P291 Weighted sparsity regularization for solving the inverse EEG problem: a case study
Ole Løseth Elvetun1 , Niranjana Sudheer*2
1Faculty of Science and Technology, Norwegian University of Life Sciences, P. O Box 5003, NO - 1432, Ås, Norway
2Faculty of Science and Technology, Norwegian University of Life Sciences, P. O Box 5003, NO - 1432, Ås, Norway
*Email: niranjana.sudheer@outlook.com
IntroductionWe present weighted sparsity regularization for solving the inverse EEG problem, which helps in the recovery of dipole sources while reducing depth bias. EEG is a non-invasive technique for monitoring cerebral activity. However, it suffers from ill-posed inverse problems due to weak signals from deep sources. Common standard regularization methods solutions have been suggested to tackle this problem but has significant spatial dispersion. This study proposes a redundant basis approach combined with a weighted sparsity term to improve the recovery and lower spatial dispersion, while reducing the depth bias. MethodsOur approach is based on theoretical results established in previous studies, but modifications are required to align with the classical EEG framework [1,2]. Generally, any dipole at a particular location can be expressed as a combination of three basis dipoles with independent orientations. We will illustrate that employing more than three dipoles, specifically a redundant basis or frame, can enhance localization accuracy. We produce simulated event-related EEG data utilizing SEREEGA [3], an open-source MATLAB toolbox with 64, 131, and 228 electrode channels. Simulations with three different dipole orientations, such as fixed, limited, and free, are conducted, and performance is analyzed using dipole localization error (DLE), spatial dispersion (SD), and Earth Mover’s Distance (EMD) [3]. Results & DiscussionThe proposed method performs better than sLORETA and Lp-norm approaches with lower DLE values and reduced spatial dispersion. The frame-based methodology guarantees effective recovery of dipoles, especially in noise-free environments. We noticed that an increase in the number of frame dipoles resulted in reduced localization errors. The localization accuracy improves when the number of EEG channels is increased, particularly in the limited orientation setup. A real-world test using EEG Motor Movement data [4,5] showed the practical application of this approach. ConclusionWeighted sparsity regularization provides an effective approach to EEG inverse problems, enhancing dipole localization and minimizing depth bias. The method is effective for various dipole orientations and adaptable for real-world applications.
Acknowledgements I would like to thank my supervisor Ole Løseth Elvetun and co - supervisor Bjørn Fredrik Nielsen for providing guidance and support throughout the research. I am also grateful to my friends and family for their encouragement and support. References1.https://doi.org/10.1515/jiip-2021-0057 2.https://doi.org/10.1090/mcom/3941 3.https://doi.org/10.1016/j.jneumeth.2018.08.001 4.https://doi.org/10.1109/TBME.2004.827072 5.https://doi.org/10.1161/01.CIR.101.23.e215
