Background: A functional group is a collection of individuals who react in a similar way to a specific intervention, such as sport shoes or foot orthosis. Matching footwear features to a functional group can potentially improve running performance and reduce the risk of movement-related injuries. To match footwear features to a functional group, it is necessary to first define the different groups based on their distinctive movement patterns. Thus, the primary purpose of this study was to develop and implement a novel deep learning method for the dynamic foot categorization of runners. Methods: A deep temporal clustering algorithm was used to identify groups based on the three-dimensional angles of the ankle joint data from 108 healthy adults running barefoot. Once the groups were identified, statistical parametric mapping was employed to determine the differences between the kinematic patterns across stance. Results: Three distinct clusters were identified. Waveform analysis showed that cluster 1 had a larger ankle joint dorsiflexion between 40% and 80% of the stance phase compared to clusters 2 and 3. However, all three groups had similar ankle joint angles in the horizontal plane during running. Both cluster 1 and cluster 2 exhibited greater ankle joint inversion between 60% and 100% of the stance phase when compared to cluster 3. Significance: Demonstrated model could classify runners based on their ankle joint motions during running. This highlights the effectiveness of the proposed pattern recognition approach in automatically categorizing individuals into groups based on their unique movement patterns. This classification allows us to address specific interventions, such as footwear requirements, for each group.
