Freezing of gait in Parkinson's disease: Classification using computational intelligence.

Area of Science:

• Neurology
• Biomedical Engineering
• Machine Learning

Background:

• Parkinson's disease (PD) is a neurodegenerative disorder characterized by dopamine neuron loss, leading to motor and non-motor symptoms.
• Distinguishing PD from other conditions can be challenging due to symptom overlap.
• Gait abnormalities are a key diagnostic indicator for Parkinson's disease.

Purpose of the Study:

• To develop and evaluate machine learning models for PD diagnosis using gait data.
• To assess the effectiveness of medication in PD patients based on gait parameters.
• To analyze classifier performance across different tasks and patient states (ON/OFF medication).

Main Methods:

• A dataset was compiled including healthy individuals and PD patients (with and without freezing of gait), recorded during ON and OFF medication states.
• Data was collected from four tasks: voluntary stop, timed up and go, simple motor task, and dual motor/cognitive task.
• Seven distinct classifiers were applied to differentiate PD patients from controls and to evaluate medication effects.

Main Results:

• Machine learning models were applied to distinguish PD patients from healthy individuals, both overall and task-specifically.
• The effectiveness of medication was assessed using the collected gait data.
• Multilayer perceptron and decision tree classifiers demonstrated the most consistent performance across the evaluated problems.

Conclusions:

• Gait analysis data, when processed with machine learning, shows promise for PD diagnosis.
• Machine learning models can effectively differentiate PD patients and assess treatment response.
• Specific classifiers like multilayer perceptron and decision tree offer reliable results for PD-related gait analysis.