Convolutional Neural Network for Freezing of Gait Detection Leveraging the Continuous Wavelet Transform on Lower
Summary
This study introduces a deep learning model to detect Freezing of Gait episodes in Parkinson's disease patients using wearable sensors. The model achieved 89.2% accuracy, offering a promising tool for managing this disabling condition.
Area of Science:
- Neurology
- Biomedical Engineering
- Machine Learning
Background:
- Freezing of Gait (FoG) is a significant motor impairment in Parkinson's disease (PD).
- Accurate detection of FoG episodes is crucial for effective management and treatment.
- Wearable sensors and machine learning offer potential for objective FoG monitoring.
Purpose of the Study:
- To develop and evaluate a deep learning model for detecting Freezing of Gait episodes.
- To utilize data from wireless Inertial Measurement Units (IMUs) for FoG detection.
- To assess the model's generalizability and accuracy in PD patients with FoG.
Main Methods:
- Recruited 67 Parkinson's disease patients experiencing FoG.
- Collected time-series sensor data using two wireless IMUs placed on the ankles during clinical assessments.
- Converted sensor data to continuous wavelet transform scalograms.
- Trained a Convolutional Neural Network (CNN) model to identify FoG episodes.
Main Results:
- The CNN model achieved a generalization accuracy of 89.2%.
- The model demonstrated a geometric mean of 88.8% for FoG detection.
- The study successfully converted time-series sensor data into scalograms for CNN analysis.
Conclusions:
- Deep learning, specifically CNNs, can effectively detect Freezing of Gait episodes from wearable IMU data.
- The proposed model shows high accuracy and geometric mean, indicating its potential clinical utility.
- This approach offers a non-invasive and objective method for monitoring FoG in Parkinson's disease.
