RESEARCH ARTICLE

Low- and High-Frequency Oscillations Reveal Distinct Absence Seizure Networks Jeffrey R. Tenney, MD, PhD,1,2 Hisako Fujiwara, BS,1 Paul S. Horn, PhD,1,3 Jennifer Vannest, PhD,1,2 Jing Xiang, MD, PhD,1,2 Tracy A. Glauser, MD,1 and Douglas F. Rose, MD1 Objective: The aim of this study was to determine the frequency-dependent, spatiotemporal involvement of corticothalamic networks to the generation of absence seizures. Methods: Magnetoencephalography recordings were obtained in 12 subjects (44 seizures) with untreated childhood absence seizures. Time-frequency analysis of each seizure was performed to determine bandwidths with significant power at ictal onset. Source localization was then completed to determine brain regions contributing to generalized spike and wave discharges seen on electroencephalogram. Results: Significant power in the time-frequency analysis was seen within 1 to 20Hz, 20 to 70Hz, and 70 to 150Hz bandwidths. Source localization revealed that sources localized to the frontal cortex similarly for the low- and gamma-frequency bandwidths, whereas at the low-frequency bandwidth (3–20Hz) significantly more sources localized to the parietal cortex (odds ratio [OR] 5 16.7). Cortical sources within the high-frequency oscillation (HFO) bandwidth (70–150Hz) localized primarily to the frontal region compared to the parietal (OR 5 7.32) or temporal (OR 5 2.78) areas. Interpretation: Neuromagnetic activity within frontal and parietal cortical regions provides further confirmation of hemodynamic changes reported using functional magnetic resonance imaging that have been associated with absence seizures. The frequency-dependent nature of these networks has not previously been reported, and the presence of HFOs during absence seizures is a novel finding. Co-occurring frontal and parietal corticothalamic networks may interact to produce a pathological state that contributes to the generation of spike and wave discharges. The clinical and pathophysiological implications of HFOs within the frontal cortical region are unclear and should be further investigated. ANN NEUROL 2014;76:558–567

A

bsence seizures are characterized as multiple, brief impairments of consciousness that have an abrupt onset and termination.1 They are unique among seizure types due to their characteristic bilaterally synchronous 3Hz spike and wave discharges (SWDs) by electroencephalography (EEG) and pharmacologic treatment. Controversy has ensued regarding whether the structures that initiate the SWDs responsible for absence seizures are in the cortex or thalamus, but recent neuroimaging and

electrophysiological techniques are helping to provide an answer.2–11 Functional magnetic resonance imaging (fMRI) is a technique that provides additional insight into the widespread networks involved in the generation of absence seizures. fMRI studies in humans and animal models have implicated numerous locations within the cortex (frontal, parietal, temporal, insula), thalamus (anterior, centromedian, parafascicular, and reticular nuclei),

View this article online at wileyonlinelibrary.com. DOI: 10.1002/ana.24231 Received May 6, 2014, and in revised form Jul 11, 2014. Accepted for publication Jul 14, 2014. Address correspondence to Dr Tenney, Division of Neurology, Cincinnati Children’s Hospital Medical Center, 3333 Burnet Ave, Cincinnati, OH, 45229. E-mail: [email protected] From the 1Division of Neurology, Cincinnati Children’s Hospital Medical Center, University of Cincinnati, Cincinnati, OH; 2Pediatric Neuroimaging Research Consortium, Cincinnati Children’s Hospital Medical Center, University of Cincinnati, Cincinnati, OH; and 3Department of Mathematical Sciences, University of Cincinnati, Cincinnati, OH.

C 2014 American Neurological Association 558 V

Tenney et al: Absence Seizure Networks

caudate nuclei, cerebellum, and brainstem reticular formation.12–25 Magnetoencephalography (MEG) allows an assessment of neural activity measured on a submillisecond time scale, which makes it a very useful tool for the study of epileptic networks. Two studies of idiopathic generalized epilepsy with MEG showed that there were frontocentral, peri-insular, and thalamic activations that occurred both locally and diffusely during SWDs.26,27 A study of children with medically refractory absence epilepsy using MEG showed focal, rather than diffusely generalized, cortical activity prior to the onset of SWDs.4 Our previous work has demonstrated that at absence seizure onset frontal and thalamic sources occurred simultaneously.28 These results lend support to the idea that MEG can detect focal and subcortical activation during absence seizures, although areas of source localization have varied. We propose that MEG recordings of generalized absence seizures contain significant spectral power changes in higher (gamma) frequency bands (20–70Hz) than usually studied (3–20Hz), including high-frequency oscillations (HFOs; 70–150Hz). Specifically, we hypothesize that sources within these different frequency bandwidths may reveal distinct corticothalamic networks.

Patients and Methods Subjects Twelve children (6–12 years old), with newly diagnosed childhood absence seizures were recruited from the Neurology Division at Cincinnati Children’s Hospital Medical Center. The research protocol received approval by the institutional review board prior to study recruitment. Informed consent was obtained from the parent of each subject, and informed assent was obtained from each child. This patient population has been described previously.28 Briefly, inclusion criteria included a diagnosis of childhood absence seizures, normal development, normal neurological examination, normal brain MRI, and bilaterally synchronous 3- to 4Hz SWDs on a normal background with at least one electroclinical seizure lasting 3 seconds or more. Children were not eligible if there was a history of seizures other than absence seizures (generalized tonic–clonic or myoclonic) or if they were currently taking an antiepileptic medication.

MEG Recordings Simultaneous EEG and MEG recordings were conducted using a 275-channel CTF magnetometer that can record signals from direct current to 12kHz per channel (CTF Systems, VSM MedTech, Coquitlam, BC, Canada). First, 25 MEG-compatible EEG electrodes were placed in the conventional 10–20 arrangement along with 1 channel for electrocardiogram (ECG). Fiducial markers were applied to 3 locations: at the nasion and each preauricular. A head localization procedure was performed

October 2014

before and after each acquisition to locate the patient’s head relative to the coordinate system fixed to the MEG system. The coordinates of the 3 fiducial points, at the nasion and left and right preauricular, are determined electronically relative to the system Dewar coordinate frame. A new coordinate frame relative to the patients’ head is derived from these positions. The change in head location before and after acquisition was required to be 10 years old at the time of diagnosis, but in retrospect the onset of symptoms occurred prior to the age of 10 years for these subjects. A total of 44 seizures occurred during MEG recording, with an average duration of 10.5 seconds (range 5 4.4–23.6 seconds). These and other clinical characteristics have been previously reported.28

Group analysis of sLORETA statistical maps for all seizures, at each frequency bandwidth, was completed. Each time window was examined for the location of activity and classified as frontal cortex, thalamus, parietal cortex, temporal cortex, or occipital cortex, as defined previously.28 These results were summed

Time-Frequency Analysis During all seizures, significant power was seen in the 1- to 20Hz bandwidth that was maximal at 3 to 4Hz (Fig 1). Power changes were also seen in the 20- to 70Hz

560

Volume 76, No. 4

Tenney et al: Absence Seizure Networks

FIGURE 1: Time-frequency analysis using short-time Fourier transform was completed for each absence seizure captured during magnetoencephalographic recording. Each electroencephalographic segment (Sz) encompassing the seizure onset can be seen at the top; below are the time-frequency spectrograms. The degree of power was qualitative and represented according to the color bar. The temporal/frequency resolution was adjusted to provide the optimal frequency resolution with the narrowest time interval.

bandwidth for all seizures, with maximal changes tending to occur between 22 and 30Hz during most seizures. In the 70- to 150Hz bandwidth, the power changes were less consistent within and between subjects. Eight of the 44 seizures had no appreciable spectral power changes between 70 and 150Hz. All subjects had at least 1 seizure with a spectral power change within this bandwidth except for Subject 3, who had no change for either of the 2 seizures that were recorded. These HFOs are unlikely to be related solely to harmonics of the spikes, given that they occur both before and after. No significant spectral power changes were seen in the 0.1- to 1Hz and 150- to 300Hz bandwidths, so these were not included for source localization analysis. Source Localization Statistical maps using an sLORETA algorithm were generated for all 44 seizures that occurred during MEG recording. Results were analyzed at each of the 3 bandwidths described (3–20Hz, 20–70Hz, 70–150Hz). These results were coregistered to each subject’s brain MRI so that the maps could be compared for each frequency bandwidth (Fig 2). At the lower-frequency bandwidth (3–20Hz), sources tended to localize within the cortex more posteriorly, in the parietal region, as well as the October 2014

thalamus. At the gamma-frequency bandwidth (20– 70Hz), there was more anterior localization of sources, within the frontal cortex and in many instances to a focal region of the lateral prefrontal and orbitofrontal cortex. At the HFO bandwidth (70–150Hz), source localization was almost exclusively confined to the prefrontal and orbitofrontal cortex as well as some subcortical localization in the thalamus. Lateralized HFOs within the prefrontal cortex were a reproducible finding across subjects, and the side of lateralization was consistent at the intrasubject but not intersubject level. Statistical Analysis To provide a group summary of source localization results, those sources that localized to each brain region (frontal cortex, parietal cortex, temporal cortex, occipital cortex, thalamus) were determined using visual inspection (Fig 3, Table 2). Sources localized to the frontal cortex in almost 100% of seizures at the gamma-frequency bandwidth (20–70Hz). However, the difference in the proportion of activity between the low (3–20Hz) and gammafrequency (20–70Hz) bandwidths was not significant (p 5 0.139) due to 1 subject who was an outlier with no frontal activity at the gamma-frequency bandwidth (see Fig 3). At the low-frequency bandwidth (3–20Hz), 561

ANNALS

of Neurology

FIGURE 2: Representative example of statistical maps from standardized low-resolution electromagnetic tomography (sLORETA) analysis for a single seizure. (Left) Butterfly plots of magnetoencephalographic tracings with the time window used for analysis. The diagonal lines show the window that was selected for analysis within each frequency band. (Middle) Contour maps from independent component analysis showing the 2 main components. (Right) Four views of the brain (top, right, left, cut plane showing thalamic activity) overlaid with sLORETA statistical maps. CDR 5 current density reconstruction.

significantly more sources localized to the parietal cortex (odds ratio [OR] 5 16.67) when compared to the gamma and HFO bandwidths. As we reported previously, very few sources localized to the occipital cortex around the time of seizure onset, regardless of frequency bandwidth. No statistics could be completed for the HFO bandwidth because there were no seizures with occipital source localization. Sources at the HFO bandwidth (70–150Hz) were more likely to localize to the frontal cortex than the parietal (OR 5 7.33), temporal (OR 5 4.43), or occipital (OR not calculated) cortices. Source localization results were also compared for individual subjects (Fig 4). These are the same data as described above but visualized on a per subject basis to examine any intrasubject and intersubject variability. At the low-frequency bandwidth (3–20Hz), most subjects had sources with more widespread localization and fairly consistent localization to parietal cortex and thalamus. It is apparent that at the gamma-frequency bandwidth (20– 70Hz), the majority of sources for all subjects consistently localized to the frontal cortex. HFOs localized specifically to the frontal cortex, but whether HFOs were present was not a consistent finding across subjects.

Discussion The findings of this study, showing spectral power changes in higher-frequency bandwidths (20–70Hz) than usually studied (3–20Hz), including HFOs (70–150Hz), in a homogenous group of subjects with new onset 562

childhood absence seizures, provide new insights into the pathophysiology of human absence seizures and confirmation of findings from invasive recordings in animal models. Specifically, we have shown that source localization at the onset of absence seizures, within low- and high-frequency bandwidths, reveals distinct, simultaneous parietal and frontal corticothalamic networks. This is in agreement with several previous EEG-fMRI studies that have highlighted the importance of the frontal cortex, thalamus, and precuneus to the development and propagation of SWDs characterizing absence seizures.14,15,31,32 Our study identified clear parietal source localization at seizure onset within the low-frequency bandwidth (3–20Hz) that was studied. This finding is consistent with the robust negative blood oxygenation level– dependent (BOLD) signal changes that have been reported in previous EEG-fMRI studies of absence seizures.14,15,31,32 It is remarkable that such similarities are seen, given the differences in temporal resolution between MEG recordings and BOLD signal changes. The mechanism driving this spatial concordance between MEG source localization and the negative BOLD response is unclear, but a recent study showed similar findings where the largest magnitude negative BOLD responses were associated with the greatest EEG power of mu rhythms.33 These findings support the idea that a negative BOLD response may not simply be the opposite of positive BOLD but rather a unique neurovascular mechanism. Volume 76, No. 4

Tenney et al: Absence Seizure Networks

FIGURE 3: Areas of source localization from standardized low-resolution electromagnetic tomography analysis for 3- to 20Hz, 20- to 70Hz, and 70- to 150Hz bandwidths. The active areas for each frequency bandwidth were summed for all 44 seizures. The proportion of seizures where the corresponding area was active is seen on the y-axis. Comparison between the frequency bandwidths was done using a mixed logistic regression model with subject as a random effect and the response as proportion of activation. *p < 0.05, **p < 0.001. HFO 5 high-frequency oscillation.

The importance of posterior cortical networks has been highlighted using dynamic causal modeling, which has shown that the corticothalamic loop and hence SWDs are dependent upon the state of the precuneus region.31 Additionally, local field potential measurements and nonlinear association analysis in the WAG/Rij model of absence seizures demonstrated a consistent cortical focus within the perioral somatosensory region, which led to the development of SWDs.3 Functionally, there are distinct parietal regions that are thought to represent part of the posterior default mode network and are associated with states of awareness; however, our results demonstrate lateral rather than medial parietal lobe localization.16,18,34,35 Therefore, these findings, along with our own, raise the possibility that fluctuations within the parietal cortex may facilitate the development of “reverberating” SWDs within the frontal corticothalamic network. Another possibility is that asynchronous oscillations in the anterior (frontothalamic) and posterior October 2014

(parietothalamic) networks define the “normal” state, but when the oscillations of these networks occur synchronously, there is a transition to an ictal state. We also report simultaneous source localization at gamma-frequency bandwidths (20–70Hz, 70–150Hz) within the frontothalamic regions. These areas have been consistently reported in previous EEG-fMRI and MEG studies and are thought to be the critical components of the rhythmic nature of SWDs during absence seizures.4,14,15,18,26,27,31,32,36 A novel finding of our study is that MEG activity in a specific region of the lateral prefrontal cortex is present within the HFO range. This consistent frontal cortical area of HFO localization is similar in location to the lateral orbitofrontal area reported to have increased resting functional connectivity in patients with childhood absence epilepsy.37 The implication of this finding is unclear, but other studies have demonstrated long-range synchrony in primary generalized seizures up to 55Hz, HFOs up to 140Hz during 563

ANNALS

of Neurology

TABLE 2. Comparison of Standardized Low-Resolution Electromagnetic Tomography Source Localization Results within Each Bandwidth in the Respective Brain Regions

Region Frontal

Parietal

Temporal

Occipital

Thalamus

Bandwidth Comparison

OR (95% CI)

p

3–20Hz vs 20–70Hz

0.11 (0.01–2.05)

0.139

20–70Hz vs 70–150Hz

27.63 (1.56–490.56)

0.024

3–20Hz vs 70–150Hz

3.14 (1.45–6.81)

0.004

3–20Hz vs 20–70Hz

16.67 (6.25–50.00)

Low- and high-frequency oscillations reveal distinct absence seizure networks.

The aim of this study was to determine the frequency-dependent, spatiotemporal involvement of corticothalamic networks to the generation of absence se...
476KB Sizes 4 Downloads 4 Views