Biomechanical modeling and simulation of human eye movement

Studying human eye movement has significant implications for understanding the oculomotor system and treating vision disorders. Existing models of the oculomotor system either simplify the geometry and mechanics of the orbit, or are restricted to static simulation. In this dissertation, we present a novel three-dimensional (3D) biomechanical modeling framework for simulation of the oculomotor plant that addresses the above limitations. We aim to lay the foundation of a biomechanical simulator that will potentially be used for scientific research on ocular motility and clinical applications.;We first propose an efficient method for building subject-specific orbit models from magnetic resonance imaging (MRI). We reconstruct 3D geometric models of the orbit by fitting a generic template model to the MRI data of individual subjects. An automatic fitting process is developed, which combines parametric surface deformation with image feature selection. The accuracy of our method is validated by comparison to manual segmentation. We also present 3D reconstruction of eyeball models from MRI using the template approach with subdivision surface fitting.;We then describe a new approach for determining the averaged longitudinal strains of cylindrical soft tissues. Our method does not rely on image features to establish tissue correspondences and uses the incompressibility property of soft tissues. We demonstrate its usefulness by estimating extraocular muscle (EOM) strains from reconstructed models. Simulated sensitivity analysis and validation on MRI of a rubber phantom show its accuracy. Integrating estimated EOM strains as deformation constraints, we register EOM models across eye positions in a physically consistent way.;Finally, we develop a 3D dynamic biomechanical model for simulating ocular motility. We model EOMs as "strands," which are modeling elements for musculotendon mechanics. Realistic muscle paths and cross sectional areas of the EOM strands are based on 3D geometric models reconstructed from human subject MRI. Nonlinear EOM mechanics are incorporated and pulley hypotheses are implemented. Simulation of fixations, smooth pursuits, and saccades are demonstrated. The model generates realistic gaze trajectories from neural control signals. We validate our simulator by comparing simulations to experimental data. Our model is the first one that simulates dynamics and includes anatomical and physiological properties.