Postoperative Gait Characteristics After Arthroscopic Ankle Arthrodesis Using a Deep Learning-Based Markerless Motion Capture System: A Pilot Observational Study

Background/Objectives: Ankle arthrodesis is an effective treatment for advanced ankle arthritis. Although postoperative gait following open ankle arthrodesis has been investigated, gait characteristics after arthroscopic ankle arthrodesis remain less well characterized. Therefore, we aimed to investigate postoperative gait characteristics after arthroscopic ankle arthrodesis using a deep learning-based markerless motion capture system in this pilot observational study. Methods: Seven patients with degenerative ankle arthritis underwent arthroscopic ankle arthrodesis. Real-time gait analysis using a deep learning-based markerless motion capture system assessed gait speed, cadence, step length, stance and swing phases, and ankle range of motion preoperatively and at 1 year postoperatively. Postoperative findings were also compared with those of a minimally symptomatic comparison group. Clinical outcomes were evaluated with the Visual Analog Scale, Foot and Ankle Ability Measure (FAAM), Lower Extremity Functional Scale, American Orthopaedic Foot and Ankle Society Ankle-Hindfoot Scale, and the Karlsson Scale. Results: At the 1-year postoperative follow-up, no significant differences in any of the spatiotemporal gait parameters were found between patients and the comparison group. Preoperative ankle range of motion during the gait cycle showed significant deviations from that of the comparison group, which were largely attenuated at the 1-year follow-up, except for persistent plantarflexion restriction during terminal stance. Numerical improvements were observed in most clinical scores, whereas the FAAM-Sports subscale remained essentially unchanged. Conclusions: At 1 year postoperatively, spatiotemporal gait parameters did not differ significantly from those of the comparison group, although terminal stance plantarflexion restriction persisted. Markerless motion capture provided objective gait data in a routine clinical setting, but its accuracy requires validation against standard marker-based systems.