From 4D Transesophageal Echocardiography to Patient Specific Mitral Valve Models

Mitral valve regurgitation is the most common valvular disease, affecting 10% of the population over 75 years old [1]. Current standard of care diagnostic imaging for mitral valve procedures primarily consists of trans- esophageal echocardiography (TEE) as it provides a clear view of the mitral valve leaflets and surrounding tissue. Heart simulator technology has been adopted widely by both industry for evaluation of technolo- gies for imaging heart valves [2], and academia for the assessment of modelled heart valves [3]. Recently, developments have been made on a workflow to cre- ate 3D, patient-specific valve models directly from trans-esophageal echocardiography (TEE) images. When viewed dynamically using TEE within a pulse duplicator simulator, it has been demonstrated that these models result in pathology-specific TEE images similar to those acquired from the patient’s valves in-vivo [4]. However, producing a mesh model of the valve geometry from TEE imaging remains a challenge. Previously, produc- ing a valve model included a labor intensive series of steps including manual leaflet segmentation, and computer-aided design (CAD) manipulation to derive a 3D printable mold from a raw segmentation. Our objective is to automate the workflow and reduce the labor requirements for producing these valve models. To address the leaflet segmentation problem, we developed DeepMitral, a fully automatic valve leaflet segmentation tool. Following leaflet segmentation, we have developed tools for automatically deriving mesh models that can easily be integrated into a mold base.