PROJECT SUMMARY
Aortic stenosis (AS)—narrowing of the aortic valve—is the most common heart valve disease. Degeneration of
leaflets in AS leads to valvular obstruction, thickening of the heart muscle, heart failure, and death. Due to an
aging population, the healthcare burden of AS continues to rise, yet it remains one of the major cardiovascular
diseases for which there are no preventive or disease-modifying treatments. Currently, the clinical strategy is
watchful waiting, until severe symptoms appear. At this late stage, the risk of death rises significantly, and valve
replacement is the only option. Currently, we lack tools to predict which patients will experience a rapid worsening
of their condition.
Computed tomography angiography (CTA) is increasingly used for noninvasive assessment of AS. Guided by
our preliminary studies, we propose to develop and validate artificial intelligence (AI)-enabled fully automated
quantification of fibrotic and calcific aortic valve tissue characteristics from CTA. We aim to derive radiomic
features of peri-aortic adipose tissue (which serve as indicators of vascular inflammation). We propose to
predict rapid aortic valve disease progression, in men and women. We will combine CTA-measured aortic valve
tissue characteristics and inflammation along with patient clinical data, into a new integrated risk score for the
prediction of mortality and stroke following valve replacement. For this project, we have assembled “real-world”
clinical, imaging, and follow-up data in 7442 patients from all stages of AS—from prospective clinical trials as
well as large multicenter, multiethnic patient registries.
We propose three specific aims: (1) To develop and validate a fully automated deep learning algorithm for the
characterization of fibrotic and calcific aortic valve tissue from standard CTA, in comparison with expert readers
and histopathological tissue analysis. (2) To assess the concordance between fibrocalcific aortic valve tissue
characteristics and functional valve hemodynamics by echocardiography and further, to predict rapid aortic valve
disease progression, in men and women, in prospective AS clinical trials. (3) To create and externally validate
an AI-powered risk score—integrating patient clinical data, fibrotic and calcific aortic valve tissue characteristics,
and peri-aortic vascular inflammation—to predict death and stroke within 1-year post-valve replacement.
The proposed work will provide, for the first time, automated, multi-parametric assessment of aortic valve tissue
and vascular inflammation characteristics from standard CTA. By predicting the likelihood of adverse
outcomes, the tool could enable the creation of personalized treatment and management plans, and guide post-
operative management, including the need for intensive monitoring, medication adjustments, and follow-up
appointments. This research will allow comprehensive phenotyping of AS, paving the way to image-informed
clinical trials of new therapies.
