Project 460140

As the most prevalent and most deadly valvular heart disease, Aortic Stenosis (AS) is a significant public health problem in Canada. The prevalence of this valvular condition increases with age, with only 1/3 of the patients remaining alive at 5-years following diagnosis. Fortunately, the rapid downhill clinical course and significant mortality associated with this valvular disorder is avoidable with recent advancements in treatment of AS using aortic valve replacement. Recent studies strongly suggest that early detection and timely intervention of clinically significant AS are crucial to achieve favourable clinical outcomes. Comprehensive echocardiography is the clinical standard for the diagnosis of AS. However, since it heavily relies on acquisition of high-quality Doppler imaging and accurate characterization of aortic valve in echo, its usage is hindered for primary care physicians. Referring all patients to comprehensive echocardiography exams is also not practical, given the significant burden it will add to the already strained healthcare system. We propose a paradigm shift in screening and early detection of clinically significant AS in point-of-care settings by streamlining the ultrasound imaging of the aortic valve. Supported by abundant evidence from clinical literature, the experience of our clinical co-applicants, as well as substantial preliminary results on more than 2,000 patient records, our goal is to demonstrate that clinically significant AS can be detected with very high sensitivity and specificity through direct analysis of ultrasound videos using machine learning. We expect that the proposed technology will help the primary care physicians to timely identify and prioritize patients that should be referred for a comprehensive echocardiogram to confirm AS diagnosis and presence of severe disease in order to provide a subsequent life-saving aortic valve intervention.