Identifying treatment non-responders based on pre-treatment gait characteristics - A machine learning approach
Rosa M S Visscher1,2, Julia Murer1, Fatemeh Fahimi1,3
1Laboratory for Movement Biomechanics, Institute for Biomechanics, Department of Health Science & Technology, ETH Zürich, Zürich, Switzerland.
Insights
Machine learning accurately identifies children with movement disorders unlikely to respond to standard treatment. This helps personalize interventions for better outcomes in paediatric cerebral palsy care.
Area of Science:
- Biomedical Engineering
- Clinical Biomechanics
- Machine Learning in Medicine
Background:
- Paediatric movement disorders, like cerebral palsy, often impair walking.
- Current gait analysis guides therapy, but not all patients benefit.
- Predicting non-responders pre-treatment can personalize interventions.
Purpose of the Study:
- To assess if machine learning can identify patients at risk of negative treatment outcomes.
- To utilize pre-treatment gait and anthropometric data for prediction.
- To improve personalized intervention strategies for paediatric movement disorders.
Main Methods:
- Retrospective analysis of 119 patients with movement disorders.
- Extracted pre- and post-treatment gait (sagittal joint angles, spatiotemporal parameters) and anthropometric data.
- Trained a support vector machine classifier with 5-fold cross-validation and Bayesian optimization.
Main Results:
- Achieved 88.2% average accuracy in identifying non-responders.
- Demonstrated a 64% true negative rate.
- Obtained an area under the curve (AUC) of 88%.
Conclusions:
- A machine learning model effectively identified patients at risk of non-response to treatment.
- Pre-treatment gait and anthropometric data are valuable predictors.
- Model output can serve as a warning for personalized intervention needs.
Background:
Paediatric movement disorders such as cerebral palsy often negatively impact walking behaviour. Although clinical gait analysis is usually performed to guide therapy decisions, not all respond positively to their assigned treatment. Identifying these individuals based on their pre-treatment characteristics could guide clinicians towards more appropriate and personalized interventions. Using routinely collected pre-treatment gait and anthropometric features, we aimed to assess whether standard machine learning approaches can be effective in identifying patients at risk of negative treatment outcomes.
Methods:
Observational data of 119 patients with movement disorders were retrospectively extracted from a local clinical database, comprising sagittal joint angles and spatiotemporal parameters, derived from motion capture data pre- and post-treatment (physiotherapy, orthosis, botulin toxin injections, or surgery). Participants were labelled based on their change in gait profile score (GPS, non-responders with a decline in GPS of <1.6° vs. responders). Their pre-treatment features (sagittal joint angles, spatiotemporal parameters, anthropometrics) were used to train a support vector machine classifier with 5-fold cross-validation and Bayesian optimization within a MATLAB-based Classification Learner App.
Results:
An average accuracy of 88.2 ± 0.5 % was achieved for identifying participants whose gait will not respond to treatment, with 64 % true negative rate and an area under the curve of 88 %.
Conclusion:
Overall, a classical machine learning model was able to identify patients at risk of not responding to treatment, based on gait features and anthropometrics collected prior to treatment. The output of such a model could function as a warning signal, notifying clinicians that a certain individual might not respond well to the standard of care and that a more personalized intervention might be needed.
