Computerized Visualization and Prediction of Coronary Artery Ischemia

Weill Medical Coll Of Cornell Univ, New York NY

Investigators

Linked publications & trials

Abstract

? DESCRIPTION (provided by applicant): Coronary artery disease (CAD) is the principal cause of morbidity and mortality. In the US, CAD affects >16 million adults and accounts for >1/3 deaths annually. Recently, computed tomography (CT) has emerged as a non-invasive option for imaging coronary arteries and myocardial perfusion. Nevertheless, diagnosis of ischemic CAD based on CT alone is less robust due to the lack of lesion-specific physiologic information. Computational fluid dynamics (CFD) has been applied to image-based modeling of patient-specific geometry from CT data, which now allows for non-invasive calculation of coronary pressure, flow and shear stress, and thus lesion-specific evaluation of coronary ischemia in higher diagnostic accuracy than the strategy based on CT alone. There is a pressing need to incorporate CT imaging, image-based modeling, and CFD into clinical practice for better diagnosis of coronary ischemia. However, progress has been thwarted by three major challenges: (1) lack of integrated tools to visualize immense data to assist diagnosis, (2) inabiliy to quantify and select salient anatomic and physiologic features for optimal prediction of ischemia, and (3) lack of comprehensive clinically relevant evaluation. In this proposal, we will develop and evaluate a novel computerized system to improve visual and predictive assessment of coronary ischemia from CT imaging and CFD. To accomplish this goal: (1) We will create tools to visualize coronary anatomy, physiology and myocardial perfusion for routine diagnosis assistance; The evaluation will be performed by comparing the diagnostic performance of two experienced cardiologists using our visualization tool and the conventional workstation using invasive ground truths; (2) We will develop methods to automatically quantify and select salient anatomic and physiologic features for maximizing predictive accuracy using machine learning. The evaluation will be assessed through cross-validation on a large existing database with respect to invasive ground truths. If successful, our developments will provide a new computerized system to assist the diagnosis of coronary ischemia by visualizing the totality of anatomic and physiologic findings over current unassisted approaches, and predict ischemia by machine learning methods that are superior to current heuristic techniques, and ultimately accelerate the translation of diagnostic performance gain into routine clinical practice.