Machine learning classification of multiple sclerosis patients based on raw data from an instrumented walkway

Area of Science:

• Biomedical Engineering
• Neurology
• Machine Learning

Background:

• Instrumented walkways with embedded sensors offer detailed gait analysis, but standard variables may overlook crucial data.
• Multiple sclerosis (MS) is a neurodegenerative disease causing gait impairments, making it suitable for machine learning studies.
• Raw sensor data from instrumented walkways can reveal subtle gait patterns.

Purpose of the Study:

• To apply machine learning techniques to raw instrumented walkway data for differentiating MS patients from healthy controls.
• To develop novel gait features that supplement standard parameters to improve classification accuracy.
• To investigate the efficacy of machine learning in identifying gait disturbances in MS.

Main Methods:

• Utilized machine learning algorithms, including Support Vector Machine (SVM), on raw data from instrumented walkways.
• Extracted standard gait variables and engineered novel features such as toe direction, hull area, and foot dimensions.
• Trained and evaluated models using a dataset comprising MS patients and healthy controls.

Main Results:

• Standard gait variables achieved 81% accuracy, 95% precision, 81% recall, and 87% F1-score using SVM.
• The addition of novel features (toe direction, hull area, base of support area, foot length, foot width, foot area) improved classification accuracy by 7%, recall by 9%, and F1-score by 6%.

Conclusions:

• Instrumented walkways generate extensive gait data, with machine learning capable of discerning MS patients from controls with high accuracy.
• Novel gait features significantly enhance the performance of machine learning models for MS gait analysis.
• Machine learning applied to raw walkway data offers a promising approach for objective gait assessment in neurological conditions.