Comparison of shoe insole materials by neural network analysis

The effects of two insole materials within the shoe are compared using neural network analysis. Seven male subjects without locomotor disorders walk on a treadmill at a controlled speed and cadence wearing a common shoe and no socks, under three conditions; these are two types of insole of the same thickness, and a no insole condition. Pressure-related data from under the foot, within the shoe, are obtained by the MICRO-EMED system during walking. A back-propagation neural network is trained to associate sets of pressure-related data with the insole conditions. Subsequently neural network analysis is performed to reveal the abstract rules that govern the decision-making processes within the neural network, based on the synergistic interactions between the measured variables. Data are also analysed using ANOVA. The neural network analysis finds trends in the way in which the trained neural network responds. The interpretation of those trends gives a delicate description of the dynamic behaviour of the insoles despite the fact that no significant differences are found using ANOVA. It is concluded that neural network analysis can distinguish between insole behaviour during use, even though these differences are not significantly different based on statistical tests.