Biomedical Signal and Image Computing Laboratory

3D Biomechanical Patient-Specific Modeling of the Tongue for Simulating the Swallowing Process


Negar Mohaghegh Harandi (Grad Student)
BiSICL, Department of Electrical and Computer Engineering, UBC

Sidney Fels (Principal Investigator)
HCT, Department of Electrical and Computer Engineering, UBC

Rafeef Abugharbieh (Principal Investigator)
BiSICL, Department of Electrical and Computer Engineering, UBC


3D Modeling, medical image processing, segmentation, registration, swallowing, 3D Static MRI, Dynamic MRI, oropharyngeal structures, tongue, inverse modeling, muscle activation, OPAL project, Artisynth project, SOFA framework


In this project, we investigate a cost-effective framework for generating 3D biomechanical subject-speciffic model of the tongue and further simulate tongue's movements during the swallowing process. The proposed pipeline will take advantage of state-of-the-art imaging techniques to expedite segmentation, generation of the volumetric biomechanical model, and activation of the related biomechanics that simulate swallowing.

A schematic pipeline for subject-speciffic simulation: data acquisition, segmentation, biomechanical volume meshing and simulation modules.

A generic simulation platform is currently being developed in the biomechanical simulation framework Artisynth, to model the interaction of the oropharyngeal organs and the food bolus during swallowing. In this project, we aim to further expand this generic platform to encompass individualized information. This will allow investigation of the subjectspeciffc variability in the morphology and the physiology of swallowing. It will also facilitate future development of the patient-specffic platform, which will be aimed at assisting the diagnosis and treatment planning of dysphagia in medical settings.

Generic Dynamic Modeling of the Oral, Pharyngeal, and Laryngeal (OPAL) Complex for Biomedical Application in ArtiSynth.

This research will focus, in particular, on the subject-speciffic modelling and simulation of the tongue as the primary organ in oropharynx. We believe that the tongue accounts for the major challenges introduced in the subject-speciffic platform. Firstly, the tongue's morphology is hard to capture with medical image and functional data, especially the site where it infuses into the adjacent soft-tissues, such as muscles, connective tissues and glands. Motion artifacts also contribute to impairing the acquisitions. Secondly, the tongue has complex biomechanics. It is a muscle-activated soft tissue, controlled both by the activation of the external muscles in the head and its own interwoven muscle fibres. These activations result in a wide range of variation in tongue shape and movements. The project is part of the OPAL framework and is divided into the three following phases:

  1. Segmentation:
    Static MRI partially resolves soft tissue details of the oropharynx, which are crucial in swallowing and speech studies. However, delineation of tongue tissue remains a challenge due to the lack of definitive boundary features. We propose a minimally interactive inter-subject mesh-to-image registration scheme to tackle 3D segmentation of the human tongue from MR volumes. A tongue surface-mesh is first initialized using an exemplar expertdelineated template, which is then refined based on local intensity similarities between the source and target volumes. A shape matching technique (Gilles et al. 2008) is applied for regularizing the deformation. We enable effective minimal user interaction by incorporating additional boundary labels in areas where automatic segmentation is deemed inadequate. We validate our method on 18 normal-subjects using expert manual delineation as the ground truth. Results indicate an average dice segmentation accuracy of 0.904 ± 0.004, achieved within an expert interaction time of 2 ± 1 minutes per volume.

  2. Mesh evolution during the segmentation process in mid-coronal (left), mid-sagittal (middle) and mid-axial (right) slices. Position of the tongue in the initial configuration (a), after initialization phase (b), after mesh-to-volume registration (c) and after user interaction (d). The red circles in (c) denote the areas in which user input is deemed necessary.

  3. FE Mesh Registration:
    The segmentation module will be followed by the generation of a finite-element (FE) volumetric mesh with respect to the generic biomechanical reference currently available in ArtiSynth. We plan to adapt the state-of-the-art FE mesh registration methods to fit the needs of the proposed pipeline. We will explore the optimum model regarding the computational limitations on one hand, and the required level of details for accurate subject-specfic registration and simulation, on the other hand.

  4. FE model of the generic tongue (right) compared to supject-specific FE model (left) built based on first time frame of Cine-MRI of speech data.

  5. Simulation:
    We plan to activate the biomechanics of the subject-specific model to simulate swallowing in the ArtiSynth. The proposed pipeline will apply the inverse modelling toolkit to estimate the virtual muscle activations from the subject-specific kinematic information of swallowing. Effective control points, computed based on anatomical features of the tongue, will be tracked in dynamic MRI. The activated model will be further validated using the physiological and functional acquisitions of swallowing.

Selected Publications

  • N. M. Harandi, R. Abugharbieh, S. Fels, “Minimally Interactive MRI Segmentation for Subject-Specific Modelling of the Human Tongue”, MICCAI workshop on Bio-Imaging and Visualization for Patient-Customized Simulations (BIVPCS), Nagoya-Japan, Sep 2013. [PDF]

a place of mind, The University of British Columbia

The University of British Columbia
Vancouver Campus
2329 West Mall
Vancouver, B.C., Canada V6T 1Z4
Tel (Directory Assistance): 604.822.2211
The University of British Columbia
Okanagan Campus
3333 University Way
Kelowna, B.C., Canada V1V 1V7
Tel: 250.807.8000