AI Driven Medical Image Analysis for Cardiovascular Risk Prediction

Heart attack is a severe cardiovascular event often associated with the clinical symptom of chest pain. The pathophysiology often includes a blood cloth in the coronary arteries, which are the arteries supplying the heart itself with oxygenated blood. When there is a lack of oxygen to a certain area of the heart, that area may die, if not treated in time, causing so called myocardial infarction (MI). MI can involve a small or larger area of the heart, depending on the location and size of the artery occluded. MI is often preceded by anatomical changes in the heart induced by traditional risk factors such as smoking, elevated blood pressure and elevated cholesterol. Additionally, having a MI increases the risk of occlusion of another coronary artery, i.e. getting another MI. Thus, detecting a minor infarction in a patient gives the opportunity for preventive measures against a severe infarction.

A previous MI can alter the shape, appearance, and function of the left ventricle, the heart’s main pumping chamber. The goal of this project is to quantify these changes and extract clinical biomarkers for assessing cardiovascular risk specifically related to the left ventricle. We will achieve this by using state-of-the-art approaches in both deep learning and medical image analysis to model the physical movement and the function of the heart. With these quantifications and clinical biomarkers, we aim to derive the correlation of the ventricle’s physical movement and appearance with later cardiovascular events, especially recurrent MI.

The project utilizes data from the General Copenhagen Population Study, with volumetric CT scans of patients dating back to 2010. Some of these patients have since experienced cardiovascular events providing valuable data for a machine learning based approach to identify biomarkers and their correlation with the cardiovascular events. This data-driven approach holds the promise of a more accurate cardiovascular risk assessment. Ultimately, this project aims to advance scientific understanding of cardiovascular risk and to improve patient diagnosis, potentially saving lives.