Functional Foot Segmentation Based on Plantar Pressure Measurements for Profiling Subjects Performing a Running Exercise

Foot plantar pressure measurements allow quantitative analysis and classification of gait patterns and can provide valuable insight into the execution of different movements, with applications in health care and sports. In this paper, we introduce a novel data-analysis methodology that allows to decompose running trials from different subjects into its constituent spatial and temporal building blocks. The method is built around non-negative matrix factorization that allows clustering the temporal activation patterns observed across all trials (inter- and intra-subject), and subsequently, the spatial components corresponding to these temporal signals. This results in a spatio-temporal decomposition of each trial based on the prototypical patterns observed across all trials. The per trial weights can be used as spatio-temporal features for profiling and classification. The proposed methodology has been validated on a large database of running trials. The resulting decomposition revealed the characteristics of the running exercise. The derived spatio-temporal features were able to capture differences in running style from heel and non-heel strikers, but did not reveal differences in running style for the physical pathologies flat/high arch and hallux valgus.