Statistical Shape and Motion Model for Myocardial Deformation from Tagged Magnetic Resonance Imaging

Thalia Seale1, Vicente Grau2, Abhirup Banerjee1
1University of Oxford, 2Institute of Biomedical Engineering, Department of Engineering Science, University of Oxford


Abstract

Aims: This study aimed to improve the characterisation of regional myocardial dynamics in cardiac pathology, where conventional strain measures may fail to capture subtle or heterogeneous dysfunction. We leverage tagged MRI, which enables direct observation of myocardial deformation beyond surface-based metrics.

Methods: We constructed point clouds capturing myocardial material deformation using dense optical flow estimated over the cardiac cycle. A principal component analysis (PCA)-based statistical shape and motion model (SSMM) was then fitted on the point clouds to capture geometric and temporal variation.

Results: The model was evaluated on 759 UK Biobank subjects. SSMM-derived features achieved an area under the receiver operating characteristic curve (AUC) of 0.706 for incident myocardial infarction (MI) and 0.766 for prevalent MI. Latent space analysis (Fig. 1) shows that these features characterise myocardial deformation patterns that are not observable with conventional cine imaging and are significantly associated with MI.

Conclusion: Modelling dense spatiotemporal myocardial deformation provides information beyond conventional metrics, enabling interpretable phenotyping and capturing clinically relevant variation associated with MI.