Recent machine learning advancements in sensor-based mobility analysis: Deep learning for Parkinson's disease...

Area of Science:

• Biomedical Engineering
• Neurology
• Machine Learning

Background:

• Wearable sensors enable long-term monitoring of movement disorders.
• Current methods for in- and out-of-clinic motor symptom assessment require improvement.

Purpose of the Study:

• To investigate deep learning for analyzing wearable sensor data.
• To assess deep learning's efficacy in monitoring motor symptoms of Parkinson's disease.

Main Methods:

• Collected data from ten idiopathic Parkinson's disease patients using inertial measurement units.
• Expert-labeled motor tasks for classification, focusing on bradykinesia detection.
• Compared deep learning (convolutional neural networks) with standard machine learning pipelines.

Main Results:

• Deep learning achieved a higher classification rate than state-of-the-art machine learning algorithms by at least 4.6%.
• Demonstrated deep learning's potential for accurate bradykinesia detection.

Conclusions:

• Deep learning is a promising method for sensor-based movement disorder assessment.
• Highlights the advantages and disadvantages of deep learning in this context.