Cognitive driven gait freezing phase detection and classification for neuro-rehabilitated patients using machine...

Area of Science:

• Biomedical Engineering
• Neurology
• Machine Learning

Background:

• Diagnosing brain cognitive and gait freezing patterns is crucial for mental health disorders.
• Current visual inspection methods for gait abnormalities are subjective and inaccurate.
• Objective gait analysis can improve prognosis and treatment for Parkinson's patients.

Purpose of the Study:

• To differentiate between gait brain disorder and typical walking patterns.
• To utilize machine learning (ML) and inertial measurement unit (IMU) sensor data.
• To enhance gait recognition for brain, hip, and leg rehabilitation.

Main Methods:

• Utilized the Daphnet freezing of Gait Data Set (237 instances, 9 attributes).
• Applied ML algorithms and feature reduction techniques for gait recognition.
• Employed supervised learning with IMU sensor data from brain, hip, and leg.

Main Results:

• Random Forest (RF) achieved the highest accuracy at 98.9%.
• Perceptron showed the lowest accuracy at 70.4%.
• LDA feature reduction combined with KNN, RF, and NB yielded promising accuracy and F1-scores.

Conclusions:

• Integrating ML algorithms provides a viable solution for distinguishing gait disorders from normal walking.
• Machine learning offers an objective approach to support clinical judgment in gait analysis.
• This research highlights the potential of ML in neurological disorder diagnosis.