Shape-Based Methods for Motor Function Analysis

Problems in biomechanics often involve nuisance variables such as varying motion speeds, orientations, and individual differences in limb sizes. In order to effectively learn from smaller, noisy datasets, these factors must be quotiented out. This dissertation introduces new tools from statistical shape analysis to address such issues. In the first study, motion trajectories in children with Duchenne Muscular Dystrophy (DMD) and Spinal Muscular Atrophy (SMA) are temporally aligned using the square root of their derivative (SRVF) and analyzed with Functional Principal Component Analysis (FPCA). The results reveals key variations in curl speed and asymmetry, with SMA patients showing greater activation of the asymmetry pattern. In the second study, Kinematic and EMG data in stroke patients are analyzed using the Transported SRVF framework in Kendall shape space, improving registration and classification of hemiplegic gaits. This approach identifies mean temporal shapes and modes of variation, enhancing the understanding of gait abnormalities and potentially informing better clinical assessments. These methods improve motion analysis in children with neuromuscular disorders and stroke patients, and holds promise for developing objective motor function assessments in clinical and remote settings.