An Explainable Machine Learning Framework for Intelligent Physiotherapy Decision Making, Movement Assessment, Rehabilitation Outcome Prediction, and Personalized Treatment Planning
Computer and Decision Making: An International Journal · 01.11.2026 · 3 (6S) · 972-982
DOI: 10.1922/comdem.v3i6s.18861 ↗Kurz gefasst
Fachartikel in der Zeitschrift Computer and Decision Making: An International Journal, online erschienen am 01.11.2026. Verfasst von Dr. Shivangi Joshi und 4 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Quality of rehabilitation exercises is of importance for physiotherapy decision making and is important when the exercises are carried out without constant supervision of the therapist. This study was aimed at creating an explainable machine learning framework for sensor-based movement assessment, immediate rehabilitation movement-outcome prediction and personalized corrective decision support. Physical Therapy Exercises data was analysed with help of five synchronized wearable units including tri-axial accelerometers, gyroscopes and magnetometers. Correct, fast and low amplitude executions explicitly marked were segmented in 1-second overlapping windows and 480 statistical and magnitude-based features were extracted. Subject independence was maintained with nested leave-one-subject-out validation and all preprocessing and hyperparameter optimization were only performed on training subjects. Logistic regression, random forest, extra trees and decision tree models were compared. Random Forest obtained the best results in terms of execution, with accuracy of 0.6417, balanced accuracy of 0.6417 and macro-F1 of 0.6426. One-vs-rest ROC-AUC scores were between 0.780 and 0.828. Explainability determined that sensor units 2 and 3 were the most significant, containing features from the accelerometer, gyroscope and magnetometer. As seen throughout the framework, noticeable inter-subject variability exists and exercise-quality feedback is still required at the clinical level, which the framework shows as a clear path from wearable movement signals to clinical exercise-quality feedback.
Zitieren
APA
Joshi, D. S., Tahoor, D. A., Sharma, D. D., Gupta, P. D. N., & Kalra, D. S. (2026). An Explainable Machine Learning Framework for Intelligent Physiotherapy Decision Making, Movement Assessment, Rehabilitation Outcome Prediction, and Personalized Treatment Planning. Computer and Decision Making: An International Journal, 3(6S), 972-982. https://doi.org/10.1922/comdem.v3i6s.18861
Vancouver
Joshi DS, Tahoor DA, Sharma DD, Gupta PDN, Kalra DS. An Explainable Machine Learning Framework for Intelligent Physiotherapy Decision Making, Movement Assessment, Rehabilitation Outcome Prediction, and Personalized Treatment Planning. Computer and Decision Making: An International Journal. 2026;3(6S):972-982. doi:10.1922/comdem.v3i6s.18861
BibTeX
@article{Joshi2026,
author = {Joshi, Dr. Shivangi and Tahoor, Dr. Aterah and Sharma, Dr. Deepti and Gupta, Prof. (Dr) Neha and Kalra, Dr. Sumit},
title = {An Explainable Machine Learning Framework for Intelligent Physiotherapy Decision Making, Movement Assessment, Rehabilitation Outcome Prediction, and Personalized Treatment Planning},
journal = {Computer and Decision Making: An International Journal},
year = {2026},
volume = {3},
number = {6S},
pages = {972--982},
doi = {10.1922/comdem.v3i6s.18861}
}
EUVIMED