Epileptic discharges affect the default mode network--FMRI and intracerebral EEG evidence

Abstract

Functional neuroimaging studies of epilepsy patients often show, at the time of epileptic activity, deactivation in default mode network (DMN) regions, which is hypothesized to reflect altered consciousness. We aimed to study the metabolic and electrophysiological correlates of these changes in the DMN regions. We studied six epilepsy patients that underwent scalp EEG-fMRI and later stereotaxic intracerebral EEG (SEEG) sampling regions of DMN (posterior cingulate cortex, Pre-cuneus, inferior parietal lobule, medial prefrontal cortex and dorsolateral frontal cortex) as well as non-DMN regions. SEEG recordings were subject to frequency analyses comparing sections with interictal epileptic discharges (IED) to IED-free baselines in the IED-generating region, DMN and non-DMN regions. EEG-fMRI and SEEG were obtained at rest. During IEDs, EEG-fMRI demonstrated deactivation in various DMN nodes in 5 of 6 patients, most frequently the pre-cuneus and inferior parietal lobule, and less frequently the other DMN nodes. SEEG analyses demonstrated decrease in gamma power (50-150 Hz), and increase in the power of lower frequencies (<30 Hz) at times of IEDs, in at least one DMN node in all patients. These changes were not apparent in the non-DMN regions. We demonstrate that, at the time of IEDs, DMN regions decrease their metabolic demand and undergo an EEG change consisting of decreased gamma and increased lower frequencies. These findings, specific to DMN regions, confirm in a pathological condition a direct relationship between DMN BOLD activity and EEG activity. They indicate that epileptic activity affects the DMN, and therefore may momentarily reduce the consciousness level and cognitive reserve.

Figure 1. EEG-fMRI results and Time-Frequency (TF) statistical maps of SEEG recordings from patient 5.

A – Scalp EEG showing electrographic seizure over perisylvian region and associated BOLD changes. Functional map is co-registered with anatomical MRI and CT showing the sampling of P-C, insula and temporal cortex by SEEG electrodes (marked by white arrows, other electrodes are not shown). Functional maps show activations in left insula and bilateral cingulate and deactivations in DMN nodes of left hemisphere. B - SEEG traces showing typical electrographic seizure originating from left insula (lower 2 traces) and SEEG in left PCC, P-C, IPL, DLPFC, inferior occipital and superior temporal regions. Each region was sampled with 2 consecutive bipolar channels 5 mm apart, and higher number denotes a more lateral channel. Note the spread of seizure activity to the IPL and temporal cortex and that the activity in IPL is identical to the activity measured on the scalp. C – Schematic representation of a typical electrographic seizure (not in scale) and 1 s epochs centered at the start, center, and end of the seizure. Epochs can overlap depending on the length of the IED. D – TF maps of statistically significant z values comparing start, center, and end of IED with baseline. They show decrease in gamma and increase in frequencies<30 Hz coinciding with the center of the seizure in PCC, P-C, DLPFC, whereas IPL and superior temporal gyrus show increase of frequencies up to 60 and 120 Hz, respectively, secondary to spread of epileptic activity. Boxed TF maps are shown in higher magnification on the right. Statistical threshold obtained after false discovery rate correction (indicated as “1” in color bar). Plots are saturated at a z value that is double the threshold value (indicated as “2”). Non-significant pixels (below false discovery rate threshold) are displayed in green.

Figure 2. EEG-fMRI results and Time-Frequency (TF) statistical maps of SEEG recordings from patient 3.

A – Scalp EEG showing right temporo-parietal IED (spike and slow wave) and the associated BOLD changes. Most significant deactivation (t = -6.7) is in right IPL and concordant with the spike field. Smaller deactivations are in posterior pre-cuneus and frontal and parietal regions ipsilateral to the IED. Activations are in the cingulate gyri, thalami and ipsilateral central region. B - SEEG traces showing typical runs of spike and slow waves in right IPL (lower 3 traces) and in right PCC and middle cingulate (MC). C – TF plots comparing IEDs and baseline in PCC, MC and the channels with IEDs (IPL). The IPL is the spiking channel and part of the DMN. TF maps centered at the center of IED run show decrease in gamma in the channels with IEDs (IPL). Boxed TF map is shown in higher magnification on the right and demonstrates power decrease in 40–160 Hz band, and increase in frequencies <30 Hz.

Figure 3. EEG-fMRI results and Time-Frequency (TF) statistical maps of SEEG recordings from patient 1.

A – Scalp EEG showing right temporal IED and the associated BOLD changes. Functional map is co-registered to anatomical MRI acquired with the SEEG electrodes, showing the sampling of PCC, IPL and ACC. Functional maps show significant activations in right frontal and parietal lobes and deactivation in cuneus. B - Right posterior temporal-occipital sharp wave was associated with right inferior temporal activation and left central deactivation (not shown in figure). EEG-fMRI did not show DMN deactivation in A&B. C - SEEG traces showing typical runs of hippocampal IEDs (lower 3 traces) and SEEG in PCC, IPL and ACC. Most medial channel in PCC (upper trace) shows propagation of epileptic activity. D – TF plots comparing IEDs with baseline in PCC, IPL, ACC and the IED-generating channel (hippocampus). TF maps are centered at the start, center and end of IED, showing statistically significant increase in frequencies <30 Hz in the PCC (reflecting spike propagation) and decrease in gamma coinciding with the hippocampal IED. Boxed TF maps shown in higher magnification on the right illustrate power change in PCC and IED-generating channel. Power decrease in PCC is in 80–140 Hz band, and starts 250 ms before IED onset.

Figure 4. Relative power spectral density in dB, for frequencies 0–50 Hz.

Power spectra in the DMN regions with gamma decrease and the non-DMN regions in all patients comparing 1 s epochs at the center of IED to baseline. Blue trace is the most medial, followed laterally by the green and red traces. S1–S6 relates to patients 1–6. Note the power increase in frequencies <30 Hz in all DMN regions with gamma decrease, but only in 2/10 non-DMN locations: Temporal cortex in patient 5 shows increase in low frequencies, secondary to propagation of epileptic activity (Fig. 1B), and LOF contacts of patient 6 also due to propagation of IEDs originating from left SMA (not shown). Comparing Fig. 4 and 5 shows an inverse relationship between the decrease in gamma and the increase in the lower frequencies in many DMN nodes: contacts that showed more gamma decrease also had more increase in lower frequencies: For examples, compare the responses in L PCC in patient 4 and L P-C in patient 5.

Figure 5. Relative power spectral density in dB, for frequencies 50–250 Hz.

Power spectra in the DMN regions with gamma decrease and the non-DMN regions in all patients comparing 1 s epochs at the center of IED to baseline. Blue trace is the most medial, followed laterally by the green and red traces. S1–S6 relates to patients 1–6. Note the consistency of gamma decrease in DMN regions and the lack of similar spectral changes in non-DMN contacts.