Multi-modal pipeline for comprehensive validation of Mitral Valve geometry and functional computational models

Valvular heart disease affects a high number of patients, exhibiting significant mortality and morbidity rates. Mitral Valve (MV) Regurgitation, a disorder in which the MV does not close properly during systole, is among its most common forms. Traditionally, it has been treated with MV replacement. However, recently there is an increased interest in MV repair procedures, providing better long-term survival, better preservation of heart function, lower risk of complications, and usually eliminating the need for long-term use of blood thinners (anticoagulants). These procedures are complex and require an experienced surgeon and elaborate pre-operative planning. Hence, there is a need for efficient tools for training and planning of MV repair interventions. Computational models of valve function have been developed for these purposes. Nevertheless, state-of-the-art models remain approximations of real anatomy with considerable simplifications, since current modalities are limited by image quality. Hence, there is an important need to validate such low-fidelity models against comprehensive ex-vivo data to assess their clinical applicability. As a first step towards this aim, we propose an integrated pipeline for the validation of MV geometry and function models estimated in ex-vivo TEE data with respect to ex-vivo microCT data. We utilize a controlled experimental setup for ex-vivo imaging and employ robust machine learning and optimization techniques to extract reproducible geometrical models from both modalities. Using one exemplary case, we demonstrate the validity of our framework.