Determination of input variables for the development of a gait asymmetry expert system in patients with idiopathic scoliosis

Abstract

The purpose of this study was to select the appropriate input variables for the development of an expert system to analyze the gait asymmetry of patients with idiopathic scoliosis. Gait experiments were performed with 12 healthy female adolescents and 16 female adolescents with untreated adolescent idiopathic scoliosis. The experimental equipment included six infrared cameras and two ground reaction force platforms. By using a 3D human model, gait elements, kinematic and kinetic data were extracted. Self-organizing map and genetic algorithm were used for proper selection of input variables, and these methods were validated by using auto regression models, which were described in previous studies. Sixty gait variables based on a literature review were selected, and Self-organizing map was used to maintain the independency between the input variables, and the 39 independent retaining variables were chosen. Also, in order to identify the inputs exhibiting a significant relationship with the output, a genetic algorithm-general regression neural network was applied; and the frequency of the solution set was measured by genetic algorithm iteration. A stepwise method was applied based on the variables with high frequency, and final 11 input variables were selected. Furthermore, a back propagation artificial neural network with high accuracy 96.3(3.2)%, which can discriminate patients from the normal subjects, was developed with selected 11 input variables. Therefore, the results of this study can be used as input variables for the development of a gait asymmetry expert system.