Seizure onset zone localization using postictal hypoperfusion detected by arterial spin labelling MRI

Localized cerebral blood flow reductions may underlie focal neurological deficits following seizures. Using arterial spin labelling, Gaxiola-Valdez et al. observe postictal hypoperfusion in roughly 70% of patients, which localizes to the seizure onset zone in 80% of cases. Postictal hypoperfusion may thus constitute a marker of the seizure onset zone.

Abstract

Neurological dysfunction following epileptic seizures is a well-recognized phenomenon. Several potential mechanisms have been suggested to explain postictal dysfunction, with alteration in cerebral blood flow being one possibility. These vascular disturbances may be long lasting and localized to brain areas involved in seizure generation and propagation, as supported by both animal and human studies. Therefore, measuring perfusion changes in the postictal period may help localize the seizure onset zone. Arterial spin labelling is a non-invasive, rapid and reproducible magnetic resonance imaging technique that measures cerebral perfusion. To this end, we measured postictal perfusion in patients with drug resistant focal epilepsy who were admitted to our seizure-monitoring unit for presurgical evaluation. Twenty-one patients were prospectively recruited and underwent arterial spin labelling scanning within 90 min of a habitual seizure. Patients also underwent a similar scan in the interictal period, after they were seizure-free for at least 24 h. The acquired scans were subtracted to identify the areas of significant postictal hypoperfusion. The location of the maximal hypoperfusion was compared to the presumed seizure onset zone to assess for concordance. Also, the localizing value of this technique was compared to other structural and functional imaging modalities. Postictal perfusion reductions of >15 units (ml/100 g/l) were seen in 15/21 patients (71.4%). In 12/15 (80%) of these patients, the location of the hypoperfusion was partially or fully concordant with the location of the presumed seizure onset zone. This technique compared favourably to other neuroimaging modalities, being similar or superior to structural magnetic resonance imaging in 52% of cases, ictal single-photon emission computed tomography in 60% of cases and interictal positron emission tomography in 71% of cases. Better arterial spin labelling results were obtained in patients in whom the seizure onset zone was discernible based on non-invasive data. Thus, this technique is a safe, non-invasive and relatively inexpensive tool to detect postictal hypoperfusion that may provide useful data to localize the seizure onset zone. This technique may be incorporated into the battery of conventional investigations for presurgical evaluation of patients with drug resistant focal epilepsy.

Introduction

Postictal dysfunction following epileptic seizures is a well-recognized phenomenon. This may involve transient and reversible motor, behavioural and cognitive manifestations lasting from minutes to days. Postictal dysfunction, while contributing to the disability of the epileptic disorder, may also provide some insight into the mechanisms of seizure termination and at times, the localization of the seizure onset zone (SOZ) (Fisher and Schachter, 2000).

Several potential mechanisms have been suggested to explain this postictal dysfunction (Fisher and Schachter, 2000; Koepp et al., 2010). Of the proposed hypotheses, alteration in cerebral blood flow (CBF) has been most widely tested, both in animals and humans. The striking similarity between Todd’s paresis, which is the classic model of postictal dysfunction, and ischaemic stroke led researchers to this hypothesis, where prolonged postictal hypoperfusion was first implied (Meyer, 1959) or confirmed by case reports using cerebral angiography (Yarnell, 1975), CT perfusion (Mathews et al., 2008) and MRI perfusion (Rupprecht et al., 2010).

Animal studies using diffusion-weighted MRI following chemically-induced status epilepticus in rats showed decreased apparent diffusion coefficient suggesting postictal ischaemia lasting up to 72 h (Righini et al., 1994; Nakasu et al., 1995a, b; Wang et al., 1996; Fabene et al., 2003; Choy et al., 2010). A recent study systematically demonstrated hypoperfusion and severe hypoxia (oxygen partial-pressure <10 mmHg) in the hippocampus lasting up to 1 h after hippocampal kindled seizures in rats (Farrell et al., 2016).

Postictal CBF alterations have also been demonstrated in human temporal lobe epilepsy. For example, single photon emission computerized tomography (SPECT) studies showed that temporal lobe seizures are associated with global temporal hyperperfusion during seizures, which switches to relative mesial temporal hyperperfusion and lateral temporal hypoperfusion immediately following seizures (<2 min), and finally global temporal hypoperfusion, 3–15 min following seizures (Rowe et al., 1991; Newton et al., 1992; Duncan et al., 1993). Hypoperfusion has been demonstrated with direct cerebral recordings (Weinand et al., 1994, 1997) and lateralized most frequently to the side of seizure onset. This hypoperfusion may be prolonged, lasting up to 60 min (Weinand et al., 1997). However, subsequent studies using diffusion-weighted MRI failed to show significant postictal changes (Wieshmann et al., 1997; Diehl et al., 2001; Hufnagel et al., 2003; Oh et al., 2004; Salmenpera et al., 2006).

Arterial spin labelling (ASL) is a sensitive MRI method for measuring cerebral perfusion that does not require exogenous tracer administration and can be readily performed during routine MRI examination. ASL has been used to measure perfusion and related haemodynamic parameters in normal subjects and in cerebrovascular disease (Petersen et al., 2006; Hendrikse et al., 2012). ASL has also been used to detect interictal and postictal alterations in brain perfusion dynamics in focal epilepsy to localize the SOZ. Most of these studies are small case series and yielded inconsistent results, including focal and hemispheric hypoperfusion, hyperperfusion, or no significant perfusion changes (Wolf et al., 2001; Lim et al., 2008; Pizzini et al., 2008, 2013; Altrichter et al., 2009; Nguyen et al., 2010; Pendse et al., 2010; Miyaji et al., 2014; Sugita et al., 2014). Some studies used asymmetries in CBF to lateralize the SOZ in temporal lobe epilepsy (Wolf et al., 2001; Lim et al., 2008; Eryurt et al., 2015; Guo et al., 2015; Oner et al., 2015). Larger studies performed interictal ASL and showed CBF alterations in regions concordant with the presumed SOZ (Storti et al., 2014; Boscolo Galazzo et al., 2015; Sierra-Marcos et al., 2016).

Many of the previous studies have limitations. Because CBF in relation to seizures is dynamic, inconsistent findings demonstrating hypoperfusion and hyperperfusion are likely due to the variable time points of CBF measurement. Furthermore, most studies looked at interictal CBF changes, which may not accurately reflect the SOZ. There are also methodological differences in how CBF differences were identified, including comparing one side to the other (Storti et al., 2014), using statistical differences between patients and healthy controls (Boscolo Galazzo et al., 2015) or qualitative visual assessment (Kim et al., 2016).

To circumvent these limitations, we measured postictal CBF in patients using ASL within 90 min of an electrographically confirmed seizure. We then compared these data to each patient’s interictal CBF to identify the region(s) showing postictal hypoperfusion. In a recent study of 10 subjects, we showed that postictal hypoperfusion can be measured by ASL (Farrell et al., 2016). In the present study, we studied the specific patterns of postictal hypoperfusion in a larger cohort of patients. We hypothesized that subtraction of a patient’s postictal ASL data from interictal ASL data would reliably demonstrate postictal hypoperfusion in regions that include the SOZ.

Materials and methods

Study design

Consecutive patients with drug-resistant focal epilepsy admitted to the Seizure Monitoring Unit at the Foothills Medical Centre for continuous scalp video-electroencephalography (VEEG) monitoring between January 2014 and March 2016 were prospectively enrolled in the study. All patients were admitted for presurgical evaluation. The study was approved by the Conjoint Health Research Ethics Board of the University of Calgary and all patients provided informed consent.

Clinical data collection

All patients underwent scalp VEEG monitoring using the 10–20 electrode placement system with additional surface sphenoidal electrodes. Five patients subsequently underwent intracranial VEEG monitoring, with depth and subdural electrodes being placed following standard clinical protocol at our centre. Demographic data including age, gender, age at seizure onset, duration of epilepsy, and seizure types were recorded. Antiepileptic medications were withdrawn depending on seizure frequency and habitual seizures were captured. As part of standard presurgical work-up, structural MRI using our centre’s standard epilepsy protocol was obtained for all patients. Ictal and interictal SPECT, PET, language and motor functional MRI, and neuropsychological testing were obtained as clinically indicated. Consensus opinion obtained from our weekly comprehensive, multidisciplinary epilepsy surgery rounds was documented. For patients who underwent epilepsy surgery, clinical outcomes and pathology were documented.

Definition of the seizure onset zone

The SOZ was determined by clinical history, interictal and ictal scalp VEEG, intracranial VEEG, structural MRI, PET, interictal and ictal SPECT, and expert consensus. As this was not supported by surgical outcomes in all cases, we referred to it as the ‘presumed SOZ’.

Arterial spin labelling MRI data collection

ASL data were collected using a 3 T GE Discovery MRI 750 system with a 12-channel phased array head receive coil. Each patient had an ASL scan performed within 90 min of seizure termination that was captured through VEEG and confirmed by an epileptologist. EEG electrodes were removed prior to scanning as our seizure monitoring unit does not use MRI-compatible EEG electrodes. A physician accompanied each patient as a safety precaution. Baseline, interictal ASL scans were obtained following a seizure-free period of >24 h. Foam padding was placed around patients’ heads to minimize motion during scanning. Two different MRI sequences were used during the ASL study: a 3D fast-spin-echo Pseudo-Continuous Arterial Spin Labelling (pCASL) with background suppression and a high-resolution T₁-weighted scan. The pCASL sequence had the following parameters: scan duration = 5:50 min, repetition time = 5513 ms, echo time = 4.7 ms, in-plane resolution = 1.875 × 1.875 mm, slice-thickness = 5 mm, 28 slices, post-label delay = 2525 ms, number of excitations = 4, and spiral acquisition with 1024 points and eight arms. The T₁-weighted images had the following parameters: repetition time = 9.4 ms, echo time = min full, in-plane resolution = 0.5 × 0.5 mm, and slice-thickness = 1.3 mm. Quantitative CBF maps were generated by the MRI scanner automatically from both ASL datasets in CBF units (ml/100 g/min) using the following formula:

$$CBF=6000\hspace{0.5em}\frac{\lambda \left(1-e^{-\frac{ST}{T1T}}\right)e^{\frac{PLD}{T1B}}}{2T1B\left(s\right)\left(1-e^{\frac{LT}{T1}}\right)\varepsilon \hspace{0.5em}\times \hspace{0.5em}NEX}\left(\frac{\Delta M}{SF\hspace{0.5em}\times \hspace{0.5em}P{D}_{REF}}\right)$$ (1)

where T1B and T1T represent blood and tissue T₁ values (1.6 s at 3 T), respectively, λ is the partial coefficient set to 0.9, ɛ is the efficiency and is set to 0.80 × 0.75, ΔM is the difference between tag and no tag images, PD_REF is the reference proton density images, NEX is the number of excitations, SF is a scaling factor of 45, and PLD is the post-labelling delay.

Arterial spin labelling data analysis

Both interictal and postictal quantitative CBF maps and high resolution images (T₁-weighted) were brain extracted using the BET brain extraction toolbox from FSL (BET – FSL v2.1) (Jenkinson et al., 2002). Subsequently, both CBF maps were registered onto each patient’s high-resolution anatomical scan using an affine transformation (12 degrees of freedom) from the FSL FLIRT toolbox (FLIRT – FSL v5.5) (Jenkinson et al., 2002). Using these co-registered CBF maps, a subtraction CBF map (interictal − postictal) was generated for each patient to identify brain areas where perfusion changed following a seizure relative to their own baseline. We located hypoperfused regions with a >15 CBF unit reduction, which approximately accounts for a ∼25–30% reduction in normal grey matter CBF (45–50 CBF units) (Wong et al., 1999). Additionally, the opposite subtraction (postictal − interictal) was performed to identify regions of postictal hyperperfusion (>15 CBF unit increase). A threshold of 15 CBF units was used based on animal studies showing that hippocampal seizures are followed by a minimum of 25–30% reduction of CBF (Farrell et al., 2016). In addition, in a preliminary exploratory analysis, we estimated the mean and standard deviation (SD) of CBF values for all voxels of the subtraction ASL (sASL) maps for each patient in all grey and white matter voxels within the brain. We found that 2 SD (P = 0.05) above and below the mean was ∼15 CBF units (17.56 CBF units, unpublished observations). Furthermore, the maps showing brain regions with CBF values >2 SD below the mean were virtually identical to those produced by using a lower threshold of 15 CBF units. Thus, the statistical analysis supported the use of a 15 CBF unit threshold, which is clinically significant, intuitive, and supported by animal data.

Region of interest cerebral blood flow analysis

FSLView (3.2.0) was used to draw 3D regions of interest in a blinded fashion by I.G. and reviewed by P.F./S.S. blinded to the clinical information. In cases where hypoperfusion was seen, regions of interest were drawn around the region that convincingly showed maximal hypoperfusion that was not felt to be artefactual in nature (Supplementary Table 1). Subtraction ASL (sASL) maps from each patient with a minimum threshold of 15 CBF units were used as a guide. In cases where no noticeable hypoperfusion was seen, 3D regions of interest were generated in the centre of the presumed SOZ. For cases where the SOZ was unknown (Patients 12 and 20) and a single generalized epilepsy case (Patient 17), no regions of interest were generated; thus 18 patients were considered for region of interest analysis. Next, these regions of interest were used to estimate the mean CBF values in the baseline and postictal CBF maps. The per cent change in postictal CBF relative to baseline was calculated for each region of interest and correlated to seizure duration.

Concordance of arterial spin labelling hypoperfusion and presumed seizure onset zone

Two epileptologists (P.F./S.S.) reviewed all clinical data for all patients. Based on this information, a consensus was reached for the localization of the presumed SOZ. Next, they reviewed and characterized the sASL data (as described below) to reach a consensus. The following principles were followed:

1. A confidence score from 1 to 5 (1 - not confident, 5 - very confident) was used to rate how confident each reviewer was that changes seen in the sASL data were genuine or not (i.e. artefact).

2. Most areas of hypoperfusion were accounted for, whether small or large, patchy or confluent. However, scattered and punctuate changes were not taken into account.

3. Findings confined to the surface or CSF spaces were considered to be artefact.

Hypoperfusion seen on sASL was classified on a three-tier basis:

1. Unilateral versus bilateral findings.

2. Distribution of the findings: focal (small area within a lobe, ∼3 cm³), lobar (involving >3 cm³ of a lobe), regional (involving two adjacent lobes), hemispheric (involving three or more adjacent lobes or two non-adjacent lobes in a single hemisphere), multifocal (involving two or more lobes in both hemispheres) and no change (no hypoperfusion >15 CBF units seen).

3. If applicable, subcategorization of areas of focal prominence within widespread regions of hypoperfusion was performed. Specifically, if amongst regional, hemispheric and multifocal areas of hypoperfusion there were areas of hypoperfusion that were prominent because of their size (e.g. confluent area of hypoperfusion within patchy hemispheric hypoperfusion) or degree of hypoperfusion (small but severely hypoperfused area within mild hemispheric hypoperfusion), these changes were subcategorized further as ‘with focal prominence’.

To assess concordance, the locations of ASL changes were compared to the location of the presumed SOZ. For this purpose, the most prominent area of hypoperfusion was considered. This area was compared to the epileptogenic lesion in cases of lesional epilepsy, or to intracranial EEG localization or surgical resection where applicable. In the absence of these, concordance was assessed with scalp EEG along with functional imaging localization. Concordance was classified as follows:

1. Concordant: if the presumed SOZ overlaps entirely with the area of hypoperfusion. The changes on ASL may be more extensive, but as long as the most prominent hypoperfusion matched the SOZ, it was considered concordant.

2. Partially concordant: if some, but not all of the SOZ overlapped with the area of hypoperfusion.

3. Discordant (ipsilateral): if the SOZ and the area of hypoperfusion are in different regions of the same hemisphere.

4. Discordant (contralateral): if the SOZ and the area of hypoperfusion are in opposite hemispheres.

We subdivided the brain into discrete regions to enable comparison between ASL changes and the SOZ localization. These regions were as follows:

1. Frontal lobe: divided into orbitofrontal (inferior surface), mesial frontal (medial to the interhemispheric fissure), anterior lateral frontal (anterior to the precentral sulcus), posterior lateral frontal (anterior to the central sulcus and posterior to the precentral sulcus.

2. Temporal lobe: divided into equal halves; anterior and posterior segment.

3. Parietal lobe: divided into superior (superior to intraparietal sulcus) and inferior parietal (inferior to intraparietal sulcus) segments.

4. Occipital lobe: no subdivisions were made.

Additionally, ASL results were compared to results obtained from anatomical MRI, SPECT and PET investigations individually to evaluate whether ASL subtraction maps offered additional information to localize the SOZ. Superiority was judged subjectively based on the ability to better localize the SOZ.

Results

Patient selection

Twenty-five patients entered the study protocol. Of these, one patient did not undergo a baseline ASL scan and one did not undergo a postictal scan. Thus, 23 patients were initially included in the study. Of these patients, ASL data for two patients were not analysable due to significant motion artefact, leaving 21 patients that were fully analysed.

Patient characteristics

Patient demographic data are summarized in Table 1. The mean age of the patients at the time of the study was 33.9 years (range 20–58 years). Eleven patients (52%) were female. The mean age of seizure onset was 16.1 years (range 1–43 years). All patients had frequent seizures ranging from multiple daily to monthly. Table 1 summarizes details of the clinical semiology, interictal EEG, ictal EEG, as well as structural and functional imaging data. MRI was non-lesional in eight patients. Three patients had cortical dysplasia, two showed post-surgical changes from temporal lobectomy, one each had mesial temporal sclerosis, subependymal heterotopia, bifrontal encephalomalacia, and bilateral hippocampal atrophy. Two patients had non-specific white matter hyperintensities and two had findings with uncertain significance (rounded and bulky amygdala).

Table 1.

Summary of patient demographics, clinical features, investigations, seizure localization, and postoperative outcomes

Patient ID	Age/ gender	Seizure onset age, years	Seizure burden	Seizure description	Interictal EEG	Ictal EEG
1	33 F	11	Weekly	Behavioural arrest, oral automatisms, dysphasia	Left temporal spikes (F7, SP1, T3)	Left mid-posterior temporal
2	41 F	13	Weekly	Hypermotor seizure from sleep	Left temporal spikes (SP1, F7, T3)	Unclear, obscured by artefact
3	25 M	5	Daily	Dysphasia, right side clonic movements	Left fronto-central spikes (F3, C3, Fp1)	Left fronto-central
4	33 M	7	Weekly	Fear, tachycardia, behavioural arrest	Left fronto-temporal spikes (FP1, F7, T3)	Left fronto-temporal
5	33 M	29	Weekly	Behavioural arrest, dysphasia	Bitemporal sharps (max Sp2 and Sp1)	Left hemisphere max posterior quadrant
6	42 M	22	Weekly	Left head version, secondary GTC	Bitemporal spikes right (F8, Sp2, T4) > left (F7, Sp1, T3)	Right posterior quadrant
7	40 M	9	Daily	Posturing of both upper limbs and dysphasia	Midline and left parasagittal spikes Cz, Fz, Pz, P3	Central and left parasagittal
8	26 F	1	Weekly	Vocalization, GTC	Bitemporal slowing	Bifrontal max left frontal
9	22 M	14	Weekly	Behavioral arrest, right arm dystonia, oral automatisms	Left anterior temporal sharps (max Sp1, F7)	Left fronto-temporal
10	20 F	1	Weekly	Bimanual automatisms, behavioural arrest	Bitemporal spikes right (F8, Sp2, T4) > left (F7, Sp1, T3)	Bitemporal, right > left
11	22 F	22	Weekly	Behavioural arrest, ictal speech	Right temporal slowing and spikes (F8, Sp2, T4)	Bilateral max left fronto-temporal
12	58 F	1	Weekly	Vocalization, GTC	Bitemporal slowing	Left fronto-temporal
13	55 M	43	Weekly	Manual automatisms, preserved speech	Right temporal slowing and spikes (F8, Sp2, T4)	Right anterior temporal
14	40 M	11	Monthly	Behavioural arrest, right head version, GTC	Bifrontal max left spikes (F3, Fp1, Fz)	Bifrontal max left frontal
15	21 F	15	Weekly	Right head version, GTC	Left temporal spikes (Sp1, F7)	Left hemisphere max fronto-temporal
16	38 M	34	Monthly	Oral and bimanual automatisms, dysphasia	Left temporal slowing and spikes (F7, SP1, T3)	Left anterior temporal
17	25 F	17	Weekly	Vocalization, GTC	3-4 Hz generalized spike, max bifrontal	Generalized
18	43 F	22	Daily	Behavioural arrest, oral automatisms	Right temporal spikes (F8, Sp2, T4)	Right hemisphere max temporal
19	30 F	13	Weekly	Behavioural arrest, dysphasia	Left temporal slowing	Left hemisphere max temporal
20	26 M	25	Daily	Hypermotor, partially responsive	Right frontotemporal spikes (Fp2, F8, F4)	Diffuse rhythmic slowing
21	39 F	25	Weekly	Staring, right head version, GTC	Left temporal spikes (F7, SP1, T3)	Left temporal

Patient ID	Structural MRI	Subtraction SPECT	PET	Intracranial EEG localization	Final presumed SOZ	Epilepsy surgery	Surgical outcome
1	Left mesial temporal cortical dysplasia	Not performed	Not performed	Not performed	Left temporal	No	N/A
2	Periventricular WMH	Right opercular, orbitofrontal	Not performed	Not performed	Right frontal	No	N/A
3	Normal	Left temporal, insula, operculum	Not performed	Left superior temporal gyrus	Left superior temporal gyrus	No	N/A
4	Normal	Left anterior mesial temporal	Left anterior mesial temporal	Not performed	Left insula, anterior temporal	No	N/A
5	Bilateral subependymal heterotopias max right	Bilateral temporal	Not performed	Bitemporal max left	Bilateral temporal max left	No	N/A
6	Right fat amygdala, occipital cavernoma	Right anterior temporal	Not performed	Not performed	Bitemporal right > left	No	N/A
7	Normal	Left temporal, left mesial parasagittal	Left temporal lobe and insula	Bilateral mesial frontal max left	Bilateral mesial frontal left > right	No	N/A
8	Left frontal cortical dysplasia, WMH	Left frontal	Left anterior mesial temporal	Not performed	Left frontal	No	N/A
9	Normal	Left anterior mesial temporal	Left anterior mesial temporal	Not performed	Left mesial temporal, frontal	No	N/A
10	Postsurgical (right temporal lobectomy)	Left posterior temporal	Site of previous surgery	Not performed	Right posterior temporal	No	N/A
11	Normal	Right temporal lobe	Right mesial temporal	Bilateral mesial temporal	Bilateral mesial temporal	No	N/A
12	Normal	Right frontal, left temporal	Multiple areas max left parietal	Not performed	Unclear	No	N/A
13	Right hippocampal sclerosis	Right anterior temporal	Not performed	Not performed	Right mesial temporal	Yes	Patient expired
14	Bifrontal, max left encephalomalacia	Left orbitofrontal, frontal operculum	Left hemisphere max frontal	Left frontal convexity	Left frontal	No	N/A
15	Postsurgical (right temporal lobectomy)	Left temporo-parieto-occipital	Site of previous surgery	Not performed	Left anterior temporal, frontal	No	N/A
16	Normal	Right opercular, right frontal	Left anterior temporal	Not performed	Left anterior temporal	Yes	Seizure free
17	Normal	Right frontal, right insula	Left frontal	Not performed	Generalized	No	N/A
18	Right hippocampus rounded	Right temporal, insula, parietal	Right hemisphere max temporal	Not performed	Right temporal	No	N/A
19	Left para-hippocampal cortical dysplasia	Left temporal	Not performed	Not performed	Left temporal	No	N/A
20	Punctate right frontal WMH	Right temporal, orbitofrontal, operculum	Right temporal, orbitofrontal, operculum	Not performed	Unclear	No	N/A
21	Bilateral hippocampal atrophy	Not performed	Left anterior temporal	Not performed	Left anterior temporal	No	N/A

GTC = generalized tonic-clonic seizures; WMH = white matter hyperintensity.

Localization of the presumed seizure onset zone

Based on scalp VEEG monitoring, the presumed SOZ was discernible in 15 patients (Table 1). In the other six patients, the SOZ could not be determined with certainty using non-invasive data. Of these patients, four underwent intracranial VEEG and the SOZ was established. No definite localization was possible for the remaining two patients. Based on all available information, the presumed SOZ was temporal in eight patients, bitemporal in three, bifrontal in one, temporal plus in three (temporal lobe plus another region), frontal in three, generalized in one, and unknown in two.

Arterial spin labelling data

Recorded seizures and postictal arterial spin labelling data

Table 2 summarizes the ASL data and features of the captured seizures. Only definite, habitual electroclinical seizures were used for the study. Fifteen patients experienced focal dyscognitive seizures and six experienced seizures evolving to bilateral convulsions. The seizures lasted an average of 83.8 s (range 26–166 s). Six patients had active interictal EEGs, defined as interictal epileptiform discharges seen at least once every 30 s.

Table 2.

Summary of sASL results showing the location and distribution of postictal hypoperfusion and its concordance with the SOZ

Patient ID	Final presumed SOZ	Seizure type	Seizure duration, s	Ictal EEG	Time to ASL, min	Subtraction ASL location	Unilateral versus bilateral	Distribution	Sub- distribution	Concordance with SOZ	Score 1–5
1	Left temporal	Focal dyscognitive	70	Left mid-posterior temporal	54	Left temporal convexity > mesial temporal	Unilateral	Focal	N/A	Concordant	5
2	Right frontal	Focal dyscognitive	26	Unclear, obscured by artifact	69	No change	No change	No change	N/A	N/A	5
3	Left superior temporal gyrus	Focal dyscognitive	87	Left fronto-central	83	No change	No change	No change	N/A	N/A	2
4	Left insula, temporal	Focal dyscognitive	114	Left fronto-temporal	53	Left inferior frontal operculum, anterior insula	Unilateral	Multifocal	Focal	Concordant	5
5	Bitemporal Left > right	Focal dyscognitive	64	Left hemisphere max posterior quadrant	56	Multifocal	Bilateral	Multifocal	N/A	Discordant	3
6	Bitemporal right > left	Secondary GTC	155	Right posterior quadrant	60	Right temporal-occipital-parietal; left frontal	Bilateral	Multifocal	Focal	Concordant	4
7	Bilateral mesial frontal left > right	Focal dyscognitive	85	Central and left parasagittal	54	Multifocal scattered frontal and temporal	Bilateral	Multifocal	N/A	Discordant	5
8	Left frontal	Secondary GTC	166	Bifrontal max left frontal	61	Left frontal, right posterior superior parietal	Bilateral	Multifocal	Focal	Concordant	5
9	Left mesial temporal, frontal	Focal dyscognitive	56	Left fronto-temporal	69	No change	No change	No change	N/A	N/A	5
10	Right posterior temporal	Focal dyscognitive	30	Bitemporal, right > left	45	Multiple areas near right temporal resection cavity	Unilateral	Lobar	N/A	Concordant	3
11	Bilateral mesial temporal	Focal dyscognitive	50	Bilateral max left fronto-temporal	58	Left mesial temporal, fronto-polar	Unilateral	Hemispheric	Lobar	Partially	4
12	Unclear	Secondary GTC	60	Left fronto-temporal	63	No change	No change	No change	N/A	N/A	5
13	Right mesial temporal	Focal dyscognitive	72	Right anterior temporal	75	Right temporal	Unilateral	Lobar	N/A	Concordant	5
14	Left frontal	Focal dyscognitive	60	Bifrontal max left frontal	90	Left posterior frontal and small right frontal	Unilateral	Lobar	N/A	Concordant	4
15	Left anterior temporal, frontal	Secondary GTC	120	Left hemisphere max fronto-temporal	55	Left temporal convexity	Bilateral	Multifocal	Lobar	Partially	5
16	Left anterior temporal	Focal dyscognitive	71	Left anterior temporal	59	Left mesial and neocortical left temporal	Unilateral	Focal	N/A	Concordant	5
17	Generalized	Secondary GTC	52	Generalized	71	No change	No change	No change	N/A	N/A	5
18	Right temporal	Focal dyscognitive	150	Right hemisphere max temporal	45	Right temporal neocortical	Unilateral	Lobar	N/A	Concordant	4
19	Left temporal	Focal dyscognitive	43	Left hemisphere max temporal	116	No change	No change	No change	N/A	N/A	5
20	Unclear	Focal dyscognitive	75	Diffuse rhythmic slowing	55	Multiple scattered areas	Bilateral	Multifocal	N/A	Unknown	3
21	Left anterior temporal	Secondary GTC	119	Left temporal	90	Left temporal	Unilateral	Regional	N/A	Concordant	4

GTC = generalized tonic-clonic seizures.

Postictal ASL scans were obtained 65.7 min following seizure termination (range 45–116 min). Maximal postictal CBF reductions of >15 CBF units were seen in 15/21 patients (71.4%) (Table 2). Based on the region of interest analysis in 18 patients, CBF was reduced by 14.2 ± 2.64 CBF units (range 6–40 CBF units) following seizures compared to baseline (Supplementary Table 1). In addition, the degree of postictal hypoperfusion was positively correlated with seizure duration (Fig. 1; r = 0.51; P = 0.031). As a control, regions of interest were drawn on the homologous contralateral side and the CBF measurements are summarized in Supplementary Table 2 and plotted in Supplementary Fig. 1. Subtraction ASL showed a small CBF change in the control region of interest (2.0 ± 1.33 CBF units) compared to the zone of maximal hyperperfusion or presumed SOZ (14.2 ± 2.64 CBF units). A 2 × 2 repeated measures ANOVA showed no main effect of hemisphere [F(1,17) = 1.108, P = 0.307]; however, a significant main effect of state was seen [F(1,17) = 18.753, P < 0.001].

Figure 1.

The degree of postictal hypoperfusion is directly related to seizure duration. Scatter plot of the relation between the percentage decrease of CBF (interictal − postictal scans) in the suspected SOZ and seizure duration. A significant positive correlation was seen between these two measurements (Pearson r = 0.51; two-tailed significance; P = 0.031).

Arterial spin labelling results: location of hypoperfusion

Table 3 summarizes the ASL results grouped according to the distribution of postictal hypoperfusion, as described below.

1. Focal or lobar: an example of focal postictal hypoperfusion is seen in Fig. 2, following a left temporal lobe seizure in a patient with left mesial temporal cortical dysplasia. Figure 3A shows a lobar pattern of hypoperfusion in a patient with right hippocampal sclerosis following a right temporal lobe seizure. All six patients with one of these distributions had a clearly defined SOZ on VEEG. In all cases, the sASL data correctly lateralized and localized the presumed SOZ. Four of these patients had a temporal SOZ (Patients 1, 13, 16 and 18), one had temporal plus (Patient 10) and one had a frontal SOZ (Patient 14). The certainty of sASL findings was high in all patients.

2. Regional or hemispheric: one patient had regional hypoperfusion following a mesial temporal seizure (Patient 21, Fig. 3B) and another had hemispheric hypoperfusion following a left temporal lobe seizure (Patient 11, Fig. 3C). More specifically, Patient 11 had lobar hypoperfusion within a hemispheric abnormality. Both patients had a clearly defined SOZ and sASL correctly lateralized the SOZ.

3. Multifocal: Fig. 4 shows multifocal hypoperfusion with focal prominence following a left frontal seizure. This patient had symptoms suggestive of insular involvement, which was confirmed by ictal SPECT and sASL. Of the seven patients with multifocal postictal hypoperfusion, the changes were symmetrical (e.g. not lateralized) in three (Table 3). One of these patients had a bitemporal SOZ (Patient 5), one had a bifrontal SOZ (Patient 7) and the SOZ was unknown in one patient (Patient 20). The remaining four patients had a sub-distribution of hypoperfusion (Table 3). Of these patients, one with bitemporal (Patient 6), insular (Patient 4) and frontal SOZ (Patient 8) each showed more focal hypoperfusion. One patient (Patient 15) with a temporal SOZ showed a lobar sub-distribution.

4. No change: six patients had no demonstrable hypoperfusion on sASL (Patients 2, 3, 9, 12, 17 and 19). Potential confounding factors for limited detection of postictal hypoperfusion existed for five of them. Two patients did not have a clearly determined SOZ on scalp VEEG data (Patients 2 and 12), one patient had generalized epilepsy (Patient 17), one had very frequent epileptiform discharges at baseline (Patient 9), and ASL data were acquired more than 90 min following a seizure in one patient (Patient 19).

Table 3.

Summary of ASL results grouped by distribution of hypoperfusion seen on sASL

ASL findings	Sub- distribution	Patient ID	SOZ localization on scalp EEG	SOZ location	ASL/ SOZ concordance	ASL superior to MRI?	ASL superior to SPECT?	ASL superior to PET?	Seizure type	Time to ASL (>90 min)	Interictal baseline
Focal and lobar	N/A	1	Clear	Temporal	Yes	Yes	N/A	N/A	Focal dyscognitive	No	Not active
	N/A	10	Clear	Temporal plus	Yes	Yes	Yes	Yes	Focal dyscognitive	No	Not active
	N/A	13	Clear	Temporal	Yes	No	Same	N/A	Focal dyscognitive	No	Not active
	N/A	14	Clear	Frontal	Yes	Yes	Same	Same	Focal dyscognitive	Yes	Not active
	N/A	16	Clear	Temporal	Yes	Yes	Yes	Same	Focal dyscognitive	No	Not active
	N/A	18	Clear	Temporal	Yes	Yes	Same	Same	Focal dyscognitive	No	Very active
Regional	N/A	21	Clear	Temporal	Yes	Yes	N/A	Same	Secondary GTC	Yes	Very active
Hemispheric	Lobar	11	Clear	Bitemporal	Partial	Yes	Yes	Yes	Focal dyscognitive	No	Not active
Multifocal	Focal	4	Clear	Temporal plus	Yes	Yes	Yes	Yes	Focal dyscognitive	No	Very active
	N/A	5	Unclear	Bitemporal	No	No	No	N/A	Focal dyscognitive	No	Not active
	Focal	6	Clear	Temporal	Yes	Yes	Same	N/A	Secondary GTC	No	Not active
	N/A	7	Unclear	Bifrontal	No	Same	No	Same	Focal dyscognitive	No	Very active
	Focal	8	Clear	Frontal	Yes	Yes	N/A	Yes	Secondary GTC	No	Not active
	Lobar	15	Clear	Bitemporal	Partial	Yes	Yes	Yes	Secondary GTC	No	Very active
	N/A	20	Unclear	Unknown	Unknown	Unknown	Unknown	Unknown	Focal dyscognitive	No	Not active
No change on subtraction ASL	N/A	2	Unclear	Frontal	N/A	Same	No	N/A	Focal dyscognitive	No	Not active
	N/A	3	Clear	Temporal	N/A	Same	No	N/A	Focal dyscognitive	No	Not active
	N/A	9	Clear	Temporal plus	N/A	Same	No	No	Focal dyscognitive	No	Very active
	N/A	12	Unclear	Unknown	N/A	Unknown	Unknown	Unknown	Secondary GTC	No	Not active
	N/A	17	Unclear	Generalized	N/A	Unknown	Unknown	Unknown	Secondary GTC	No	Not active
	N/A	19	Clear	Temporal	N/A	No	No	N/A	Focal dyscognitive	Yes	Not active

GTC = generalized tonic-clonic seizures; N/A = not available.

Figure 2.

Example of focal hypoperfusion seen on sASL in a 33-year-old right-handed female with intractable epilepsy. (A) Ictal EEG recording on a longitudinal bipolar montage showing seizure onset over the left anterior-midtemporal region. (B) Coronal MRI (FLAIR and T₂-weighted) showing poor grey white matter differentiation and abnormal morphology of the left parahippocampal gyrus and hippocampus, consistent with cortical dysplasia. (C) Subtraction CBF map (interictal − postictal) superimposed onto the patient’s T₁-weighted anatomical image showing areas of hypoperfusion >25 CBF units (∼40% reduction compared to normal grey matter CBF).

Figure 3.

Examples of different patterns of postictal hypoperfusion seen on sASL. (A) Coronal T₂ MRI from Patient 13 showing right mesial temporal sclerosis (arrow). Postictal ASL shows a ‘lobar’ pattern of hypoperfusion; the presumed SOZ was in the right mesial temporal lobe. (B) Interictal PET image from Patient 21 showing left temporal hypometabolism (arrow). Postictal ASL showed a ‘regional’ pattern of hypoperfusion; the presumed SOZ was in the left temporal lobe. (C) Patient 11 had a non-lesional MRI scan and independent, bilateral mesial temporal SOZs, based on intracranial VEEG monitoring. Postictal ASL from a left mesial temporal seizure showed a ‘hemispheric’ pattern of hypoperfusion on the left side of the brain.

Figure 4.

Example of multifocal hypoperfusion seen on sASL in a 33-year-old right-handed male with intractable epilepsy. (A) Ictal EEG recording on a longitudinal bipolar montage showing seizure onset in the left frontal region (arrows). (B) Ictal SPECT showing hyperperfusion in the left frontal operculum and insula. (C) Subtraction CBF map (interictal – postictal) superimposed onto the patient’s T₁-weighted anatomical image demonstrating areas of hypoperfusion >15 CBF units (top) and >25 CBF units (bottom, ∼40% reduction compared to normal grey matter CBF). ‘Multifocal’ hypoperfusion is seen on sASL (top) that is maximal in the left frontal operculum (bottom, arrow).

Concordance with seizure onset zone

In 12/15 (80%) of patients in whom hypoperfusion was seen, the location of the hypoperfusion was concordant with the presumed SOZ (Table 2). Interestingly, these patients had at least some focality to their pattern of postictal hypoperfusion (Table 3). Specifically, sASL correctly localized the presumed SOZ in all seven patients with focal, lobar, or regional hypoperfusion (Patients 1, 10, 13, 14, 16, 18 and 21). Furthermore, sASL correctly localized the presumed SOZ in all three patients with multifocal hypoperfusion with an area of focal prominence (Patients 4, 6 and 8). Subtraction ASL also lateralized (partial concordance) the SOZ in one patient with a hemispheric hypoperfusion (Patient 11) and one patient with multifocal hypoperfusion with lobar prominence (Patient 15). Thus, sASL had lateralizing and/or localizing value when localized or lateralized patterns of postictal hypoperfusion were seen.

Subtraction ASL showed multifocal, non-lateralized changes in three subjects (Patients 5, 7 and 20), which therefore did not provide information about a possible SOZ. Interestingly, all three subjects had an unclear SOZ based on scalp VEEG monitoring.

Comparison with imaging data

ASL provided better localization of the SOZ in 6/9 patients with lesional MRIs (Patients 1, 6, 8, 14, 18 and 21) and in three (Patients 5, 13 and 19), MRI was superior to ASL (Table 3). Out of 12 patients with non-lesional MRIs, sASL provided better localization of the SOZ than MRI in five cases (Patients 4, 10, 11, 15 and 16) and was the same in four (Patients 2, 3, 7 and 9). In three cases, a comparison could not be made as the SOZ was unknown (Patients 12, 17 and 20).

Ictal SPECT was performed in 15 patients. ASL provided superior localization of the SOZ in five patients (Patients 4, 10, 11, 15 and 16), similar localization in four (Patients 6, 13, 14 and 18) and inferior localization in six (Patients 2, 3, 5, 7, 9 and 19). No comparisons were made for three patients as the SOZ was unknown (Patients 12, 17 and 20).

In 11 patients (Patients 2, 4, 5, 9, 10, 11, 15, 17, 18, 19 and 20) sASL was performed on the same seizure as the ictal SPECT (Table 3). This facilitated direct comparison of postictal ASL to ictal SPECT. The localization provided by sASL was superior to ictal SPECT in four cases, inferior to ictal SPECT in four, and similar in three.

PET was performed in 14 patients (Table 3). ASL was superior to PET in five patients (Patients 4, 8, 10, 11 and 15), similar in five (Patients 7, 14, 16, 18 and 21), and inferior in one (Patient 9). No comparisons were made for three patients as the SOZ was unknown (Patients 12, 17 and 20).

Special cases: unclear seizure onset zone on scalp VEEG

Six patients had an unclear SOZ based on scalp VEEG. Of these patients, three showed no significant postictal hypoperfusion (Patients 2, 12 and 17) and three had multifocal changes without focal features (Patients 5, 7 and 20; Table 3). Two of these patients subsequently underwent intracranial VEEG, which revealed complex bilateral SOZs (Patients 5 and 7).

Special cases: patients undergoing intracranial EEG

Five patients underwent iEEG for a variety of indications. sASL showed no significant hypoperfusion in one case (Patient 3), multifocal changes without focality in two (Patients 5 and 7), lateralization of the SOZ in one (Patient 11) and localization of the SOZ in one (Patient 14).

Special cases: patients with epilepsy surgery

Two patients underwent epilepsy surgery. Patient 16 underwent a left anterior temporal lobectomy and is 1-year seizure free. Pathology showed cortical dysplasia. Patient 13 underwent right anterior temporal lobectomy, but had complications in the postoperative period and expired. Pathology showed hippocampal sclerosis. In both cases, sASL co-localized with the epileptogenic zone (Table 3 and Supplementary Fig. 2).

Special cases: patients with postictal hyperperfusion

Only two patients showed regions of hyperperfusion >15 CBF units (Patients 3 and 14; Fig. 5). Patient 3 had a non-lesional MRI and left superior temporal gyrus SOZ determined by intracranial EEG and hyperperfusion was seen over both frontal lobes, maximal on the left (Fig. 5A). Patient 14 had post-traumatic bilateral frontal encephalomalacia and a left frontal SOZ based on intracranial VEEG. Postictal hypoperfusion was seen over the left hemisphere and postictal hyperperfusion was seen over the right hemisphere (Fig. 5B). Thus, postictal hyperperfusion incorrectly lateralized the SOZ.

Figure 5.

Two patients in whom hyperperfusion was detected after seizures. (A) Twenty-five year old male with refractory epilepsy and SOZ in the left posterior temporal lobe. Postictal hyperperfusion is seen over both frontal lobes, maximally on the left. No significant postictal hypoperfusion was seen. (B) Forty year old male with refractory epilepsy and a left frontal SOZ. First row shows left hemispheric hypometabolism on PET (arrow) and left frontal hypoperfusion on sASL. Second row shows postictal hyperperfusion over the right hemisphere with a patchy distribution.

Interictal and postictal cerebral blood flow measurements

Review of the interictal state showed that only 2/21 patients (Patients 7 and 8) showed hypoperfusion in interictal CBF maps. Patient 7 had a suspected SOZ in the mesial frontal regions and sASL showed scattered multifocal hypoperfusion discordant with the presumed SOZ. The interictal and postictal ASL scans, however, both showed hypoperfusion in the left supplementary motor area (40.7 and 29.2 CBF units, respectively; control region of interest = 44.7 and 43.4 CBF units, respectively) concordant with the SOZ. Patient 8 had a presumed left frontal SOZ and interictal and postictal hypoperfusion in a right parietal lesion related to a perinatal infarct that was unrelated to the SOZ and which cancelled out on sASL. Only 10/21 patients (Patients 1, 3, 4, 7, 8 and 12–16) showed hypoperfusion in the raw postictal CBF maps, which again were concordant with the sASL results with the exception of Patient 7 as described above.

Eleven patients (Patients 1, 2, 5, 6, 8, 13, 14 and 18–21) had abnormal MRI scans. Of these, only one (Patient 8) showed interictal hypoperfusion, which was related to a known perinatal infarct distant from the SOZ. Four patients with MRI lesions showed postictal hypoperfusion at the lesion that was concordant with the SOZ in all cases (Patients 1, 8, 13 and 14).

Potential confounders

A number of confounders were considered. These included late ASL data acquisition (>90 min postictal), active interictal discharges during baseline ASL data collection, postictal ASL scanning following secondarily generalized seizures, seizure duration, and duration of epilepsy. Of these, the first three confounders may have contributed to the absence of significant postictal hypoperfusion. Specifically, postictal ASL studies were performed more than 90 min after seizures in three patients. In one patient (Patient 14), lobar changes were seen, in another (Patient 21) regional changes were seen, and in the third patient (Patient 19) no hypoperfusion was seen. Six patients had abundant interictal epileptiform discharges at baseline. Three of these patients showed multifocal postictal hypoperfusion (Patients 4, 7 and 15) and one (Patient 9) showed no hypoperfusion on sASL. Six patients had secondarily generalized seizures captured by postictal ASL. Two of these patients showed no hypoperfusion on sASL (Patients 12 and 17) and three demonstrated multifocal changes (Patients 6, 8 and 15).

Discussion

This is the first systemic study of postictal ASL. We have shown that sASL can be safely and rapidly performed in patients in the postictal state and is able to detect localized hypoperfusion. When subtracted from the patient’s baseline perfusion scan, ASL is able to demonstrate >15 CBF unit reductions (∼25% of normal grey matter CBF) in 15/21 (71%) patients. Furthermore, in 12/15 (80%) of the patients in whom hypoperfusion was seen, the location of the hypoperfusion was concordant with the presumed SOZ. The selected CBF threshold accounts for ∼2 SD of sASL map measurements in each patient (P = 0.05, unpublished observations). Overall, this fixed threshold was optimized by taking into account previous animal work, exploring different sASL thresholds in all patients visually, and calculating standard deviation measurements.

Determining the seizure onset zone

Our study is strengthened by comparing sASL data to ictal VEEG recordings, which are considered to be the gold standard for SOZ localization. Most previous ASL studies used interictal EEG or electrical source imaging of interictal discharges to determine the SOZ. A few studies (Storti et al., 2014; Boscolo Galazzo et al., 2015; Sierra-Marcos et al., 2016) used scalp VEEG to determine the SOZ, but the ASL studies were performed during the interictal period. Interictal haemodynamic changes may not correlate with and cannot be accepted as a proxy measure of the SOZ, as shown by comparisons of interictal and ictal SPECT data (Devous et al., 1998; Spanaki et al., 1999). Indeed, no previous study used continuous VEEG to accurately localize the seizure after which an ASL study was performed. This is especially important for patients with multiple SOZs. As our study was performed with accurate VEEG information on seizure duration and localization, we had the advantage of assessing concordance of the ASL data with the SOZ for that particular seizure.

Arterial spin labelling technique

It is crucial to understand how ASL functions in order to better analyse and interpret data. Most previous studies used statistical comparisons with controls to quantify the CBF changes. Additionally, most of these studies used interictal CBF data, which may not be ideal for localizing the SOZ. Indeed, minor differences in interictal CBF maps between patients and controls may provide inaccurate localization of the SOZ. Highlighting the importance of using postictal rather than interictal CBF data, recent animal work from our centre showed that CBF is reduced by 53% relative to baseline and oxygen partial pressure is reduced to the severe hypoxic range (<10 mmHg) up to 60 min following seizures (Farrell et al., 2016). Therefore, we determined the absolute CBF differences between the interictal and postictal state to identify hypoperfused areas that could identify the SOZ.

A few studies measured interictal CBF asymmetries in patients with temporal lobe epilepsy and measured CBF only over the temporal lobes (Wolf et al., 2001; Lim et al., 2008; Eryurt et al., 2015; Guo et al., 2015; Oner et al., 2015). The asymmetric perfusion indices were used to reliably identify the abnormal temporal lobe. However, in the absence of ictal EEG recordings, these studies assumed that a structural lesion and interictal EEG could accurately localize the SOZ. Furthermore, these studies did not provide whole brain coverage and therefore extratemporal perfusion changes were not studied.

Like every physiological parameter, CBF dynamics have a normal range and variations that could be unique to individuals and influenced by structural abnormalities, SOZ location, and seizure frequency amongst other variables. Most ASL studies compared CBF maps to controls and calculated statistical differences. We subtracted patients’ postictal CBF maps from their baseline CBF maps and provided a unique measure of hypoperfusion seen in the postictal state that circumvents these limitations.

Lastly, interictal and postictal CBF maps provided information extra to sASL in only one patient (Patient 7), who had extremely frequent interictal discharges and seizures.

Timing of arterial spin labelling

The timing of ASL data acquisition with respect to seizure occurrence is important. This situation is complicated by the fact that there exists no clear definition of the postictal state, with studies using variable timelines, from 5 min to 60 days after a seizure. However, a recent animal study that systematically evaluated postictal hypoperfusion and hypoxia showed that these changes could be consistently seen at the SOZ for ∼60 min. Given these results, our aim was to obtain postictal ASL studies within 60 min of seizure termination.

The fact that human studies assessing CBF in relation to seizures yielded conflicting results by reporting both postictal hypoperfusion and hyperperfusion may be a result of the different time points at which ASL scanning was performed. In addition, previous ASL studies were retrospective and the precise timing of seizure termination was not recorded. We verified seizures by VEEG, so the time to ASL imaging was accurately measured.

No study has reported on hypoperfusion on ASL in the immediate postictal period (e.g. <60 min). Most studies focused on the interictal period, except two (Pizzini et al., 2013; Kim et al., 2016), where most patients underwent ASL 1–10 h following a seizure. Interestingly, one study showed hypoperfusion in 81% of patients from several hours to 60 days after seizures (Kim et al., 2016). However, the other study showed hyperperfusion 1–5 h after seizures and hypoperfusion interictally (Pizzini et al., 2013). None of these patients underwent continuous VEEG immediately prior to the ASL study, so the undetected seizures may have confounded these studies. Thus, it is important to perform postictal ASL studies prospectively, within a predetermined time window, and under continuous VEEG monitoring.

Hypoperfusion on arterial spin labelling

Previous ASL studies showed that CBF changes (hypo- or hyperperfusion) co-localize to the SOZ in 60–100% of patients and 15–21% showed no CBF changes (see ‘Introduction’ section). In our study, 71.4% showed significant postictal hypoperfusion and 28.6% did not. Of the latter patients, one had a late ASL scan (Patient 19) at 2 h, possibly explaining the lack of hypoperfusion. Another patient (Patient 9) had a suspected SOZ in the mesial frontal regions and abundant epileptiform discharges at baseline. The interictal and postictal ASL studies both showed hypoperfusion in the left supplementary motor area that was not present on sASL. This is the only patient in whom interictal or postictal ASL alone provided more information than sASL, probably due to the patient’s very active interictal discharges. The remaining four patients had unclear SOZ localization on scalp VEEG or very complex epileptogenic networks on subsequent intracranial VEEG, which may suggest the presence of widespread epileptic networks.

We observed different patterns of hypoperfusion in the 15 patients who exhibited significant postictal hypoperfusion (Table 3). This variability can be potentially explained. All patients who showed focal and lobar hypoperfusion were scanned within 90 min of seizure termination. Patient 21, who showed regional changes, was scanned after 90 min and experienced a secondarily generalized seizure. The three patients (Patients 5, 7 and 20) who showed truly multifocal hypoperfusion (with no focality) had complex network epilepsies. One had an extensive bilateral mesial frontal network based on intracranial EEG (Patient 7). One had a bitemporal-frontal network and extensive bilateral subependymal heterotopias (Patient 5). One had unclear, late ictal EEG changes in the right frontal and left temporal region independently (Patient 20).

Concordance of arterial spin labelling to the seizure onset zone

Subtraction ASL provided good localizing information in 12/15 cases where hypoperfusion was documented (Table 3). The best concordance was seen in temporal lobe epilepsy, where 8/11 patients showed focal and regional hypoperfusion concordant to the SOZ. Patients with temporal lobe epilepsy showed a lobar pattern of hypoperfusion, with more prominent hypoperfusion over the lateral temporal lobe and relative sparing of the mesial temporal lobe. This is similar to a previous study that showed lateral temporal hypoperfusion occurring 3–15 min following mesial temporal lobe seizures (Newton et al., 1992).

Subtraction ASL data also showed concordance with the SOZ in the five patients who subsequently underwent intracranial EEG. Of the three patients with a discrete SOZ, ASL was concordant in two (Patients 11 and 14) and showed no change in one (Patient 3). Two patients with complex bilateral epileptogenic networks (Patients 5 and 7) showed multifocal changes. ASL was also concordant to the presumed epileptogenic zone in the two patients that underwent epilepsy surgery.

Arterial spin labelling versus other neuroimaging data

In our study, ASL offered better localization than MRI in 52% cases. ASL offered similar or better localization compared to ictal SPECT (when performed) in 60% (9/15) of cases, across temporal and extratemporal cases. For the remaining six patients, sASL was inferior to ictal SPECT, possibly for the following reasons: unclear EEG onset or complex epileptogenic network (Patients 2 and 5), active interictal baseline (Patients 7 and 9), late postictal ASL scan (Patient 19), and unknown reasons (Patient 3). The yield of ictal SPECT has varied in studies from 66% to 97% based on temporal or extratemporal localization (Weil, 2001; Lee et al., 2008; Zaknun et al., 2008). When compared to PET, ASL offered similar or better localization in 71% cases. Studies have shown that the yield of PET varied from 30–60% based on the location of the SOZ (Ryvlin et al., 1998) and increases to 80% in temporal lobe epilepsy (Won et al., 1999). Thus, ASL may offer extra SOZ localizing information compared to PET and SPECT and it does not require ionizing radiation. ASL also has a greater spatial resolution, lower cost, and greater accessibility.

Hyperperfusion on arterial spin labelling

We analysed the ASL data to look for hyperperfusion. Only two patients showed significant postictal hyperperfusion. In one patient with a left temporal SOZ (Patient 3), hyperperfusion was seen in both frontal lobes, maximally over the left. In the other case (Patient 14), with a left frontal SOZ, hyperperfusion was seen in the right frontotemporal region. Thus, in both cases, hyperperfusion did not correctly localize the SOZ. Previous reports of hyperperfusion were single case reports or case series studies that produced inconsistent results in a variety of settings, including status epilepticus, post-stroke, and single seizures (Pizzini et al., 2008; Altrichter et al., 2009; Nguyen et al., 2010; Miyaji et al., 2014).

Confounding factors

Secondary generalized tonic-clonic seizures

Studies during electroconvulsive therapy demonstrated hyperperfusion in the thalamus and hypoperfusion in the anterior cingulate as well as dorsolateral and medial frontal lobes (Takano et al., 2007). In our study, all six patients except one (Patient 21) who had secondarily generalized tonic-clonic seizures exhibited widespread multifocal hypoperfusion or no sASL changes, possibly reflecting recruitment of bilateral networks.

Timing of arterial spin labelling

Based on recent animal work, it has been shown that the most severe hypoxia following a seizure occurs between 20–60 min (Farrell et al., 2016). This suggested that the timing of postictal scans is an important consideration. Indeed, most patients who had late postictal ASL scans (>60 min) showed no significant hypoperfusion.

Interictal baseline

Although not a factor in all cases, an active interictal baseline EEG needs to be considered. We hypothesize that frequent interictal epileptiform discharges could possibly lead to mild CBF reductions as documented in recent studies (Storti et al., 2014; Boscolo Galazzo et al., 2015). Thus, comparing such a patient’s baseline CBF to postictal CBF may not yield significant CBF differences.

Other factors

Other factors such as medication changes while undergoing VEEG, circadian rhythms, differences in diet and caffeine consumption may influence global CBF. However, sASL maps showed localized CBF changes specifically related to the SOZ, suggesting no significant influence of these factors.

Limitations

Although we showed good concordance of postictal hypoperfusion to the presumed SOZ, our results could not be validated by post-surgical outcome as only two patients underwent surgery to date. Also, since EEG electrodes were removed ∼20 min prior to each ASL scan, there is a small possibility of missing electrographic seizures during this time and in the scanner. As we included consecutive patients undergoing presurgical investigations, our patient population was heterogeneous and this may have contributed to some variability in our results. We have justified the use of our postictal hypoperfusion CBF threshold based on previous animal work and on the estimation of the standard deviation of our sASL maps, but recognize that other methods of analysing postictal CBF data exist. Indeed, our method may be optimal only when analysing CBF ASL data (e.g. not in CBF CT data) and be specific to our pseudo-continuous arterial spin labelling sequence. We were able to demonstrate significant postictal CBF changes despite selecting a rather long post-labelling delay of 2525 ms. Use of a more optimal post-labelling delay should further improve the sensitivity of this approach. Larger patient numbers and studying specific subgroups (e.g. temporal lobe epilepsy) may lead to more consistent results. Another limitation is the observation that patients with seizures with unclear ictal patterns did not have localized hypoperfusion. The reasons for this are not clear.

Rapid access to MRI scanners poses a challenge for obtaining postictal ASL scans. Because the entire postictal ASL protocol takes only 6.5 min to complete, our centre agreed to accommodate these scans in-between previously scheduled scans, with minimal impact on the scanning schedule. In addition, ASL scans can only be performed in our centres during regular hours of operation.

Conclusions

Subtraction ASL is a safe, feasible and cost-effective imaging modality that can be used in the postictal period to assist with the localization of the SOZ. Provided that the study is performed within 60 min of seizure termination and the interictal baseline is not active, hypoperfusion is seen in up to 80% of patients. Although it is not always specific, ASL appears to be a sensitive modality, with no false localization when compared to the SOZ. The best concordance is achieved for patients with a discrete SOZ and not with patients with widespread epileptogenic networks. Better concordance was also seen in temporal and temporal-plus cases. Postictal ASL also compared favourably with ictal SPECT and PET. Thus, sASL may be an additional tool for identifying the SOZ. Once more studies are able to reproduce our findings, sASL might eventually be considered as a replacement for ictal SPECT in appropriately selected patients.

Glossary

Abbreviations

ASL

arterial spin labelling

CBF

cerebral blood flow

sASL

subtraction ASL

SOZ

seizure onset zone

SPECT

single photon emission computerized tomography

VEEG

video-EEG