Localization of spatially distributed brain sources after a tensor-based preprocessing of interictal epileptic EEG...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Accurate localization of spatially distributed epileptic sources from electroencephalography (EEG) is crucial for understanding and treating epilepsy.
• Traditional tensor models for EEG data can be challenging when dealing with multiple extended sources or noisy signals.
• Developing robust source localization techniques is essential for clinical applications.

Purpose of the Study:

• To investigate and compare different tensor-based preprocessing strategies for localizing spatially distributed epileptic sources from EEG data.
• To evaluate the efficacy of the Canonical Polyadic (CP) model for space-time-spike tensors derived from interictal epileptic activity.
• To establish an efficient source localization scheme that overcomes limitations of correlated sources and background noise.

Main Methods:

• Utilizing tensor-based preprocessing on interictal epileptic EEG data.
• Constructing a space-time-spike tensor from amplitude-modulated spikes originating from epileptic regions.
• Investigating and comparing three source localization strategies: the 'disk algorithm', standardized Tikhonov regularization, and a fused LASSO scheme.
• Testing the methods on realistic simulated data.

Main Results:

• The study demonstrates the feasibility of using space-time-spike tensors for source localization when the CP model is applicable.
• Performance comparison of the 'disk algorithm', Tikhonov regularization, and fused LASSO revealed varying efficiencies under different noise and source correlation conditions.
• The proposed tensor-based preprocessing and subsequent localization schemes show promise for improved spatial resolution of epileptic foci.

Conclusions:

• Tensor-based preprocessing, particularly with space-time-spike tensors, offers a viable approach for localizing epileptic sources.
• The choice of localization strategy (disk algorithm, Tikhonov, or LASSO) depends on the specific characteristics of the EEG data, such as source correlation and noise levels.
• Further research and validation on clinical data are warranted to translate these findings into practical diagnostic tools for epilepsy management.