Artificial neural networks in the selection of shoe lasts for people with mild diabetes.

This research addressed the selection of shoe lasts for footwear design to help relieve the pain associated with diabetic neuropathy and foot ulcers. A reverse engineering (RE) technique was used to convert point clouds corresponding to scanned shoe lasts and diabetic foot data into stereo lithograph (STL) meshes. A slicing algorithm was developed and was used to find relevant girth features of diabetic foot and the shoe lasts. An artificial neural network, termed self-organizing map (SOM), classified 60 sets of shoe lasts into similar groups. Foot shapes of three mild diabetic patients were entered into the SOM feature categories to match with suitable shoe lasts. By conducting expert questionnaire analysis of the characteristic girths featured data with analytic hierarchy process (AHP), the weights of the girths were obtained. Grey relational analysis (GRA) was then used to calculate the correlation between foot girth and the corresponding range of shoe lasts. The most suitable shoe last for each patient with a mild diabetic foot can be determined by calculating the relative fitness function for each patient. By correlating diabetic foot with suitable shoe lasts, this study demonstrated an effective strategy for designing shoes for patients with mild diabetes, which can then be manufactured to meet customized requirements.