Leveraging Sexual Dimorphism to Predict Cardiac Remodeling and Enhance Treatment in Women with Severe Aortic Stenosis

ABSTRACT Women with severe aortic stenosis (AS), a condition characterized by the narrowing of the aortic valve opening, often experience delayed diagnosis, undertreatment, and higher mortality rates compared to men, indicating both delayed care-seeking and a lack of appropriate diagnostics and monitoring for female patients. Yet, the

influence of anatomical and functional differences in the female population on AS presentation, management, and outcomes remains poorly understood. Furthermore, despite the prevalence of symptoms, women with severe AS receive less aortic valve replacement (AVR) treatment and have higher excess mortality rates over a

five-year period compared to men. Our proposed project integrates innovative medical image processing and computational modeling methods, such as statistical shape analysis (SSA), convolutional neural networks (CNN), and inverse finite element analysis (FEA), to gain sex-specific insights into cardiac remodeling and dysfunction, with a

specific focus on severe AS in women. By focusing on cardiac remodeling, a consequence of prolonged aortic valve disease, our goal is to enhance AS treatment for women by considering sex-specific differences in ventricular responses to AVR-induced afterload. To achieve this, we will develop a personalized, mathematical

approach that leverages sex-differentiating anatomical and functional characteristics of the left ventricle (LV), ultimately aiming to improve survival outcomes. Additionally, we will compare the predictive value of these sex- differentiating measures to traditional indices, enhancing our understanding of their effectiveness in guiding

clinical management. We hypothesize that advanced anatomical metrics (e.g., shape scores) and material characteristics (e.g., cardiac stiffness) are superior predictors of post-intervention cardiac events and dysfunction compared to traditionally collected clinical measures. Our research consists of two main aims. Aim 1 involves

developing a fully automated, neural network pipeline to segment clinical images, creating an advanced SSA model to extract hidden geometrical features, and establishing a correlation between shape scores and post- intervention clinical events. This analysis will assess the predictive power of sex-specific measures compared to

the male population, cases where sex is not considered in the model training, and universal clinical indices. Aim 2 focuses on developing a computational tool to estimate patient-specific stiffness of the inhomogeneous LV tissue, with an examination of its potential value in predicting diastolic dysfunction and AVR outcomes.

Our research serves as a steppingstone to guide clinicians in preprocedural patient selection, optimize surgical timing, and improve survival outcomes. By developing a sex-specific risk stratification tool and a mechanistic framework for effective prognosis, we aim to provide valuable means to enhance treatment and mitigate

devastating events associated with severe AS, particularly for female patients.