|
Affiliation
|
Epworth Healthcare and The University of Melbourne, Melbourne, Australia (Dr Williams); Victoria University, Melbourne, Australia (Dr Lai); School of Mechanical Engineering, The University of Melbourne, Australia (Dr Schache); and School of Allied Health, La Trobe University, Melbourne, Australia (Dr Morris). |
|
Abstract
|
OBJECTIVE:: To determine the extent to which gait disorders associated with traumatic brain injury (TBI) are able to be classified into clinically relevant and distinct subgroups.
DESIGN:: Cross-sectional cohort study comprising people with TBI receiving physiotherapy for mobility limitations. PARTICIPANTS:: One hundred two people with TBI. OUTCOME MEASURES:: The taxonomic framework for gait disorders following TBI was devised on the basis of a framework previously developed for people with cerebral palsy. Participants with TBI who were receiving therapy for mobility problems were assessed using 3-dimensional gait analysis. Pelvis and bilateral lower limb kinematic data were recorded using a VICON motion analysis system while each participant walked at a self-selected speed. Five trials of data were collected for each participant. Multiclass support vector machine models were developed to systematically and automatically ascertain the clinical classification.
RESULTS:: The statistical features derived from the major joint angles from unaffected limbs contributed to the best classification accuracy of 82.35% (84 of the 102 subjects). Features from the affected limb resulted in a classification accuracy of 76.47% (78 of 102 subjects).
CONCLUSIONS:: Despite considerable variability in gait disorders following TBI, we were able to generate a clinical classification system on the basis of 6 distinct subgroups of gait deviations. Statistical features related to the motion of the pelvis, hip, knee, and ankle on the less affected leg were able to accurately classify 82% of people with TBI-related gait disorders using a multiclass support vector machine framework.
Language: en |