Evaluation of cortical current density imaging methods using intracranial electrocorticograms and functional MRI

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Electroencephalography (EEG) source imaging noninvasively estimates neural activity.
• Accurate source localization is crucial for understanding brain function.
• Evaluating different source reconstruction algorithms is essential for improving EEG analysis.

Purpose of the Study:

• To compare the performance of various EEG source imaging algorithms.
• To assess techniques using intracranial electrocorticograms (ECoGs) and functional MRI (fMRI) as ground truth.
• To determine the most effective algorithm for imaging well-localized neural sources.

Main Methods:

• Evaluated five source imaging algorithms: MNLS, LORETA with L(p)-norm (p=1, 1.5, 2), sLORETA, minimum L(p)-norm (p=1, 1.5), and L(1)-norm.
• Used realistic head models derived from individual MRIs.
• Validated source localization using ECoGs and fMRI data from 10 subjects undergoing somatosensory stimulation.

Main Results:

• L(p)-norm and LORETA-L(p) methods produced smoother current density distributions with higher p values.
• L(1)-norm based LORETA outperformed LORETA-L(2) for imaging localized sources like the P30 SEP component.
• L(1)-norm and minimum L(p)-norm (p=1) algorithms achieved superior localization accuracy compared to MNLS and other LORETA variants.

Conclusions:

• L(1)-norm based algorithms demonstrate superior performance for cortical current density imaging of localized sources.
• These findings are particularly relevant for analyzing somatosensory evoked potentials.
• The study highlights the advantage of L(1)-norm regularization in EEG source imaging.