Clinical Implications of Machine Learning, Artificial Intelligence, and Radiomics in Cardiac Imaging

Purpose of review Rapid advancements in technology and electronic medical record systems have given rise to massive amounts of cardiac imaging data with the potential to alter medical practices. The rise of machine learning (ML) and radiomics – the concept that images contain invaluable data regarding disease processes beyond what the eyes can see – promises increased precision and accuracy to the current standard of care. Recent advancements in major cardiac imaging modalities, such as echocardiography, cardiac CT and cardiac MRI, have uncovered promising diagnostic and prognostic information through the application of ML. Recent findings In echocardiography, ML has been successfully applied to identify views, make right- and left-sided heart measurements, and detect certain diseases like hypertrophic cardiomyopathy. Application of ML in cardiac CT has seen success in quantifying coronary plaque burden, identifying significant coronary stenosis, and predicting mortality in coronary artery disease (CAD). In cardiac MRI, efforts have been made to automatize segmentation for chamber measurements and detecting fibrosis. For nuclear imaging, ML has been applied to not only make measurements like the left ventricular ejection fraction, but also to identify perfusion abnormalities and predict obstructive CAD. Summary While there are many milestones still to be reached before ML can be widely integrated to current clinical practices, there is optimism for ML to advance the field of cardiovascular imaging through enhanced image analysis and improved efficiency.