Related Experiment Video
Updated: Aug 28, 2026
3D Kinematic Gait Analysis for Preclinical Studies in Rodents
Published on: August 3, 2019
LLM-Assisted Interpretation of Kinematic Gait Data in Children with Cerebral Palsy: A Pilot Study on Gait Deviation
Mehrdad Davoudi1,2, Jacqueline Romkes1, Michèle Widmer1
1Neuroorthopaedics and Centre for Clinical Motion Analysis, Department of Orthopaedics, University Children's Hospital Basel, 4056 Basel, Switzerland.
Abstract:
Three-dimensional instrumented gait analysis is widely used to guide surgical decision-making in children with cerebral palsy (CP), but its interpretation is time-consuming and prone to inter-rater variability. In this single-centre pilot study, we investigated whether a generative large language model (LLM) could consistently generate gait deviation findings and surgical procedure suggestions that align with expert judgement. Kinematic features for lower-limb joints across the gait cycle, stance, and swing were extracted from eight children with unilateral CP using the open-source GaitSharing Toolkit and a structured prompt, then submitted three times per patient to OpenAI's GPT-5.5 model. The model assessed 28 kinematic deviations and 12 surgical procedure groups using majority voting. One gait analyst and two paediatric orthopaedic surgeons independently rated outputs on a 0-2 ordinal scale, blinded to all clinical information beyond the kinematic curves and diagnosis. Agreement was summarised descriptively as the percentage of the maximum attainable score with 95% confidence intervals (CIs), and quadratic-weighted Cohen's kappa was used to quantify inter-surgeon agreement. Agreement with the gait expert was highest at the hip (90.6%) and lowest at the knee, particularly in the transverse plane (65.2%). For surgical procedures, agreement with the LLM reached 83.9% and 73.4% for the two surgeons, with the tibialis anterior procedure showing the lowest concordance. Inter-surgeon agreement was 79.2% (95% CI 71.9-85.4) with a kappa of 0.59 (0.47-0.70), indicating moderate agreement. The LLM showed high self-consistency (>90% across runs). These preliminary findings suggest that generative LLMs may be feasible as assistive tools in clinical gait analysis for deviation detection and future treatment planning and should be interpreted as hypothesis-generating, warranting confirmation in larger, more diverse cohorts.
