volume | PIER Journals

Abstract

This paper presents a novel high-resolution brain source reconstruction method for Brain-Computer Interface (BCI) applications using a deep learning-based direct inversion approach. The proposed framework integrates electroencephalography (EEG) data simulated via the FieldTrip toolbox and leverages a modified U-Net architecture trained to directly estimate the active and inactive cortical source regions. Unlike traditional inverse methods such as Minimum Norm Estimation (MNE), LORETA, and Lasso the proposed method bypasses the computational complexity of analytical solutions and offers faster inference times once trained. Experimental results using a database of 50,000 synthetic models demonstrate a reconstruction accuracy of up to 61.66% under optimized conditions, with a validation loss of 0.6372 and an F1 score of 61.12%. The method shows improved detection of active brain regions in central cortical areas and delivers robust spatial reconstructions compared to conventional numerical techniques. Although performance on certain edge cases remains limited, the proposed framework offers a promising direction for scalable, real-time source localization in diagnostic and neurorehabilitation applications.

1. Zorzos, Ioannis, Ioannis Kakkos, Errikos M. Ventouras, and George K. Matsopoulos, "Advances in electrical source imaging: A review of the current approaches, applications and challenges," Signals, Vol. 2, No. 3, 378-391, 2021.
doi:10.3390/signals2030024 Google Scholar

2. Vorwerk, Johannes, Robert Oostenveld, Maria Carla Piastra, Lilla Magyari, and Carsten H. Wolters, "The fieldtrip-simbio pipeline for EEG forward solutions," Biomedical Engineering Online, Vol. 17, No. 1, 37, 2018.
doi:10.1186/s12938-018-0463-y Google Scholar

3. Waldert, Stephan, Hubert Preissl, Evariste Demandt, Christoph Braun, Niels Birbaumer, Ad Aertsen, and Carsten Mehring, "Hand movement direction decoded from MEG and EEG," Journal of Neuroscience, Vol. 28, No. 4, 1000-1008, 2008.
doi:10.1523/jneurosci.5171-07.2008 Google Scholar

4. Ahn, S., D. Kim, J. H. Hong, and S. C. Jun, "Effect of realistic human head modelling on brain source distribution," Electronics Letters, Vol. 48, No. 18, 1095-1097, 2012.
doi:10.1049/el.2012.1569 Google Scholar

5. Khemakhem, Rafik, Wassim Zouch, A. Ben Hamida, Abdelmalik Taleb-Ahmed, and Imed Feki, "EEG source localization using the inverse problem methods," IJCSNS International Journal of Computer Science and Network Security, Vol. 9, No. 4, 408, 2009. Google Scholar

6. Pantazis, Dimitrios and Amir Adler, "MEG source localization via deep learning," Sensors, Vol. 21, No. 13, 4278, 2021.
doi:10.3390/s21134278 Google Scholar

7. Borra, Davide, Francesco Bossi, Davide Rivolta, and Elisa Magosso, "Deep learning applied to EEG source-data reveals both ventral and dorsal visual stream involvement in holistic processing of social stimuli," Scientific Reports, Vol. 13, No. 1, 7365, 2023.
doi:10.1038/s41598-023-34487-z Google Scholar

8. Abibullaev, Berdakh, Aigerim Keutayeva, and Amin Zollanvari, "Deep learning in EEG-based BCIs: A comprehensive review of transformer models, advantages, challenges, and applications," IEEE Access, Vol. 11, 127271-127301, 2023.
doi:10.1109/access.2023.3329678 Google Scholar

9. Sun, Rui, Abbas Sohrabpour, Gregory A. Worrell, and Bin He, "Deep neural networks constrained by neural mass models improve electrophysiological source imaging of spatiotemporal brain dynamics," Proceedings of the National Academy of Sciences, Vol. 119, No. 31, e2201128119, 2022.
doi:10.1073/pnas.2201128119 Google Scholar

10. Oostenveld, Robert, Pascal Fries, Eric Maris, and Jan-Mathijs Schoffelen, "FieldTrip: Open source software for advanced analysis of MEG, EEG, and invasive electrophysiological data," Computational Intelligence and Neuroscience, Vol. 2011, No. 1, 156869, 2011.
doi:10.1155/2011/156869 Google Scholar

11. Knösche, Thomas R. and Jens Haueisen, EEG/MEG Source Reconstruction, Springer, 2022.
doi:10.1007/978-3-030-74918-7

12. Morik, Marco, Ali Hashemi, Klaus-Robert Müller, Stefan Haufe, and Shinichi Nakajima, "Enhancing brain source reconstruction through physics-informed 3D neural networks," ArXiv Preprint ArXiv:2411.00143, 2024.
doi:10.48550/arXiv.2411.00143 Google Scholar

13. Raissi, M., P. Perdikaris, and G. E. Karniadakis, "Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations," Journal of Computational Physics, Vol. 378, 686-707, 2019.
doi:10.1016/j.jcp.2018.10.045 Google Scholar

14. Enßlin, Torsten A., Mona Frommert, and Francisco S. Kitaura, "Information field theory for cosmological perturbation reconstruction and nonlinear signal analysis," Physical Review D, Vol. 80, No. 10, 105005, 2009.
doi:10.1103/physrevd.80.105005 Google Scholar

15. Nielsen, Frank, "An elementary introduction to information geometry," Entropy, Vol. 22, No. 10, 1100, 2020.
doi:10.3390/e22101100 Google Scholar

Mr. Babak Ojaroudi Parchin

Dr. Mehdi Nooshyar

Dr. Mohammad Ojaroudi