Enhanced spatial resolution in EEG via super-resolution
Poster No:
1317
Submission Type:
Abstract Submission
Authors:
Seong Jin Cho1, Kwang-Ho Choi1, Joong Il Kim1, Yeonhee Ryu1
Institutions:
1Korea Institute of Oriental Medicine, Yuseong-gu, Daejeon, Korea, Republic of
First Author:
Co-Author(s):
Introduction:
Electroencephalography (EEG) is a non-invasive neurophysiological technique that records electrical brain activity (Teplan, 2002). It has been extensively used in various fields, including neurofeedback, brain-computer interfaces, and cognitive neuroscience. While traditional EEG systems often rely on wired setups, there is a growing demand for wireless and portable devices to enable EEG measurements in diverse environments and conditions. However, wireless EEG systems typically require a reduction in the number of measurement electrodes to minimize power consumption and weight, which can compromise spatial resolution (Niso, 2023). To address this limitation, we propose a novel approach to enhance EEG spatial resolution through super-resolution techniques. Super-resolution aims to reconstruct high-resolution signals from low-resolution input. In recent years, generative models have been explored for EEG super-resolution, but their stochastic nature may not be ideal for precise signal prediction (Corley, 2018; Kwon, 2019). In this study, we leverage a deterministic approach based on a U-Net architecture to achieve accurate EEG super-resolution. By training the U-Net on a large dataset of EEG recordings, we aim to learn the underlying patterns and dependencies between electrodes, enabling the prediction of missing or under-sampled EEG signals.
Methods:
We collected resting-state EEG data from 40 healthy participants. The EEG signals were recorded using a 64-channel system at a sampling rate of 512 Hz. Raw data were preprocessed, including bandpass filtering (2-50 Hz) and epoch extraction (2-second epochs). To determine the optimal number and placement of electrodes for super-resolution, we employed functional principal component analysis (FPCA) to identify the principal components that capture most of the information in the EEG signals. By analyzing the eigenvalues of the principal components, we determined that 20 electrodes were sufficient to retain approximately 95% of the information. To optimize electrode placement, we utilized k-means clustering to group electrodes with similar signal patterns. The electrode with the smallest distance to the cluster centroid was selected as the representative electrode for that cluster. This approach resulted in a 19-electrode configuration that identically aligned with the standard 10-20 electrode system (Homan, 1988). We employed a 1D U-Net architecture to implement the super-resolution algorithm (Hossain, 2023). The U-Net consists of an encoder-decoder structure with skip connections, allowing for efficient feature extraction and propagation (Fig. 1). The network was trained on a dataset of 12,000 epochs, with 7,200 for training, 2,400 for validation, and 2,400 for testing. The Adam optimizer was used with a learning rate of 0.001. The performance of the super-resolution model was evaluated using Pearson's correlation coefficient between the predicted and ground-truth EEG signals. A higher correlation coefficient indicates better prediction accuracy.
·Figure 1. 1D U-Net architecture designed to enhance the spatial resolution of EEG signals.
Results:
The proposed U-Net model demonstrated impressive performance in super-resolving EEG signals from 19 to 64 channels (Fig. 2a, b). The average Pearson's correlation coefficient across all epochs was 0.79±0.19, with 77% of epochs achieving a correlation coefficient of 0.7 or higher. The model was particularly effective in predicting signals in most electrode locations, with slightly lower performance observed in some occipital lobe channels (Fig. 2c).
·Figure 2. Performance results of super-resolution.
Conclusions:
This study presents a novel approach to enhance EEG spatial resolution using a U-Net-based super-resolution algorithm. By carefully selecting the optimal number and placement of electrodes, we were able to reconstruct high-resolution EEG signals from low-resolution input effectively. This advancement has the potential to improve the accuracy and reliability of EEG-based brain-computer interfaces and neuroimaging studies, enabling more precise and informative insight into brain function.
Modeling and Analysis Methods:
EEG/MEG Modeling and Analysis 1
Methods Development 2
Other Methods
Novel Imaging Acquisition Methods:
EEG
Imaging Methods Other
Keywords:
Acquisition
Development
Electroencephaolography (EEG)
Machine Learning
Modeling
1|2Indicates the priority used for review
Abstract Information
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.
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:
Please indicate below if your study was a "resting state" or "task-activation” study.
Healthy subjects only or patients (note that patient studies may also involve healthy subjects):
Was this research conducted in the United States?
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.
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.
Please indicate which methods were used in your research:
Provide references using APA citation style.
Homan, R. W. 1988. 'The 10-20 electrode system and cerebral location.' American Journal of EEG Technology, 28(4): 269–279.
Hossain, M. S., S. Mahmud, A. Khandakar, N. Al-Emadi, F. A. Chowdhury, Z. B. Mahbub, and M. E. Chowdhury. 2023. 'MultiResUNet3+: A full-scale connected multi-residual UNet model to denoise electrooculogram and electromyogram artifacts from corrupted electroencephalogram signals.' Bioengineering, 10(5): 579.
Kwon, M., S. Han, K. Kim, and S. C. Jun. 2019. 'Super-resolution for improving EEG spatial resolution using deep convolutional neural network-feasibility study.' Sensors, 19: 5317.
Niso, G., E. Romero, J. T. Moreau, A. Araujo, and L. R. Krol. 2023. 'Wireless EEG: A survey of systems and studies.' NeuroImage, 269: 119774.
Teplan, M. 2002. 'Fundamentals of EEG measurement.' Measurement Science Review, 2(2): 1–11.