Printhotics Predict: Machine Learning to Support Digital Ankle-Foot Orthoses Production

Abstract

Ankle-foot orthoses (AFOs) are thermoplastic braces encompassing the foot and ankle, often prescribed to improve walking ability of children and adults with neuromuscular diseases and other movement disorders. Traditionally, AFO fabrication begins with a plaster cast of the lower leg, followed by modification of the positive cast including additions for soft tissue expansion, realignment of foot and ankle deformities, smoothing for comfort and shaping for shoe fit. These manual tasks are dependent on clinician expertise and purpose-built workspaces, can be time consuming and generate significant waste material. 3D scanning technologies and machine learning techniques have the potential to replace plaster casting and contribute to a digital manufacturing workflow, involving 3D printing. The aim of this PhD thesis was to map and digitise the plaster cast modifications made by an orthotist and develop a machine learning algorithm to predict these modifications to support a fully digital AFO workflow. This thesis forms one component of the larger Printhotics program of research to evolve the fabrication of orthotic devices for children by advancing the knowledge and application of technology to digitalise and improve 3D printing service delivery. This PhD thesis has demonstrated the feasibility of machine learning to progress a digital AFO workflow, and reveals the potential for artificial intelligence to improve the efficiency of AFO fabrication.