CAROTID-MAP Carotid Atherosclerotic Risk Assessment Through Imaging andPrediction - PROJECT SUMMARY/ABSTRACT Despite advances in diagnosis, one-fifth of ischemic strokes have no clear cause. A significant proportion of such vascular events are thought to arise from clots that originate in undiagnosed sources outside the brain, including from nonstenosing carotid artery plaques. We have shown that expert human readers can extract several high-risk plaque features of carotid artery plaques on carotid computed tomographic angiography (CTA), such as soft plaque or spotty calcifications, which are strongly associated with ipsilateral cryptogenic stroke. Although there has been a major recent increase in the utilization of carotid CTA in stroke patients and widespread use of CT for vascular event risk stratification in the coronary arteries, CT markers of vascular event risk in carotid disease have not been widely adopted in practice given questions regarding reliability, reproducibility, and whether a single or combination of imaging features is optimal for identifying culprit plaques. Determining which plaque imaging phenotypes are associated with brain infarction is important, as it would enable testing of new treatments to reduce the risk of recurrent stroke. Our overarching objective in this R01 project is to enable clinicians to harness the rich information on stroke risk embedded in carotid CTA scans but which is typically either ignored or too cumbersome to systematically extract in daily practice. Our central hypothesis is that there are reliable imaging-derived plaque features in the carotid artery ipsilateral to the cryptogenic stroke that are different from the contralateral artery. By using a novel within-subject design to extract high-risk features, we will minimize confounding by systemic risk factors and thereby enable robust risk prediction. We have an extensive track record of leading projects in carotid disease and cryptogenic stroke and will leverage this experience to ascertain which expert reader determined carotid CTA based carotid plaque features are most strongly associated with ipsilateral cryptogenic stroke. Next, we will use a large patient dataset from our institution to develop an advanced computer vision algorithm to evaluate these images. Finally, we will evaluate the ability of our algorithms to predict recurrent stroke events in a prospective longitudinal study of patients admitted to our hospital with cryptogenic stroke. A successful outcome of this proposal will make our novel imaging strategy ready for immediate deployment in large-scale multisite clinical trials. Our study promises to lay the foundation of a tool that will enable clinicians to leverage fully the vascular risk data embedded in carotid CTA studies and thereby help reduce the individual and societal burden vascular disease caused by atherosclerosis.
