Vertical ground reaction force marker for Parkinson's disease

Area of Science:

• Biomedical Engineering
• Neurology
• Machine Learning

Background:

• Parkinson's disease (PD) is characterized by abnormal gait patterns.
• Automated gait analysis offers potential for early PD diagnosis and treatment monitoring.

Purpose of the Study:

• To differentiate PD patients from healthy controls using plantar vertical ground reaction force (VGRF) data.
• To evaluate the efficacy of machine learning classifiers for PD gait prediction.

Main Methods:

• Features extracted from VGRF data, including swing time and stride time variability, were selected using sequential forward feature selection.
• Support Vector Machine (SVM), K-nearest neighbor (KNN), random forest, and decision tree classifiers were employed.
• A prediction model was developed and evaluated for classification accuracy.

Main Results:

• The Support Vector Machine (SVM) classifier with a cubic kernel achieved the highest performance.
• Achieved accuracy of 93.6%, sensitivity of 93.1%, and specificity of 94.1%.
• Improved classification performance by approximately 10% compared to previous studies using the same dataset.

Conclusions:

• VGRF data from wearable devices, combined with SVM and selected features, can effectively diagnose PD.
• This approach shows potential for monitoring the effectiveness of PD therapies non-invasively.
• The study highlights the utility of machine learning in analyzing gait for neurological disorder assessment.