidemia disease were predicted using radioanatomical and biochemical findings. We believe that these results will contribute to the early diagnosis and early treatment protocol of hyperlipidemia and other related diseases.