Subject-Specific Prediction of Muscle Activations

Computer-aided modeling and simulation of the oropharyngeal structures are beneficial for 3D visualization, and for the understanding of the associated physiology. Generic biomechanical models of the Oral, Pharyngeal, and Laryngeal (OPAL) structures are adopted into the ArtiSynth framework. Forward-dynamics tracking of FE model of the tongue was previously addressed through solving the inverse problem. The estimated biomechanics were evaluated using either the average motion reported in the literature or those of a different subject. We expand the existing generic platform to allow for subject-specific simulations, in order to (1) better evaluate the simulated biomechanics, (2) investigate the inter-subject variability and (3) provide additional insight into the speech production.

This project involves creation and simulation of subject-specific coupled biomechanical models of the tongue, jaw, and hyoid. Models are created based on the anatomy of the subject, captured in volumes of dynamic cine-MRI. We simulate our model based on the motion of tissue points of the tongue, extracted from the tagged-MRI of a speech sequence. The preliminary results show a plausible explanation for the known features of speech such as co-articulation.

In order to track the tongue's tissue points during the speech task, we combine the estimated motion from tagged- MRI, calculated by the harmonic phase algorithm (HARP), with surface information from cine-MRI using dieomorphic demons. A 3D dense and incompressible deformation field from tMRI is reconstructed based on divergence-free vector splines with incomplete data samples. It was observed that the tracking results are excellent at measuring internal tissue motion within the tongue, but are poorer near the boundary. Interestingly, the opposite situation occurs for tracking results using 3D deformation registration methods. Thus we enhance the tracking by adding boundary tissue tracking information derived from 3D deformable registration to the internal tissue tracking information derived previously.

We first evaluate the subject-specific model by activating the muscle groups separately. The observed motion was in the range of movements reported in the literature. The recorded tagged MRI experienced some noise in the left half of the subject’s tongue, so the target points (40 uniformly distributed FE nodes) were selected in the right half of the tongue, while forcing identical activations in the left. To test the accuracy of the inverse model, we designed a set of synthetic muscle activations. The displacements of the target points were recorded and fed back as input to the inverse model. The results show high accuracy. In the experiment with the tagged MRI, we calculated the average displacement in the neighbouring region of each target point, in order to reduce noise. The estimated activations are shown in Figure 3. The average of maximum tracking error was 2.88 mm. The work in progress includes running the simulation on the coupled tongue-jaw-hyoid model. We are further providing a more detailed analysis of motion of the articulatory model, including activation of the muscles, the hyoid and jaw. The experiments will be performed for more normal subjects, to measure the inter-subject variability, and investigate certain effects of gravity and head posture.

Relevant Publications

Negar M. Harandi, Jonghye Woo, Maureen Stone, Rafeef Abugharbieh and Sidney Fels (2014) Subject-Specific Biomechanical Modelling of the Tongue: Analysis of Muscle Activations During Speech. In Proceedings of the 10th International Seminar on Speech Production. "Cologne, April. (BibTeX)

Sanchez, C. Antonio, Lloyd, John E., Fels, Sidney and Abolmaesumi, Purang (2013) Embedding digitized fibre fields in finite element models of muscles.. . (URL) (BibTeX)

Ian Stavness, John E. Lloyd and Sidney Fels (2012) Automatic prediction of tongue muscle activations using a finite element model.. . (URL) (BibTeX)

Fangxu Xing, Jonghye Woo, Emi Z. Murano, Junghoon Lee, Maureen Stone and Jerry L. Prince (2013) 3D Tongue Motion from Tagged and Cine MR Images. In Proceedings of Medical Image Computing and Computer-Assisted Intervention(MICCAI)., pages 41-48, . Springer Berlin Heidelberg. (BibTeX)