Page 134 - MEGIN Book Of Abstracts - 2023

to compare independent component analysis (ICA) Novel noninvasive identification of patient-specific
followed by dipole source localization and the linearly epileptic networks in focal epilepsies: Linking
constrained minimum-variance beamformer (LCMV- single-photon emission computed tomography
BF) for characterizing regions with interictal epileptic perfusion during seizures with resting-state
activity and their dynamic connectivity. After a simula- magnetoencephalography dynamics (2023)
tion study, we compared ICA and LCMV-BF results with
intracerebral EEG (stereotaxic EEG, SEEG) recorded Krishnan, Balu; Tousseyn, Simon; Wang, Zhong Irene;
simultaneously in 8 epileptic patients, which provide a Murakami, Hiroatsu; Wu, Guiyun; Burgess, Richard;
unique 'ground truth' to which non-invasive results can Iasemidis, Leonidas; Najm, Imad; Alexopoulos, Andreas
be confronted. We compared the signal time courses V
extracted applying ICA and LCMV-BF on MEG data to
that of SEEG, both for the actual signals and the dy- Neurological Institute, Epilepsy Center, Cleveland Clinic,
namic connectivity computed using cross-correlation Cleveland, Ohio, USA; Academic Center for Epileptology, Kem-
(evolution of links in time). With our simulations, we il- penhaeghe and Maastricht UMC+, Heeze, The Netherlands;
lustrated the different effect of the temporal and spatial Department of Neurology, Barrow Neurological Institute,
correlation among sources on the two methods. While Scottsdale, Arizona, USA
ICA was more affected by the temporal correlation but
robust against spatial configurations, LCMV-BF showed ABSTRACT Single-photon emission computed to-
opposite behavior. Moreover, ICA seems more suited to mography (SPECT) during seizures and magnetoen-
retrieve the simulated networks. In case of real patient cephalography (MEG) during the interictal state are
data, good MEG/SEEG correlation and good localization noninvasive modalities employed in the localization of
were obtained in 6 out of 8 patients. In 4 of them ICA the epileptogenic zone in patients with drug-resistant
had the best performance (higher correlation, lower focal epilepsy (DRFE). The present study aims to inves-
localization distance). In terms of dynamic connectiv- tigate whether there exists a preferentially high MEG
ity, the evolution in time of the cross-correlation links functional connectivity (FC) among those regions of
could be retrieved in 5 patients out of 6, however, with the brain that exhibit hyperperfusion or hypoperfusion
more variable results in terms of correlation and dis- during seizures. We studied MEG and SPECT data in 30
tance. In two patients LCMV-BF had better results than consecutive DRFE patients who had resective epilepsy
ICA. In one patient the two methods showed equally surgery. We parcellated each ictal perfusion map into
good outcomes, and in the remaining two patients ICA 200 regions of interest (ROIs) and generated ROI time
performed best. In conclusion, our results obtained by series using source modeling of MEG data. FC between
exploiting simultaneous MEG/SEEG recordings suggest ROIs was quantified using coherence and phase-
that ICA and LCMV-BF have complementary qualities locking value. We defined a generalized linear model
for retrieving the dynamics of interictal sources and to relate the connectivity of each ROI, ictal perfusion
their network interactions. z score, and distance between ROIs. We compared the
coefficients relating perfusion z score to FC of each ROI
Keywords: Beamformer, Brain connectivity, Epileptic and estimated the connectivity within and between
network, ICA, MEG, Simultaneous recordings, Source resected and unresected ROIs. We found that perfusion
reconstruction z scores were strongly correlated with the FC of hyper-,
and separately, hypoperfused ROIs across patients.
NeuroImage (2023), Vol. 265 (36513288) (0 citations) High interictal connectivity was observed between
hyperperfused brain regions inside and outside the
resected area. High connectivity was also observed be-
tween regions of ictal hypoperfusion. Importantly, the
ictally hypoperfused regions had a low interictal con-
nectivity to regions that became hyperperfused during

 ontents  Index 113
C