Poster No:
1685
Submission Type:
Abstract Submission
Authors:
Mingai Li1, Xiang Yu Li2
Institutions:
1School of Information Science and Technology, Beijing University of Technology, Beijing, China, 2School of Information Science and Technology, Beijing University of Technology, Beijing, Beijing
First Author:
Mingai Li
School of Information Science and Technology, Beijing University of Technology
Beijing, China
Co-Author:
Xiang Yu Li
School of Information Science and Technology, Beijing University of Technology
Beijing, Beijing
Introduction:
Motor imagery EEG (MI-EEG), as an electrophysiological signal, has become a typical brain computer interface (BCI) paradigm in neurological rehabilitation system, and it is crucial to acquire movement intention by decoding MI (Pichiorri, 2015). However, decoding from EEG is hindered by the limited electrodes and personalized differences among subjects. It is potential to combine EEG source imaging technology (ESI), which can estimate MI-induced source activity from a mass of equivalent current dipoles, with deep learning at cortical level. Some researchers have engaged in selecting time of interest (TOI) of dipoles (Fang, 2022, Li, 2021) or regions of interest (ROIs) based on Brodmann partition (Dong, 2023, Hou, 2022, Mammone, 2022], but this may result in loss of partial valuable features. ROI importance is developed to highlight the distinctive contributions of Destrieux partitions for each subject, it is helpful to demonstrate the individualized features with all partitions [Wang, 2024]. Inspired by this idea, we will focus on quantifying the frequency band importance of Desikan-Killiany (DK) partition equivalent dipole to enhance temporal-frequency-spatial features for improving MI decoding performance.
Methods:
In this paper, we will propose a novel source domain MI decoding method, denoted as MIDsd. The standardized low resolution electromagnetic tomography algorithm (sLORETA) is applied to convert the multi-electrode MI-EEGs from scalp space into cortex space for dipole estimation. Then, based on DK partitions, a partition equivalent dipole (PED) is obtained by averaging all dipoles within a partition to represent each partition, and its 3-demensional (3D) coordinates are thought as the cube center of a partition. Next, continuous wavelet transform (CWT) is used to calculate the time-frequency map of PED, which is divided into multiple frequency bands ( (1-3Hz), (4-7Hz), (8-12Hz) and (13-30Hz)), and input into random forest algorithm (RF) to obtain the importance of each band (BI). Furthermore, BI is utilized to weight the energy sequence of each band to generate a comprehensive energy sequence for each partition, which is separated into multiple time segments, and the total energy of each segment is calculated. Thus, a 4D dipole feature map is constructed by combining 3D coordinates of PED. Finally, a Convolutional Neural Network (CNN) with 3D Convolutional Block Attention Module (3DCBAM) (denoted as 3DCBAM-CNN) is designed for feature extraction and classification.
Results:
Based on the public BCI Competition IV 2a dataset with four MI tasks, extensive experiments are conducted. The proposed MIDsd achieves the average accuracy of 88.50%, and the statistical characteristic indicators, such as standard deviation, Kappa value and confusion matrix, are calculated for statistical analysis.
Conclusions:
A novel source domain MI decoding method (MIDsd) is proposed in this paper. The results show that PED can effectively represent the comprehensive performance of induced dipoles by MI within a partition, it is reasonable to reflect the different contributions of multi bands by introducing BI, and BI is capable of enhancing subject-based temporal-frequency-spatial features of PED as well. Also, 3DCBAM-CNN can match the characteristics of 4D dipole feature maps, improving decoding performance in source domain. The statistical analysis indicates better stability and consistency of MIDsd.
Modeling and Analysis Methods:
EEG/MEG Modeling and Analysis 2
Motor Behavior:
Brain Machine Interface 1
Keywords:
Computational Neuroscience
Design and Analysis
Electroencephaolography (EEG)
Machine Learning
Other - Motor Imagery
1|2Indicates the priority used for review
By submitting your proposal, you grant permission for the Organization for Human Brain Mapping (OHBM) to distribute your work in any format, including video, audio print and electronic text through OHBM OnDemand, social media channels, the OHBM website, or other electronic publications and media.
I accept
The Open Science Special Interest Group (OSSIG) is introducing a reproducibility challenge for OHBM 2025. This new initiative aims to enhance the reproducibility of scientific results and foster collaborations between labs. Teams will consist of a “source” party and a “reproducing” party, and will be evaluated on the success of their replication, the openness of the source work, and additional deliverables. Click here for more information.
Propose your OHBM abstract(s) as source work for future OHBM meetings by selecting one of the following options:
I do not want to participate in the reproducibility challenge.
Please indicate below if your study was a "resting state" or "task-activation” study.
Task-activation
Healthy subjects only or patients (note that patient studies may also involve healthy subjects):
Healthy subjects
Was this research conducted in the United States?
No
Were any human subjects research approved by the relevant Institutional Review Board or ethics panel?
NOTE: Any human subjects studies without IRB approval will be automatically rejected.
Yes
Were any animal research approved by the relevant IACUC or other animal research panel?
NOTE: Any animal studies without IACUC approval will be automatically rejected.
Not applicable
Please indicate which methods were used in your research:
EEG/ERP
Provide references using APA citation style.
Dong Y. (2023). Applying correlation analysis to electrode optimization in source domain. Medical & Biological Engineering & Computing, 61(5), 1225-1238.
Fang T. (2022). Decoding motor imagery tasks using ESI and hybrid feature CNN. Journal of Neural Engineering, 19(1), Art. no. 016022.
Hou Y. (2020). A novel approach of decoding EEG four-class motor imagery tasks via scout ESI and CNN. Journal of Neural Engineering, 17(1), Art. no. 016048.
Li M.-A. (2021). A novel decoding method for motor imagery tasks with 4D data representation and 3D convolutional neural networks. Journal of Neural Engineering, 18(4), Art. no. 046029.
Mammone N. (2020). A deep CNN approach to decode motor preparation of upper limbs from time-frequency maps of EEG signals at source level. Neural Networks, 124, 357-372.
Pichiorri F. (2015). Brain-computer interface boosts motor imagery practice during stroke recovery. Annals of Neurology, 77(5), 851-865.
Wang L. (2024). Cortical ROI Importance Improves MI Decoding from EEG Using Fused Light Neural Network. IEEE Transactions on Neural Systems and Rehabilitation Engineering, 32, 3636-3646.
No