Edited by: Patrick William Carney, The Florey Institute of Neuroscience and Mental Health, Australia
Reviewed by: Andy P. Bagshaw, University of Birmingham, UK; Francesca Pittau, Service de Neurologie/Hôpitaux Universitaires de Genève, Switzerland
*Correspondence: Julia Jacobs, Department of Neuropediatrics and Muscular Diseases, University Medical Center Freiburg, Mathildenstrasse 1, 79106 Freiburg, Germany e-mail:
This article was submitted to Brain Imaging Methods, a section of the journal Frontiers in Neuroscience.
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
EEG-fMRI is a non-invasive method to identify epileptic networks activated by IEDs in patients with focal and generalized epilepsy (Gotman et al.,
Additionally to positive BOLD changes, negative BOLD changes also called deactivations are observed related to IEDs (Archer et al.,
An improved understanding of negative BOLD responses is important to facilitate the interpretation of BOLD responses in a clinical setting. Moreover negative BOLD responses may provide additional information about the effect of IEDs on the patient's brain. Many of the observed negative BOLD responses occur in the precuneus, posterior cingulate, bilateral inferior parietal and mesial prefrontal cortices, structures which are known to be part of the default mode network (DMN) (Archer et al.,
In epilepsy, the strongest negative BOLD in the DMN was observed during absence seizures in idiopathic generalized epilepsy (Moeller et al.,
Recently a number of fast fMRI sequences have been developed (Lin et al.,
Patients with focal epilepsies who were admitted to the Epilepsy Centre Freiburg were included in this study. All patients signed informed consent and the study was approved by the Research Ethics Committee of the University of Freiburg.
EEG-fMRI data were only acquired in patients who fulfilled the following criteria:
ability to stay calmly in the MRI scanner over a period of 1 h and
frequent IEDs (>10 in 60 min) recorded on routine EEG outside the scanner.
All patients underwent scanning with the EEG-MREG sequence for 20–40 min depending on ability to cooperate.
A 64-channel scalp EEG was continuously recorded inside the MRI scanner (3-Tesla Trio Tim, Siemens Healthcare, Erlangen, Germany) with a reference located between Fz and Cz. Sintered Ag/AgCl ring electrodes were attached using a “BrainCap” (Easycap, Herrsching, Germany), which is part of the MR-compatible EEG recording system “BrainAmp-MR” (Brain Products, Munich, Germany). Electrode impedances were kept below 15 kΩ. An electrode was placed perivertebrally on the left for acquisition of the electrocardiogram. Data was transmitted from the amplifier (5 kHz sampling rate synchronized with the 10 MHz scanner clock, 0.016–250 Hz band-pass filter) via an optic fiber cable to a computer located outside the scanner room (Mandelkow et al.,
A 3D, T1-weighted anatomical acquisition (MPRAGE,
Gradient artifacts were corrected offline by an averaged artifact subtraction method (Allen et al.,
IEDs were marked by two independent reviewers (Julia Jacobs and Katharina Körbl) and were classified into distinct types for each patient according to spatial distribution and morphology (if more than one type was present), verifying that they were similar to epileptic discharges recorded in routine EEG outside the scanner. IED-like transients occurring in a window of 150–500 ms following the QRS complex in the ECG were not marked to avoid including residual ballistocardiographic (BCG) artifact in the analysis (Flanagan et al.,
fMRI images were reconstructed from the raw MREG data (Hugger et al.,
IEDs with distinct spatial distribution were analyzed as separate regressors. Motion parameters obtained from the motion correction step and cardio-respiratory regressors based the synchronized recording of the physiological unit of the MRI scanner (Glover et al.,
The following brain regions were considered to be part of the default-mode network: precuneus, posterior cingulate, bilateral inferior parietal and mesial prefrontal cortices (Raichle et al.,
- medial prefrontal regions: frontal superior orbital, frontal superior medial, frontal medial orbital, rectus, cingulum anterior
- lateral inferior parietal regions: parietal inferior, angular, supramarginal
- posterior cingulate:
- precuneus:
The atlas was co-registered to each patient's anatomical image, resulting in individual default-mode network templates on which clusters of significant IED-related negative BOLD responses were overlaid.
For each patient, the volume of the default-mode templates, regions of significant negative BOLD responses, and overlap between the two were used to generate the following two measures:
- Percentage of the overall negative BOLD associated with a given IED type, which is located within the structures of the default-mode network
- Percentage of default-mode structures covered by significant negative BOLD changes
The calculated percentages of overlap were then correlated with the number of IED for each IED type using a Spearman correlation. Percentage of overlap was compared for temporal vs. extra-temporal IED using a
Fifteen consecutive patients were included. Six patients showed one, four patients two, four patients three and one patient four distinct IED types. Thus, a total of 30 distinct IED types could be analyzed in this study, 12 of which were classified as temporal IED. Clinical details of all patients are given in Table
1 | 26 | m | 7y | Structural | CPS | Hypothalamic harmatoma | LEV, OXC, LCM |
2 | 36 | m | 7y | Structural TLE | CPS | MTS R | LTG, LCM |
3 | 17 | m | 13y | Unclear | SPS/CPS GTCS | Normal | LTG, OXC |
4 | 17 | f | 16y | Structural TLE | CPS | Unclear mass in the L superior T gyrus | OXC |
5 | 27 | m | 11y | Structural FLE | SPS/CPS GTCS | Surgical cavity F L | LEV, OXC |
6 | 12 | m | 9y | Structural FLE | CPS | Caveroma F R | none |
7 | 9 | m | 4y | Structural FLE | SPS/CPS | Extensive R polymicrogyria | LEV, VPA |
8 | 28 | f | 11y | FLE of unclear origine | SPS/CPS GTCS | Unclear lesion F R, including insular cortex. | LTG, LCM |
9 | 71 | m | 70y | Structural TLE | CPS | Cystic tumor mesio-temporal L | VPA |
10 | 31 | f | 31y | TLE | CPS | Normal | OXC |
11 | 60 | m | 40y | Structural TLE | CPS | Defect /sclerosis T pole L. | OXC |
12 | 23 | m | 14y | Structural TLE | SPS, CPS | FCD T L | LTG |
13 | 40 | f | 23y | Bilateral TLE | CPS, GTCS | Malrotation HC R | LTG |
14 | 14 | f | 7y | Structural FLE | CPS | FCD F R | VPA, OXC |
15 | 16 | m | 1y | Structural PLE | CPS | Tuberous sclerosis | LEV, ZNS |
Details about all IED types and resulting BOLD responses are given in Table
1 | 1 | TP right | T | 6 | 11 | 6 |
2 | 2 | F T right | T | 4 | 12 | 14 |
3 | T right | T | 1 | 19 | 9 | |
3 | 4 | F right | EX-T | 7 | 10 | 5 |
5 | PO Left | EX-T | 2 | 15 | 6 | |
6 | F left | EX-T | 13 | 11 | 3 | |
4 | 7 | T right | T | 1 | 18 | 4 |
8 | P right | Ex-T | 2 | 21 | 7 | |
9 | F right | Ex-T | 3 | 9 | 6 | |
5 | 10 | C left | Ex-T | 46 | 17 | 5 |
11 | T left | T | 8 | 16 | 32 | |
12 | F pole left | Ex-T | 9 | 10 | 2 | |
13 | F right | Ex-T | 3 | 10 | 4 | |
6 | 14 | T left | T | 2 | 12 | 7 |
7 | 15 | F pole right | Ex-T | 1 | 0 | 0 |
16 | FP right | Ex-T | 1 | 24 | 2 | |
8 | 17 | F left | Ex-T | 3 | 3 | 2 |
18 | FP right | Ex-T | 21 | 4 | 5 | |
19 | T right | T | 2 | 5 | 0 | |
9 | 20 | F pole left | Ex-T | 2 | 6 | 5 |
21 | F left | Ex-T | 2 | 6 | 4 | |
22 | T right | T | 2 | 3 | 2 | |
10 | 23 | T right | T | 1 | 6 | 2 |
11 | 24 | FT left | T | 2 | 10 | 5 |
25 | CP right | Ex-T | 1 | 7 | 0 | |
12 | 26 | FT left | T | 10 | 10 | 14 |
13 | 27 | T right | T | 3 | 15 | 29 |
28 | P right | Ex-T | 12 | 14 | 2 | |
14 | 29 | F pole right | Ex-T | 1 | 21 | 13 |
15 | 30 | F pole right | Ex-T | 4 | 8 | 5 |
Mean and SD | 11.2 ± 6.2 | 6.7 ± 8 |
The average size of negative BOLD in temporal IED was with 154.9 ± 126.1 cm3 was significantly larger than in extra-temporal with 78.7 ± 56.4 cm3 (
The average percentage of overall negative BOLD responses found within the DMN structures was 11.2 ± 6.1%. Again a large variation was seen between IED types. Seventeen patients showed between 10 and 20% overlap and 3 patients more than 20% overlap (see Table
There was no significant correlation between the number of single IED per IED type and the amount of overlap. There was no significant difference between IEDs of temporal (11.5 ± 5%) and extra-temporal origin (10.9 ± 6.6%) (Figure
The average percentage of DMN structures covered by negative BOLD was 6.7 ± 7 cm3. Three patients showed between 10 and 20% overlap and two patients more than 20% overlap (see Table
There was no significant correlation between the number of IEDs per IED type and the amount DMN structures covered by negative BOLD. There was a significantly larger area of DMN structures covered by negative BOLD for IED with temporal origin (10.4 ± 10.3%) that IED of extra-temporal origin (4.2 ± 2.9%,
Six patients had both temporal and extra-temporal IEDs (for details see Table
The present study confirms the observation that a high number of patients with focal epilepsy show alterations in the DMN during focal IED occurrence. This observation is mainly possible due to the use of MREG which increases the sensitivity of event-related fMRI for IEDs. The amount of negative BOLD in the DMN was highly variable between distinct types of IED and stronger in IEDs generated in the temporal lobe. This suggests that different IEDs may affect attention and consciousness to variable degrees and it may be of clinical interest for patients with epilepsies to identify those subtypes with a large effect on important networks such as the DMN.
The present analysis of single patients was to a large extent only possible as a result of the increased sensitivity of the MREG sequence (Zahneisen et al.,
As with all EEG-fMRI studies, it is important to exclude sources of artifact which may result in incorrect BOLD responses. One advantage of MREG is the ability to measure un-aliased physiological artifacts such as respiration and ECG, which could then be corrected as part of our analysis (LeVan et al.,
A second potential source for mistakes during EEG-fMRI is the false detection of motion artifacts such as the ballistocardiogram as IED (Flanagan et al.,
In the present study IEDs occurring during the time window of the ballistocardiogram were not included in the analysis. This measure likely resulted in the exclusion of true IEDs and thus in a decreased sensitivity of the identified BOLD responses (Flanagan et al.,
It should be pointed out that we did not preselect patients in regard to whether they showed any negative BOLD for this analysis. Therefore, it is not surprising that the area covered by negative BOLD varied between zero and 380 cubic-centimeters depending on patient and spike type. The underlying physiological mechanism of negative BOLD is largely unknown. One theory has suggested that it might result from a “vascular steal” phenomenon, which implies that neighboring areas of increased blood flow and BOLD cause a decreased blood flow and negative BOLD (Harel et al.,
Another possible explanation for the occurrence of negative BOLD might be a decreased neuronal activity at the time of IEDs in these regions. This explanation is in line with the observation of increased concentrations of the inhibitory transmitter GABA in regions of negative BOLD (Chatton et al.,
The DMN was originally discovered as a network of structures which are active or show positive BOLD during periods of rest in contrast to the actual activity under examination (Shulman et al.,
Our study confirmed previous observations that the degree to which IEDs affect the DMN is largely variable. While a clear correlation between cognitive decline and disruption of the DMN has been observed in Alzheimer's dementia (Broyd et al.,
The first hypothesis would suggest a short repetitive interruption of consciousness or normal function of the region generating the IED. Such loss of consciousness is especially observed in generalized epilepsies where trains of generalized spikes and waves lead to interrupted consciousness during absence seizures. These seizures are one of the very few seizure types which can be observed during EEG-fMRI as they occur frequently and without excessive motion. Absence seizures are associated with strong negative BOLD changes in the DMN (Blumenfeld,
Short impairments of cognitive function as described in TCI (Binnie,
The rare and scarce occurrence of DMN changes in the analysis of single patients due to the low sensitivity of classical fMRI sequences has been a major challenge for such a study design. The fact that we saw negative BOLD changes in the DMN in all patients suggests that MREG will facilitate this type of investigation and hopefully enable us to answer the question whether focal IED have the potential to interrupt consciousness or cognition.
The alternative hypothesis suggests that negative BOLD in the DMN reflects negative long-term effect of IED on cognition. This would mean that IEDs associated with DMN changes reflect stronger long-term interference with cognition than IEDs which do not deactivate the DMN. Again this question could only be answered by having long-term studies correlating DMN deactivation and cognitive decline in patients with epilepsy, as has been done for Alzheimer's dementia (Broyd et al.,
IED are often believed to reflect the epileptic potential of the underlying tissue and are usually monitored in EEG recordings to assess treatment control. While it is certainly true that frequent inter-ictal activity is associated with severity of disease and cognitive decline in some epileptic syndromes such as continuous spike wave status (CSWS) in sleep (Pera et al.,
It was the aim of the present study to investigate the overlap between DMN and negative BOLD occurrence. As we hypothesized this overlap was larger in temporal than neocortical IEDs, as has been shown in previous group analysis (Laufs et al.,
Recently TLE is increasingly believed to be a network disease and permanent changes such as atrophy of brain regions have been described even far from the focus (Spencer,
In the present study quantification of overlap between DMN regions and negative BOLD occurrence after focal IED revealed involvement of DMN structures to varying extent in all patients. MREG as a method of fast fMRI allows very sensitive detection of BOLD changes in the DMN structures. Interestingly the frequency of IEDs did not affect the occurrence of negative BOLD in the DMN, but the origin of IED did. Thus, preexisting network structures seem to be a relevant factor for the ability of an IED to interfere with the DMN. IED generated from the mesial temporal structures which are part of the DMN or their vicinity result in stronger interruption of the DMN activity. The clinical implications of these findings are unknown, but if spontaneous repetitive IEDs interrupt the resting state networks of the brain in similar but less directed way as external stimuli, a resulting impairment of consciousness and cognition are likely.
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
This work was supported by the ERC grant agreement 232908 and the DFG cluster of excellence EXC-1086 BrainLinks-BrainTools. Julia Jacobs was supported by grant JA 1725/2-1 of the German Research Foundation.
Blood Oxygenation Level Dependent
default mode network
Echo Planar Imaging
functional magnetic resonance imaging
inter-ictal epileptic discharge
Magnetic-Resonance-Encephalography
Temporal Lobe Epilepsy.