Subject-specific identification of three dimensional foot shape deviations using statistical shape analysis

The high prevalence of foot pain, and its relation to foot shape, indicates the need for an expert system to identify foot shape abnormalities. Yet, to date, no such expert system exists that examines the full 3D foot shape and produces an intuitive explanation of why a foot is abnormal. In this work, we present the first such expert system that satisfies those goals. The system is based on the concept of model-based outlier detection: a statistical shape model (SSM) is generated from 186 3D optical foot scans of healthy feet. This model acts as a knowledge base which is then parameterized by one's demographic characteristics (e.g., age, weight, height, shoe size) through a multivariate regression. This regression introduces model flexibility as it allows the model to be fine tuned to a specific individual. This fine tuned model is then used as a baseline to which the individual's 3D foot scan can be compared using standard statistical tests (e.g. t-tests). These statistical tests are performed at each vertex along the foot surface to identify the degree and location of shape outliers. Our expert system was validated on foot scans from patients with hallux valgus and abnormal foot arches. As expected, our results varied per patient, confirming that feet with the same clinical classification still show high shape variability. Additionally, the foot shape abnormalities identified by our system not only agreed with the expected location and severity of the tested foot deformities, but our analysis of the full 3D foot shape was able to completely characterize the extent of those abnormalities for the first time. These results show that the combination of statistical shape modelling, multivariate regression, and statistical testing is powerful enough to perform outlier detection for 3D foot shapes. The resulting insights provided by this system could prove useful in both shoe design and clinical diagnosis. (C) 2020 Elsevier Ltd. All rights reserved.

Publication year:2020

Keywords:A1 Journal article

BOF-keylabel:yes

Accessibility:Closed