A low-cost machine learning-based cardiovascular/stroke risk assessment system: integration of conventional factors with image phenotypes

Annually, about 17.7 million people are affected by cardiovascular (CV) diseases including heart attack and stroke events (1). Atherosclerosis is the major contributor to such CV/stroke events (2). One way of predicting the occurrence of these events is by performing risk assessment using conventional risk factors (CRF) that are responsible for the growth of atherosclerosis (3). However, CRF alone does not explain the elevated risk of CV/stroke events (4). This is because of the morphological variations in the atherosclerotic plaque that cannot be captured using CRF alone but which can easily be assessed using imaging modalities (5,6). Thus, there is a need to look beyond the scope of CRF and search for preventive healthcare solutions that can provide an accurate routine risk assessment at an affordable cost