A Validation Study of Freezing of Gait (FoG) Detection and Machine-Learning-Based FoG Prediction Using Estimated Gait...

Area of Science:

• Biomedical Engineering
• Neurology
• Wearable Technology

Background:

• Freezing of gait (FoG) is a common Parkinson's disease symptom impacting mobility and increasing fall risk.
• Objective quantification and automatic detection of FoG are crucial for personalized treatment strategies.

Purpose of the Study:

• To objectively quantify gait parameters using wearable accelerometer data.
• To compare accelerometer-derived gait parameters with 3D motion capture system data.
• To develop a machine learning model for automatic discrimination of FoG patients.

Main Methods:

• Gait parameters were quantified using data from wearable accelerometers.
• Estimated gait parameters were compared against a 3D motion capture system using mean error rate and Pearson's correlation coefficient (PCC).
• Machine learning classifiers, including Support Vector Machine (SVM), were employed for FoG detection.

Main Results:

• High agreement between accelerometer-derived and 3D motion capture system gait parameters was observed (PCC: 0.961–0.984).
• The mean error rate for estimated gait parameters was below 10%.
• The SVM classifier achieved approximately 88% accuracy in discriminating FoG patients.

Conclusions:

• The proposed accelerometer-based approach provides objective and accurate gait parameter quantification for freezing of gait.
• This method demonstrates applicability in real-world scenarios for assessing and monitoring FoG in Parkinson's disease.
• Wearable accelerometer systems are recommended for FoG assessment and management.