The gait patterns of patients with multiple sclerosis (MS) were found to be discriminable from those of healthy controls using machine learning methods, according to an article published in BioMedical Engineering OnLine.

The study utilized a standard set of gait features with a support vector machine learning model to differentiate the gaits of patients with MS from those of healthy controls with an accuracy of 81%, precision of 95%, recall of 81%, and F1-score of 87%.

When an additional set of novel gait features (toe direction, hull area, base of support area, foot length, foot width, and foot area) were added to the support vector machine, the accuracy increased to 88%, recall to 90%, and F1-score to 93%. The precision level was unchanged.

Continue Reading

“These results demonstrate that machine learning models trained on new features from raw walkway data can more effectively separate patient and control targets and could potentially be served as an alternative method for identifying gait abnormalities in MS,” the authors said.

Read more about MS symptoms

During the study, 3 different machine learning algorithms were tested including logistic regression, XGBoost, and support vector machine. Support vector machine performed the best of the 3 algorithms on the metrics of accuracy, precision, recall, and F1-score.

The area underneath the precision-recall (AUPRC) and the area underneath the receiver operating characteristic (AUROC) curves were also compared between the 3 algorithms. When using the standard feature set, all 3 curves performed similarly in AUPRC and AUROC. Utilization of the augmented feature set provided the largest improvement and ultimately highest scores for the support vector machine model which increased its mean AUPRC from 0.89 to 0.96 and its AUROC from 0.88 to 0.96.

Gait data from a total of 72 patients with MS and 16 healthy controls were utilized during the study. The average Expanded Disease Severity Scale for the patients was 2.74±2.06. The patients with MS selected for the study scored a 3 or higher on gait-related questions on the Multiple Sclerosis Impact Scale survey.


Hu W, Combden O, Jiang X, et al. Machine learning classification of multiple sclerosis patients based on raw data from an instrumented walkway. Biomed Eng Online. 2022;21(1):21. doi:10.1186/s12938-022-00992-x