A Decision Support System to Facilitate Identification of Musculoskeletal Impairments and Propose Recommendations Using Gait Analysis in Children With Cerebral Palsy

The identification of musculoskeletal impairments from gait analysis in children with cerebral palsy is a complex task, as is formulating (surgical) recommendations. In this paper, we present how we built a decision support system based on gait kinematics, anthropometrics, and physical examination data. The decision support system was trained to learn the association between these data and the list of impairments and recommendations formulated historically by experienced clinicians. Our aim was 2-fold, train a computational model that would be representative of data-based clinical reasoning in our center, and support new or junior clinicians by providing pre-processed impairments and recommendations with the associated supportive evidence. We present some of the challenges we faced, such as the issues of dimensionality reduction for kinematic data, missing data imputations, class imbalance and choosing an appropriate model evaluation metric. Most models, i.e., one model for each impairments and recommendations, achieved a weighted Brier score lower than 0.20, and sensitivity and specificity greater than 0.70 and 0.80, respectively. The results of the models are accessible through a web-based application which displays the probability predictions as well as the (up to) 5 best predictors.