Regression Networks for Neurophysiological Indicator Evaluation in Practicing Motor Imagery Tasks

Area of Science:

• Neuroscience
• Machine Learning
• Rehabilitation Engineering

Background:

• Motor Imagery (MI) is crucial for motor learning in skills, sports, and rehabilitation.
• Significant inter- and intra-subject variability hinders MI training effectiveness for up to 30% of users.
• Understanding individual brain network efficiency in MI is key to overcoming these limitations.

Purpose of the Study:

• To develop a data-driven estimator, the Deep Regression Network (DRN), for assessing individual brain network efficiency during MI tasks.
• To analyze the distinctiveness between subject groups with similar variability in MI performance.
• To enhance the understanding of Brain-Computer Interface (BCI) inefficiency in subjects.

Main Methods:

• A novel data-driven estimator, Deep Regression Network (DRN), was developed.
• The DRN employs a double-stage approach: pattern extraction using deep learning and subsequent regression analysis.
• The model infers distinctiveness between subject groups with similar variability in MI tasks.

Main Results:

• The DRN successfully predicts bi-class accuracy response based on pre-training neural desynchronization and initial training synchronization.
• The estimator demonstrates effectiveness on real-world MI data.
• The findings highlight the DRN's capability to foster synchronization patterns crucial for successful MI practice.

Conclusions:

• The Deep Regression Network (DRN) offers a powerful tool for assessing individual brain network efficiency in Motor Imagery (MI).
• This approach can help identify and potentially mitigate factors contributing to Brain-Computer Interface (BCI) inefficiency.
• The findings pave the way for more personalized and effective MI training protocols.