Decoding trajectories of imagined hand movement using electrocorticograms for brain-machine interface.
Sang Jin Jang1, Yu Jin Yang2, Seokyun Ryun2
1Korea Advanced Institute of Science and Technology, Bio and Brain Engineering, 411 E16-1(YBS Building) Korea Advanced Institute of Science and Technology, 291 Daehak-ro, Yuseong-gu, Daejeon 34141, Republic of Korea.
Journal of Neural Engineering
|August 19, 2022
Summary
Researchers successfully decoded imagined hand movement trajectories using electrocorticography and a machine learning model. This advance is crucial for developing brain-computer interfaces (BCIs) for movement-free control in individuals with motor impairments.
Area of Science:
- Neuroscience
- Biomedical Engineering
- Machine Learning
Background:
- Brain-computer interfaces (BCIs) are vital for restoring function in individuals with motor impairments.
- Decoding imagined hand movements is essential for developing movement-free control systems.
- Hand trajectory, representing 3D hand positions, is a key component of motor skill analysis.
Purpose of the Study:
- To decode the trajectory of imagined hand movements using electrocorticography (ECoG) data.
- To evaluate the effectiveness of a variational Bayesian least squares model for this decoding task.
- To compare decoding accuracy between kinesthetic movement imagination (KMI) and a mixed movement execution and imagination (MEKMI) paradigm.
Main Methods:
- ECoG data were collected from 18 epilepsy patients during movement execution (ME) and kinesthetic movement imagination (KMI) of reach-and-grasp actions.
- A variational Bayesian least squares machine learning technique was employed to analyze the neural signals.
- Decoding accuracy was assessed by comparing predicted hand trajectories with actual imagined trajectories using Pearson's correlation coefficient.
Main Results:
- The variational Bayesian decoding model successfully predicted imagined hand movement trajectories above chance levels.
- Pearson's correlation coefficients of 0.3393 (KMI only) and 0.4936 (MEKMI paradigm) were achieved.
- Higher decoding accuracy for imagined trajectories was observed in the MEKMI paradigm compared to the KMI-only paradigm.
Conclusions:
- This study demonstrates the feasibility of accurately predicting imagined hand movement trajectories from ECoG signals.
- The MEKMI paradigm shows enhanced decoding performance, suggesting potential benefits for BCI applications.
- These findings represent a significant step towards developing advanced BCIs for assistive control and rehabilitation.
