Clinical Spasticity Assessment Assisted by Machine Learning Methods and Rule-Based Decision.

Area of Science:

• Biomedical Engineering
• Rehabilitation Science
• Clinical Measurement

Background:

• The Modified Ashworth Scale (MAS) is a standard clinical tool for spasticity assessment.
• The qualitative nature of MAS leads to inconsistencies and ambiguity in spasticity evaluation.
• Objective, quantitative methods are needed to improve the reliability of spasticity assessment.

Purpose of the Study:

• To develop a data-driven approach for spasticity assessment using wearable sensor data.
• To enhance the accuracy and inter-rater reliability of spasticity classification.
• To create a machine learning model that integrates clinical expertise with quantitative measurements.

Main Methods:

• Collected clinical data from 50 subjects using wireless wearable sensors (goniometers, myometers, EMG).
• Extracted 18 features (8 kinematic, 6 kinetic, 4 physiological) from sensor data.
• Trained and evaluated machine learning classifiers (SVM, RF) and developed a hybrid Logical-SVM-RF model incorporating physician decision logic.

Main Results:

• The proposed Logical-SVM-RF classifier achieved an accuracy of 91% on an independent test set.
• Individual SVM and RF classifiers achieved accuracies ranging from 56% to 81%.
• The hybrid model significantly outperformed conventional machine learning classifiers in spasticity classification.

Conclusions:

• The developed quantitative, data-driven approach significantly improves spasticity assessment accuracy.
• Integrating machine learning with clinical decision-making enhances diagnostic reliability and inter-rater consistency.
• This method provides a foundation for more objective and reproducible spasticity evaluations.