Premium
Comparison of magnetic source estimation to intracranial EEG , resection area, and seizure outcome
Author(s) -
Tenney Jeffrey R.,
Fujiwara Hisako,
Horn Paul S.,
Rose Douglas F.
Publication year - 2014
Publication title -
epilepsia
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 2.687
H-Index - 191
eISSN - 1528-1167
pISSN - 0013-9580
DOI - 10.1111/epi.12822
Subject(s) - concordance , magnetoencephalography , retrospective cohort study , electroencephalography , medicine , epilepsy , magnetic resonance imaging , radiology , nuclear medicine , audiology , psychology , surgery , psychiatry
Summary Objectives Magnetoencephalography ( MEG ) is used to guide intracranial electroencephalography ( ICEEG ) monitoring and determine areas for resection. The purpose of this retrospective cross‐sectional study was to report our experience using dipole modeling/dipole scanning, current density reconstructions, and beam‐forming methods in a large cohort of pediatric patients with intractable epilepsy. Methods Source localization results for each algorithm and seizure‐onset zone, defined by ICEEG , were described by three blinded reviewers according to five location criteria. The accuracy of each algorithm was then compared to ICEEG . The relationships between the accuracy of these algorithms (discordant, lobar concordant, sublobar concordant) and long‐term seizure outcome was calculated using positive and negative predictive values. Results Thirty‐two patients (mean age ± SD, 10.8 ± 5 years) were included in this retrospective review. No algorithms had sublobar concordance with ICEEG in all patients, including when algorithms were grouped by type (dipole modeling/dipole scanning, current density reconstruction, beam forming). Synthetic aperture magnetometry ( SAM ) with excess kurtosis tended to be the most accurate, but there were no significant differences between algorithms. When comparing the source modeling with ICEEG findings, significantly more patients with a seizure‐free outcome were found to have lobar or sublobar concordance of multiple signal classification ( MUSIC ) (61.1%) and standardized low resolution brain electromagnetic tomography (s LORETA ) (52.9%). Positive predictive values were highest for MUSIC (61.9%) and equivalent current dipole ( ECD ) (57.1%). Negative predictive values were highest for SAM ( g 2 )‐ VS (83%), minimum norm estimate ( MNE ) (75%), MUSIC (73.7%), and ECD (73.5%). Significance This study describes the use of multiple MEG source estimation techniques and demonstrates that all algorithms have similar rates of concordance with ICEEG . Also, the concordance or discordance of MUSIC with ICEEG was the best predictor of long‐term seizure outcome.