Related Experiment Video
Updated: Jun 19, 2025
08:05
Design and Analysis for Fall Detection System Simplification
Published on: April 6, 2020
10.6K
Multi-Shared-Task Self-Supervised CNN-LSTM for Monitoring Free-Body Movement UPDRS-III Using Wearable Sensors.
Mustafa Shuqair1, Joohi Jimenez-Shahed2, Behnaz Ghoraani1
1Department of Electrical Engineering and Computer Science, Florida Atlantic University, Boca Raton, FL 33431, USA.
Bioengineering (Basel, Switzerland)
|July 27, 2024
Summary
This study introduces a deep learning framework using wearable sensors to accurately estimate Parkinson
Area of Science:
- Bioengineering and Biomedical Signal Processing
- Artificial Intelligence in Healthcare
- Neurology and Movement Disorders
Background:
- The Unified Parkinson's Disease Rating Scale (UPDRS) is essential for diagnosing Parkinson's disease (PD) and monitoring its progression.
- Accurate UPDRS assessments are critical for effective PD management and treatment adjustments.
- Current UPDRS evaluation methods can be subjective and time-consuming, necessitating more objective and continuous monitoring solutions.
Purpose of the Study:
- To develop an innovative framework integrating deep learning and wearable sensor technology for precise UPDRS Part III score estimation.
- To enhance the accuracy of Parkinson's disease severity assessments during naturalistic daily activities.
- To establish a new benchmark for objective and continuous PD symptom monitoring.
Main Methods:
- Utilized a novel Multi-shared-task Self-supervised Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) framework.
- Processed raw gyroscope signals and their spectrogram representations from wearable sensors.
- Trained and validated the model using 526 minutes of motion data from 24 Parkinson's disease patients during everyday activities.
Main Results:
- Achieved a strong correlation of 0.89 between estimated and clinically assessed UPDRS-III scores.
- Demonstrated superior performance compared to single and multichannel CNN, LSTM, and standard CNN-LSTM models.
- Established a new state-of-the-art in UPDRS-III score estimation for free-body movements.
Conclusions:
- The developed deep learning framework offers a robust and reliable method for continuous Parkinson's disease monitoring.
- This bioengineering application significantly advances PD management by enabling precise symptom assessment in real-world settings.
- The findings pave the way for improved patient care through objective and data-driven insights into PD progression.
