Automatic Personalization of the Mitral Valve Biomechanical Model Based on 4D Transesophageal Echocardiography

Abstract

Patient-specific computational models including morphological and biomechanical models based on medical images have been proposed to provide quantitative information to aid clinicians for Mitral Valve (MV) disease management. Morphological models focus on extracting geometric information by automatically detecting the mitral valve structure and tracking its motion from medical images. Biomechanical models are primarily used for analyzing the underlying mechanisms of the observed motion pattern. The recently developed patient-specific biomechanical models have integrated the personalized mitral apparatus and boundary conditions estimated from medical images to predicatively study the pathological changes and conduct surgical simulations. As a next step towards transitioning patient-specific models into clinical settings, an automatic personalization algorithm is proposed here for biomechanical models extracted from Transesophageal Echocardiography (TEE). The algorithm achieves the customization by adjusting both the chordae rest length and material parameters such as Young’s modulus which are challenging to estimate or measure directly from the medical images. The algorithm first estimates the mitral valve motion from TEE using a machine learning method and then fits the biomechanical model generated motion into the image-based estimation by minimizing the Euclidean distances between the two. The algorithm is evaluated on 4D TEE images of five patients and yields promising results, with an average fitting error of 1.84±1.17mm.