Introducing the Concept of Error Vectors to Improve Source Localization Results of Epileptic Discharges
Year: 2025
Type: article
Abstract: PURPOSE: To improve EEG source localization results of interictal epileptic discharges (IED) by applying postprocessing step to electrical source imaging (ESI). METHODS: Localization error of ESI was evaluated in comparison to known sources of stimulation potentials (ESP) by recording simultaneous stereo-EEG/scalp EEG. Error vectors were defined as the offset of the ESI-dipole of ESP to the stereo-EEG contacts used for stimulation. The inverted error vector was applied to the ESI-dipole of IED (IED-dipole). RESULTS: Seven IED clusters were evaluated. Corrected IED-dipoles were located closer to IED-onset contacts on stereo-EEG than uncorrected IED-dipoles ( median [ IQR ]: 7.8 [2.5] versus 18.7 [10.7] mm, P = 0.02). Anatomically, for high skull conductivities, all corrected IED-dipoles were located in cortical structures or adjacent to epileptogenic lesion, whereas uncorrected IED-dipoles were located in white matter or CSF ( P = 0.02). Physiologically, cortical extent of IED generators estimated from corrected IED-dipoles was 16.5 cm 2 ( IQR = 10.4 cm 2 ) and 7.4 cm 2 (range 5.8-9.2 cm 2 ) in the group of anterior temporal IED and prefrontal IED, respectively; the former was concordant with the extent estimated by subdural electrodes. In addition, the relationship of stereo-EEG IED amplitude (a) drop with increasing distance (d) from corrected IED-dipole could be modeled as a negative power equation a(d)∝1/d b ( R2 = 0.87, P < 0.01), with b ranging from 0.79 to 2.3, median: 1.57, consistent with a simulation model of the sensitivity of intracerebral electrode. CONCLUSIONS: Application of inverted error vector reduces localization error and shifts IED-dipole to an anatomically and physiologically plausible location.
Authors: Kanjana Unnwongse, Carsten H. Wolters, Tim Wehner, Lia Theophilo Krüger, Stefan Rampp, Jörg Wellmer
Language: English
Subfield: Cognitive Neuroscience
Field: Neuroscience
Domain: Life Sciences
Open Access status: hybrid
OpenAlex ID: W4410823896