From real-time acquisition to mesh morphing of foot at different positions

This work is aimed to set-up a methodology for foot shape prediction at different flexion angles, overcoming limitations encountered when different poses are required but a limited set of acquisitions can be performed. The basic idea was to identify a fitting law able to interpolate positions of foot anatomical landmarks, and then use this information to guide the deformation of an average foot shape. First of all, mesh correspondence between foot geometries was accomplished by an established procedure based on mesh morphing. Then Procrustes analysis was applied to the dataset to remove rigid motions and estimate the average shape. Two interpolation laws (linear and quadratic) were investigated and the best one in terms of prediction of 3D landmarks’ coordinates was identified. Finally, shape geometries at any flexion angle were predicted performing a second mesh morphing guided by interpolated landmarks’ displacements from the average shape. These analyses proved that a limited number of interpolation angles provides a prediction accuracy comparable to that obtained using all the angles available in the dataset. Moreover, predicted shapes have been compared to the actual scans in terms root mean square error between corresponding nodes, obtaining a mean value of 4.03 ± 1.39 mm, in accordance with data reported in literature.