Minimally Interactive MRI Segmentation for Subject-Specific Modelling of the Tongue

Abstract

AbstractSubject-Specific biomechanical modelling of the human tongue is beneficial for investigating the inter-subject variability in the physiology of the speech, chewing and swallowing. Delineation of the tongue tissue from MRI is essential for modelling, but still remains a challenge due to the lack of definitive boundary features. In this paper, we propose a minimally interactive inter-subject mesh-to-image registration scheme to tackle 3D segmentation of the tongue from MR volumes. An exemplar expert-delineated template is deformed to match the target volume, constrained based on a shape matching regularization technique. We enable effective minimal user interaction by incorporating additional boundary labels in areas where automatic segmentation is deemed inadequate. We validate our method on 12 normal-subjects. Results indicate an average dice overlap of 0.904 with the ground truth, achieved within 3 min of the expert interaction.KeywordsHyoid BoneManual SegmentationShape MatchGoal PositionMesh DeformationThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.