TEMPLATE-BASED RECONSTRUCTION OF HUMAN EXTRAOCULAR MUSCLES FROM MAGNETIC RESONANCE IMAGES

Understanding the mechanisms of eye movement is difficult without a realistic biomechanical model. We present an efficient and robust computational framework for building subject-specific models of the orbit from magnetic resonance images (MRIs). We reconstruct three-dimensional geometric models of the major structures of the orbit (six extraocular muscles, orbital wall, optic nerve, and globe) by fitting a template to the MRIs of individual subjects. A generic template captures the anatomical properties of these orbital structures and serves as the prior knowledge to improve the completeness and robustness of the model reconstruction. We develop an automatic fitting process, which combines parametric surface fitting with successive image feature selections. Reconstructed orbit models from different subjects are demonstrated. The accuracy of the proposed method is validated through comparison of reconstructed extraocular muscle cross sections with manual segmentation. The Dice coefficient is used as the metric and good agreement is observed.