An Automated Pipeline to Create Patient Specific 3D LV Geometry Models of Patients with Mitral Annular Disjunction

Gabriel Balaban1, Eivind Aabel2, Margareth Ribe2, Anna Castrini2, Kristina Haugaa2, Mary Maleckar1
1Simula Research Laboratory, 2Oslo University Hospital


Background: Mitral annular disjunction (MAD) is characterized by an abnormal insertion of the posterior mitral leaflet on the atrial wall. Despite its often subtle presentation, the presence of MAD is a potential warning sign of future ventricular arrhythmia (VA) and sudden cardiac arrest (SCA): recent studies have shown association among MAD-related imaging metrics and VA/SCA. Nevertheless, the precise mechanisms leading to VA/SCA in patients with MAD are poorly understood. A comprehensive 3D shape analysis of the left ventricles (LV) of patients with MAD may provide further insight and help to elucidate mechanisms.

Aims: To create a patient-specific 3D LV geometry modelling pipeline for patients with MAD, to enable future morphological studies based on cardiac short axis magnetic resonance imaging (MRI).

Methods: Short axis cardiac MRI scans were obtained for 25 patients presenting with MAD to Oslo University Hospital. The left ventricular endocardial and epicardial contours were traced with Segment software, and a simple polynomial fitting algorithm was applied to correct for possible slice shifts during image acquisition. An initial 3-D ellipsoid model was fitted via least squares to each patient’s contours, and then warped to fit the patient contours using free form B-spline image registration.

Results: Patient-specific geometries were created for all patients, indicating a robust pipeline. An example 3D model and MRI contours is shown in Figure 1.

Conclusions: Personalized 3D left ventricular shape models can be constructed for patients with MAD, to enable morphological studies examining the relationship between LV shape in MAD, and VA/SCD.