Gait-Based Machine Learning for Classifying Patients with Different Types of Mild Cognitive Impairment

Area of Science:

• Neurology
• Biomedical Engineering
• Machine Learning in Healthcare

Background:

• Mild cognitive impairment (MCI) has diverse etiologies including Alzheimer's disease and Parkinson's disease (PD).
• Early diagnosis and treatment of dementia are crucial due to potential remission.
• Cognitive decline in MCI can manifest as impaired walking performance.

Purpose of the Study:

• To predict different subtypes of MCI patients using gait information.
• To investigate the utility of machine learning models for MCI classification based on gait parameters.
• To differentiate between Parkinson's disease with MCI (PD-MCI) and non-PD-MCI.

Main Methods:

• Participants underwent gait analysis using a portable system, performing walk, Time Up and Go, and jump tests.
• Gait parameters were processed using machine learning classification models, specifically Support Vector Machine (SVM) with Principal Component Analysis (PCA).
• A polynomial kernel function was employed within the PCA-SVM model.

Main Results:

• The machine learning classification model successfully predicted different types of MCI patients.
• The PCA-SVM model achieved 91.67% accuracy in classifying PD-MCI versus non-PD-MCI.
• The PCA-SVM model demonstrated a ROC AUC of 0.9714 for PD-MCI classification.

Conclusions:

• Gait analysis combined with machine learning offers a promising approach for classifying MCI subtypes.
• The PCA-SVM model shows high efficacy in distinguishing PD-MCI from other forms of MCI.
• Objective gait metrics can serve as valuable biomarkers for early detection and subtyping of cognitive impairment.