Cookies on this website

We use cookies to ensure that we give you the best experience on our website. If you click 'Accept all cookies' we'll assume that you are happy to receive all cookies and you won't see this message again. If you click 'Reject all non-essential cookies' only necessary cookies providing core functionality such as security, network management, and accessibility will be enabled. Click 'Find out more' for information on how to change your cookie settings.

Echocardiography is the first-line imaging test in virtually all cardiovascular conditions, due to the advantages of low cost, widespread diffusion, portability, versatility, low environmental footprint, and radiation-free nature. However, the overwhelming depth of information that can now be extracted from echocardiograms can be confusing for the cardiologist. In a 15-30-minute examination we obtain unique data on cardiovascular anatomy, function, flow, structure, coronary supply, and lung appearances, with different techniques including M-mode, 2D, color-, continuous wave-, pulsed-, tissue-Doppler, contrast, 3D, and deformation (strain) imaging. Yet the technique remains reliant on the observer and prone to high inter-observer variability for key aspects of diagnosis from left ventricular volumes to ejection fraction and regional wall motion analysis, with significant drop-off of performance in poor quality echocardiograms. Artificial intelligence provides a solution for automated handling of this wealth of imaging information, both at rest and during stress echo. In recent years, deep learning and hybrid imaging approaches have been developed with four main aims: (1) to make objective what is currently done by the “naked eye, " for instance, regional wall motion analysis, or by “hand measurements, " for instance, ejection fraction from left ventricular volumes; (2) to detect what is undetectable by standard analysis and natural intelligence, for instance, myocardial fibrosis or carotid plaque histology from standard gray level analysis of 2D image texture; (3) to combine in single image information coming from different imaging sources, for instance, fluoroscopy and transesophageal echocardiography in hybrid imaging during valvuloplasty; (4) to mine big data with techniques such as network analysis to identify interconnected variables and thereby optimize risk stratification for individual patients. This approach is effectively illustrated in the use of echocardiography in coronary artery disease and heart failure; (5) clusters of information is obtainable during stress echo (ABCDE protocol) that can be used in data analysis to identify ischemia and regional wall motion (step A), lung water (step B), myocardial scar or necrosis (step C), coronary microcirculation (step D) or non-imaging, EKG-based, chronotropic reserve (step E). Each cluster can be combined with clinical and other imaging data to identify specific phenotypes that are likely to require targeted therapeutic approaches.

Original publication





Book title

Hybrid Cardiac Imaging for Clinical Decision-Making: From Diagnosis to Prognosis

Publication Date



29 - 38