Implementation of Multiple Regression Technique for Detection of Gait Asymmetry Using Experimental Gait Data

Abstract

This study reports the implementation of different regression (multivariate and step-wise) techniques towards determining the gait asymmetry and comparison with the symmetry indices (SI) values. To predict the gait asymmetry between two different legs, a set of gait trials (thirty-five participants) via a three-dimensional motion capture setup and force platform available at the parent institute, has been acquired. Two separate regression fit models are prepared to indicate the significant gait parameters for the right and left legs utilizing the recorded foot data. The significant sets of coefficients for the right and left leg parameters are compared with the SI values to validate the gait asymmetry. The calculated mean SI from the experimental results correspond to the predicted regression model responses, and 18 of the 27 regression fits present different sets of significant coefficients for the right and left leg parameters. The regression fits show gait parameter dependency on the different sets of predictor variables. The method can be adopted for different patient data sets to detect the influence of the causative factors on pathologic gait.