Convolutional neural networks-based early Parkinson's disease classification using cycling data from a steerable...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Machine Learning

Background:

• Parkinson's disease (PD) diagnosis relies on clinical assessment, which may miss early-stage functional changes.
• Detecting short-term variations in disability is challenging during standard medical evaluations for PD.

Purpose of the Study:

• To investigate the feasibility of using a Convolutional Neural Network (CNN) model for early Parkinson's disease detection.
• To evaluate the effectiveness of analyzing cycling dynamics for PD identification.

Main Methods:

• Utilized a specialized bicycle platform (Ultiracer) with lateral movement capabilities to simulate outdoor cycling.
• Equipped the bicycle with 6-axis force-torque sensors in the headset and seat post.
• Input data included 30-second cycling metrics (force, moment, speed, lateral movement) and personal information (sex, age, height, weight) for a CNN model.

Main Results:

• The CNN model achieved approximately 86% accuracy in classifying Parkinson's disease patients versus healthy controls.
• Classification performance was evaluated using 5-fold cross-validation on data from 29 PD patients and 36 healthy individuals.

Conclusions:

• The proposed CNN model demonstrates significant potential for early Parkinson's disease detection using cycling-based biomechanical data.
• This approach offers a novel, objective method to supplement traditional diagnostic techniques for PD.