Autonomous Estimation of Foot Bone Structure Based on Scanned Foot Surface Topography

The main goal of this master’s thesis was to develop an autonomous algorithm for the construction of an approximate surface model for a foot that would also generate the corresponding underlying bone structure model through non-invasive means for the purpose of enhancing finite element analysis (FEA) simulations for functionally personalized footwear. The algorithm consists of the following components: (1) measurement of the foot surface topography using LED light stereo imaging to obtain a scanned foot model; (2) machine learning and post-processing sub-algorithms for the localization of a set of anatomical landmarks on the given scanned foot model; (3) an optimization procedure to determine the most optimal transformation to approximate a given surface scan utilizing these anatomical landmarks; and (4) the application of the optimal transformation on a reference bone model to generate a patient-specific bone structure model. This algorithm was tested on synthetic data that was generated from scanning a physical foot model. The results demonstrated the robustness of the algorithm to construct approximate bone structures from a variety of white light scanned foot surface geometries. This work provides a protocol to synthesize various techniques into one comprehensive autonomous algorithm which is further enhanced by the introduction of machine learning to determine the anatomical landmarks of any given foot scan without the need for human intervention.