Using Position-Based Dynamics for Simulating the Mitral Valve in a Decision Support System

Abstract

In mitral valve interventions, surgeons have to select an optimal combination of techniques for every patient. Especially less experienced physicians would benefit from decision support for this process. To support the visual analysis of the patientspecific valvular dynamics and an in-silico pre-intervention simulation of different therapy options, a real-time simulation of the mitral valve is needed, especially for the use in a time-constrained clinical environment. We develop a simplified model of the mitral valve and propose a novel approach to simulate the mitral valve with position-based dynamics. As input, a mesh representation of the open-state mitral valve, two polygons representing the open and closed annulus states, simplified chordae tendineae, and a set of forces for approximating the surrounding blood are required. The mitral valve model can be deformed to simulate the closing and opening as well as incorporate changes caused by virtual interventions in the simulation. For evaluation, ten mitral valves were reconstructed from transesophageal echocardiogram sequences of patients with normal and abnormal physiology. Experts in cardiac surgery annotated anatomical landmarks for valve reconstruction. The simulation results for closing the valve were qualitatively compared to the anatomy depicted in the image sequences and, if present, the reproduction of a prolapse was verified. In addition, two virtual interventions (annuloplasty and clipping) were performed for one case and provided new insights about changes in valve closure and orifice area after modification. Each simulation ran at interactive frame rates. Our approach enables an efficient simulation of the mitral valve with normal and abnormal valve closing behavior as well as virtual interventions. The simulation results showed good agreements with the image data in general and reproduced valve closure in all cases. In three cases, prolapse was not or not correctly reproduced. Further research is needed to parameterize the model in pathologic cases.