Neuroimaging techniques such as MRI, FDG-PET, SPECT, and magnetoencephalography (MEG) are key tools in the presurgical evaluation of pediatric patients with focal epilepsy.1 Localizing the epileptogenic zone (EZ) remains challenging, particularly for poorly defined cases (PDCs).2 The presurgical workup may fail to delineate the EZ, despite evidence suggesting a focal onset. Seizure outcome is worse for these patients. The current procedures for presurgical evaluation are relatively invasive for children as they involve repeated exposure to radiation through CT-based imaging, as well as radioactive tracers, contrast agents, and/or intravenous injections. In addition, PDCs often require invasive intracranial recordings. Therefore, it would be beneficial to find less invasive methods to localize the EZ.
Arterial spin labeling (ASL) is a completely noninvasive MRI perfusion method that uses magnetically labeled blood as an endogenous tracer to produce quantitative cerebral blood flow (CBF) maps. The foundation for using perfusion-based methods is that the EZ is expected to express abnormal function and therefore abnormal perfusion. Several papers have described ASL perfusion abnormalities correlated with the EZ in focal epilepsy, including studies in adults with temporal lobe epilepsy3–8 and a small number of recent retrospective studies in children.9 However, the ASL literature on pediatric patients remains sparse, particularly regarding PDCs of focal epilepsy in children, and several questions remain. In this study, we prospectively assessed the utility of ASL perfusion in the localization of the EZ in children with MRI-positive and MRI-negative poorly defined focal epilepsy.
Methods
Patients
This prospective study received full approval by the McGill University Health Centre Research Institute Ethics Board, and all involved patients and/or parents/guardians signed an informed consent form. Twenty-five consecutive children with PDCs of focal epilepsy referred for presurgical evaluation from January 1, 2015, to December 31, 2017, were included. These PDCs had negative or positive MRI findings either with subtle ill-defined MRI signal abnormalities or with apparent lesions thought to extend beyond what is seen on MRI, like focal cortical dysplasia (FCD). Five patients with prior resections but residual uncontrolled seizures were also included. Well-defined lesional cases with visible borders (i.e., cavernous malformations, tumors) were excluded.
Each patient underwent an exhaustive presurgical evaluation (history, semiology discussion, neuropsychology assessment, video-EEG telemetry, structural brain 3T-MRI, FDG-PET, and ictal/interictal SPECT, as well as MEG and intracranial recording in some cases) and each case was presented in a multidisciplinary seizure conference. Resective surgery was recommended if a consensus could be reached regarding a strong EZ hypothesis. For resective surgery cases, only patients with at least 18 months of postsurgical follow-up were included. Patients who underwent surgery were prospectively followed for long-term seizure outcome.
MRI and Postprocessing
Both 3D T1-weighted and pseudocontinuous ASL MRI with a post-label delay of 1500 msec10 were performed on a clinical 3T system (Achieva X, Philips Healthcare) as previously described.11,12 Cortical segmentation and anatomical labeling were obtained from individual T1-weighted MR images using FreeSurfer (version 5.3).13 The average CBF was estimated according to the formula recommended by the International Society for Magnetic Resonance in Medicine perfusion study group.10 The ASL CBF map was then coregistered with the T1-weighted MRI14 and the anatomical location of a perfusion abnormality was determined by referencing the T1-weighted and/or the Desikan-Killiany atlas as implemented in FreeSurfer.15
We divided all cases into MRI-positive and MRI-negative based on the radiological review of 3T structural MRI. An MRI-positive case was defined as a patient presenting a focal signal abnormality reported unequivocally by a neuroradiologist on two consecutive distinct 3T MRI examinations without the use of uncertain terms. An MRI-negative case was defined as a patient presenting no observed signal abnormality as reported by a neuroradiologist on two consecutive 3T MRI examinations.
Surgical and Nonsurgical EZ Location Determination
The EZ location was hypothesized based on concordance between the majority of presurgical tests and confirmed postoperatively if good postsurgical seizure outcome was achieved with ≥ 1-year follow-up and if the resected tissue was histopathologically abnormal. ASL perfusion results were evaluated with respect to these surgically proven EZ cases. We designated true-positive cases as those with perfusion abnormalities found within the surgically proven EZ; false-positive cases as those with perfusion abnormalities outside the surgically proven EZ; true-negative cases as those without perfusion abnormalities where no EZ was surgically proven; and false-negative cases as those without perfusion abnormalities when an EZ was surgically proven. Sensitivity and specificity, as well as positive and negative predictive values, were calculated. Resection true-positive cases were cases in which patients had successfully undergone surgery with good (Engel class I/II) postsurgical outcome and positive pathology. Extraresection false-positive cases were parts of the brain outside the resection area that had ASL perfusion abnormalities found in them.
Assessment of ASL Perfusion Abnormalities
ASL CBF maps were visually inspected prospectively in all axial planes for obvious qualitative perfusion abnormalities. The person performing this visual inspection (P.T.) was blinded to any hypotheses regarding the EZ. A suspected perfusion abnormality had to be seen on ≥ 2 consecutive slices to be considered positive.
For retrospective quantitative analysis of the ASL perfusion maps, we first created asymmetry index (AI) maps for the ipsilateral hemisphere of each patient, which were created using the following formula to calculate AI in a voxel-wise manner: AI (%) = 100 × [(mCBFipsi − mCBFcontra)/(mCBFipsi + mCBFcontra)], where mCBF is the mean CBF (ml/100 g/min), ipsi is ipsilateral to the suspected perfusion abnormality, and contra is contralateral to the suspected perfusion abnormality. Positive and negative AI values indicate relative hyperperfusion and hypoperfusion, respectively. We only displayed the AI map for the hemisphere thought to contain the EZ based on the hypothesis generated from the rest of the presurgical evaluation.
To determine which AI clusters were significant, we created voxel-wise z-score maps showing only significant z-score values for the ipsilateral hemisphere. We first determined the mean and standard deviation (SD) of AI values in the gray matter of the ipsilateral hemisphere (hemi), and we then calculated a z-score for each voxel in the hemisphere using the following formula: z-score = [AI – mean(hemi)]/SD(hemi). We used the cases in which patients had successfully undergone surgery with good (Engel class I/II) postsurgical outcome and positive pathology (resection true-positive cases) to optimize the z-score cluster significance threshold by finding the z-score and cluster size giving the greatest difference between resection true-positive cases and extraresection false-positive cases. We linearly registered the postresection structural MRI to the preresection MRI and manually delineated the resection volume, which was reviewed for accuracy by the attending neurosurgeon (R.W.R.D.). The difference between the average proportion of true-positive voxels within the resection to the total number of resection voxels and the average proportion of false-positive voxels within the extraresection area to the total number of extraresection voxels was plotted iteratively across a range of z-scores from 0 to 3 and across a range of cluster size cutoffs. The z-score and cluster cutoff percentage at which the maximal difference between the two curves occurred were determined to be 1.5 and 5%, respectively (Fig. 1). These values were then used to generate threshold z-score maps of all subjects, including patients who did not undergo surgery, and objectively determine significant AI clusters that could potentially be used to localize the EZ.
Optimization analysis of the z-score. The difference between the average true-positive rate in the resection and the average false-positive rate in the extraresection space in the ipsilateral space is plotted iteratively across a range of z-scores between 0 and 3 and across a range of cluster size cutoffs. The z-score and cluster cutoff percentage where the maximal difference between the two curves (z-score = 1.5 and cluster cutoff = 5% largest clusters) was determined (red arrow), and these values for z-score and cluster size cutoff were used as the thresholds here and in Fig. 2. Pos = positive.
Concordance of perfusion abnormalities with scalp EEG, structural 3T-MRI, FDG-PET, and SPECT results, as well as MEG findings, intracranial recording results, and the surgically proven EZ if available, was evaluated qualitatively: perfusion abnormalities were labeled “concordant” if there was complete or partial overlap with the other imaging modality, and “discordant” if there was no overlap. Concordances of ASL results with structural 3T-MRI and MEG were assessed at the scale of gyrus/gyri, which is accessible to these modalities.16 Comparisons of ASL results with FDG-PET and ictal/interictal SPECT findings, which both have relatively poor spatial resolution, were obtained at the scale of “subregions” within lobes. For comparisons of ASL with scalp EEG, concordance was determined at the lobar level only.
Microvessel Immunohistochemistry
Microvessel density analysis was performed on all surgical specimens. Portions deemed histologically normal by the neuropathologist were used as an internal negative control. The specimens were fixed in 10% neutral buffered formalin and processed into paraffin. Sections were immunostained with a monoclonal mouse antihuman CD34 antibody, slides were digitized on an Aperio ImageScope scanner (Leica Microsystems), and five 1-mm2 areas were manually drawn within the gray matter of each specimen. Any CD34+ rounded, oval, or tubular structure clearly separated from neurons and glial cells was counted as one microvessel. The mean number of microvessels/mm2 per specimen was calculated, and specimen group means were compared to the internal negative control group means using IBM SPSS Statistics for Windows software (version 23.0, 2015; IBM Corp.).
Results
We evaluated 25 consecutive PDCs of pediatric focal epilepsy (20 MRI-positive, 5 MRI-negative; 12 females; mean age 10.3 ± 4.6 years, age range 2–18 years, mean age at seizure onset 5.3 ± 3.81 years) for perfusion abnormalities and compared these to the other presurgical evaluation tests and the final surgically proven EZ for the operated patients. All 10 operated ASL-positive patients had good postsurgical seizure outcomes, and 7 of these patients were completely seizure free at the time of this report. The mean follow-up for all patients was 39.3 ± 11.5 months (range 18–54 months); however, 1 patient died due to unrelated complications of pancreatitis 18 months after his epilepsy surgery, and 18 months was used for his follow-up time. Table 1 summarizes age and presurgical test findings in all 25 patients, and Table 2 summarizes surgical results and seizure outcome.
Clinical, neurophysiological, and neuroimaging data for all patients
| Pt No. | Age at Seizure Onset (yrs) | Age at Eval (yrs) | Semiology Localization | Scalp EEG | MRI Finding | MRI Localization | ASL Qual | ASL Quant | PET | SPECT | MEG | icEEG |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | 1 | 13 | Nonloc | Multifoc | Multiple tubers | Multifoc | Lt mid T | Pos | Rt P, lt T | Rt T, lt F | Nonloc | SEEG multifoc |
| 2 | 10 | 12 | Rt T | Rt pst Q | Rt PO gliosis | Rt pst Q | Rt pst Q | Pos | Rt pst Q | Rt pst Q | Nonloc | — |
| 3 | 9 | 12 | Lt T | Lt T | Lt pst T sig ab | Lt pst T | Lt pst T | Pos | Lt pst T | Lt pst T | Lt pst T | ECoG lt pst T |
| 4 | 2 | 13 | Rt T | Rt F | Subtle rt OF sig ab | Rt OF | Rt OF | Pos | Rt OF | Rt OF | Nonloc | SEEG rt OF |
| 5 | 1 | 10 | Rt T | Rt T | Rt mes T sclerosis | Rt mes T | Rt mes T | Pos | Rt T | Rt T | Nonloc | — |
| 6 | 11 | 15 | Rt F | Rt H | Rt FP sig ab & atrophy | Rt pst P | Rt pst P | Pos | Rt P | — | Rt pst T | — |
| 7 | 1 | 2 | Lt F | Lt F | Lt inf C sig ab, likely FCD | Lt pst C | Lt pst C | Pos | Lt F | Lt F | — | ECoG lt F |
| 8 | 3 | 5 | Nonloc | Multifoc | Bilat PeV heterotopias | PeV | PeV | Pos | — | Nonloc | Nonloc | — |
| 9 | 3 | 3 | Lt F | Lt F | Subtle lt SMA sig ab | Lt sup F | L sup F | Pos | Lt F | Lt F | — | ECoG lt F |
| 10 | 9 | 13 | Nonloc | Multifoc | Subtle rt ant-ins sig ab | Rt ins | Neg | Pos | Nonloc | Rt F | Nonloc | Grids multifoc |
| 11 | 12 | 14 | Rt F | Rt F | Subtle rt C sig ab | Rt C | Rt C | Pos | Rt F | Rt F | Rt C | ECoG rt C |
| 12 | 5 | 5 | Lt T | Lt T | Subtle lt pst T sig ab | Lt sup T | Lt sup T | Pos | Lt T | Lt T | Lt sup T | ECoG C T |
| 13 | 2 | 13 | Rt P, rt T | Lt F, lt P, lt T | Neg | Nonloc | Neg | Neg | Lt F | Lt F, lt T | Nonloc | SEEG nonloc |
| 14 | 11 | 18 | Lt F | Lt F | Lt F mixed ischemia hemorrhagic clastic lesion | Lt F | Lt F | Pos | Lt F | Lt F | Lt F | — |
| 15 | 5 | 6 | Rt F | Rt F | Rt F basal cortical sig ab | Rt OF | Neg | Pos | Rt OF | Nonloc | Nonloc | — |
| 16 | 9 | 16 | Lt F, lt ins | Nonloc | Neg | Nonloc | Neg | Neg | Lt F | — | Nonloc | — |
| 17 | 0.5 | 7 | Nonloc | Rt F | Rt FP gliosis | Rt FP | Neg | NA | — | — | — | — |
| 18 | 9 | 16 | Lt F | Nonloc | Neg | Nonloc | Neg | Pos | Nonloc | Lt T | — | — |
| 19 | 8 | 11 | Nonloc | Nonloc | Neg | Nonloc | Neg | Pos | Nonloc | — | — | — |
| 20 | 5 | 5 | Nonloc | Rt pst Q | Rt pst Q sig abs | Rt pst Q | Rt pst Q | Pos | Rt pst Q | Rt pst Q | Rt pst Q | — |
| 21 | 3 | 6 | Nonloc | Nonloc | Subtle rt ant T sig ab | Rt ant T | Rt ant T | Neg | Rt T | Nonloc | — | — |
| 22 | 2 | 9 | F | Lt F | Neg | Nonloc | Neg | Pos | Lt F | Lt F | Lt pre-F | SEEG lt SMA |
| 23 | 3 | 14 | Lt F, lt ins | Lt F | Lt inf F sig ab | Lt inf F | Lt inf F | Pos | Lt inf F | — | Nonloc | Grids lat pre-F |
| 24 | 7 | 10 | Rt H | Rt T, lt T | Bilat mes T sig abs | Rt mes T, rt T | Rt H | Pos | Rt T | Rt ant T | Rt pst T | SEEG rt T |
| 25 | 1 | 5 | Nonloc | Rt O | Multiple bilat tubers | Multifoc | Rt O | Pos | Rt O | — | — | — |
Ant = anterior; C = central; eval = evaluation; F = frontal; FP = frontoparietal; grids = subdural grids; H = hemisphere; icEEG = ictal EEG; inf = inferior; ins = insula; mes = mesial; multifoc = multifocal; NA = not applicable; neg = negative; nonloc = nonlocalizing; O = occipital; OF = orbitofrontal; P = parietal; PeV = periventricular; PO = parietooccipital; pos = positive; pre-F = prefrontal; pst = posterior; pt = patient; Q = quadrant; qual = qualitative; quant = quantitative; sig ab = signal abnormality; sup = superior; T = temporal; — = not performed.
Gray shading indicates an EZ localizing finding addressed in the study.
Surgical data for all patients
| Pt No. | Op | Pathology | Op Outcome (Engel class) | Follow-Up (mos) |
|---|---|---|---|---|
| 1 | — | — | — | — |
| 2 | — | — | — | — |
| 3 | Lt pst T | MOGHE | Id | 45 |
| 4 | Rt OF | FCD IIa/b | Ia | 42 |
| 5 | Rt ant T | HS | IIa | 50 |
| 6 | — | — | — | — |
| 7 | Lt inf C | FCD IIb | Ia | 54 |
| 8 | — | — | — | — |
| 9 | Lt sup F | FCD IIb | Ia | 46 |
| 10 | — | — | — | — |
| 11 | Rt sup C | FCD IIa | Ia | 41 |
| 12 | Lt pst T | FCD IIb | Ia | 39 |
| 13 | — | — | — | — |
| 14 | Lt F | FCD IIId | Ia | 36 |
| 15 | — | — | — | — |
| 16 | — | — | — | — |
| 17 | — | — | — | — |
| 18 | — | — | — | — |
| 19 | — | — | — | — |
| 20 | Rt pst Q | FCD Ia/b | Ia | 18* |
| 21 | — | — | — | — |
| 22 | Lt SMA | FCD IIa | Ia | 19 |
| 23 | — | — | — | — |
| 24 | Rt T | HS | IIb | 22 |
| 25 | Rt O | — | — | — |
HS = hippocampal sclerosis; — = not performed.
Patient died 18 months after surgery due to complications of acute pancreatitis, which was unrelated to his epilepsy surgery.
Structural MRI Findings
In 5 of the 25 patients the MR images were completely negative. Of the 20 MRI-positive cases, 13 showed only subtle FLAIR, T2, or T1 signal abnormalities which could not be diagnosed further on MRI. The other 7 cases included 2 cases of suspected gliosis, 2 cases of tuberous sclerosis, 1 case of an apparent ischemic-hemorrhagic lesion, 1 case of bilateral periventricular heterotopias, and only 1 case in which FCD was strongly suspected on MRI.
ASL Findings
By qualitative visual inspection, perfusion abnormalities were found in 17/25 cases (68.0%) (Fig. 2). For the MRI-positive PDCs, a perfusion abnormality was identified and was concordant with the MRI signal abnormality in 17/20 patients (85.0%), appearing as hypoperfusion in 14 patients, hyperperfusion in 2 patients, and misplaced normoperfusion (i.e., periventricular heterotopias) in 1 patient. No perfusion abnormalities were found by qualitative visual inspection in the MRI-negative cases. Using an objective quantitative analysis, based on our optimization of z-score significance thresholds from operated cases (Fig. 1), we confirmed the perfusion abnormalities found on qualitative visual inspection and found 4 additional ASL-positive cases, including 3 of the 5 MRI-negative cases (Fig. 2). Overall, our quantitative method found perfusion abnormalities in 21/24 cases (87.5%).
ASL, AI, and z-score results for all patients. Numbers indicate the patient number. White arrows indicate perfusion abnormality and presumed EZ. Asymmetry maps show decreases (blue) or increases (red) in perfusion compared to the contralateral side, and z-score maps show regions of significant increases or decreases in perfusion relative to the contralateral hemisphere. The z-scores were thresholded between ± 1.5 and ± 4.92, and significant clusters were thresholded to include only the top 5% largest clusters. Patients 10, 13, 15, 16, 17, 18, 19, and 21 had nonlocalizing ASL results (gray numbering). N/A = not applicable.
Concordance With Other Presurgical Tests and Surgically Proven EZ
Perfusion abnormalities were concordant in 17/20 cases (85%) with structural 3T MRI findings, 15/20 cases (75%) with FDG-PET findings, 12/18 cases (66.7%) with scalp EEG findings, 10/16 cases (62.5%) with SPECT findings, 6/8 cases (75%) with MEG findings, and 2/5 cases (40%) with localizing intracranial recording results. In 11/25 cases (44%) overall, a hypothesis about the potential EZ was developed that was strong enough to support a recommendation of focal surgery. In all 11 of these cases, the EZ was confirmed by at least 1 year of good postsurgical seizure outcome and positive histopathology results. Qualitative perfusion abnormalities were concordant with the final surgically proven EZ in 10/11 cases (10 true positives, 1 false negative; sensitivity 90.9%). Figure 3 shows the presurgical structural MRI for these 10 patients with resection volume overlaid on top of z-score cluster maps. The specificity of ASL for finding the surgically proven EZ was 50%; the positive predictive value was 58.7%, while the negative predictive value was 87.5%. Retrospective quantitative analysis found 1 true positive that was missed by qualitative analysis and 3 additional false positives (i.e., sensitivity 100%, specificity 23%).
Engel class I/II patient z-score results. Numbers indicate the patient number. The most recent resected area is indicated by the white shading overlying the MRI, and z-score maps show regions of significant increases or decreases in perfusion relative to the contralateral hemisphere. The z-scores were thresholded between ± 1.5 and ± 4.92, and clusters were thresholded to include the top 5% largest clusters based on the threshold optimization analysis.
Patients 1, 8, 10, and 25 were considered not to be good surgical candidates because they appeared to have multiple potential EZs. For the remaining nonoperated ASL-positive patients (false positives; patients 2, 6, 15, 18, 19, and 23), the ASL finding was not concordant with the majority of other presurgical evaluation tests and/or a strong hypothesis could not be generated about the EZ. One of the ASL-negative patients underwent surgery after stereoelectroencephalography (SEEG) localized the EZ and has been seizure free for 22 months. This was the single false negative on qualitative analysis, which was detected by retrospective quantitative analysis.
Surgical Details
In all ASL-positive surgical cases, the ASL results contributed to the presurgical hypothesis but were not used to plan the resections. For most surgeries, the resection borders were based on a combination of the structural MRI and FDG-PET findings. For patient 11, the resection was based on a subtle MRI signal abnormality and was limited by the adjacent precentral gyrus. For patient 5, a standard anterior temporal resection was performed. For patients 4 and 24, the surgical plan was based on the SEEG-determined seizure onset zone and areas of earliest electrographic spread. Electrocorticography (ECoG) was performed in all surgeries except for patients 4, 5, 14, and 24. All surgeries, except for patient 14, were performed using intraoperative MRI. The extent of resection of the MRI-visible lesion was considered complete (per a certified pediatric neuroradiologist) in all patients, except for patients 3 and 11, in whom the surgery was limited by adjacent eloquent cortex, and patient 5, who underwent a standard anterior temporal resection irrespective of posterior hippocampal signal abnormalities. Patient 11, a 15-year-old female, suffered approximately 18 months of left leg weakness due to resection adjacent to the right precentral gyrus and posterior supplementary motor area (SMA). Patient 12, a 5-year-old female, had only subtle transient (< 2 weeks) word-finding difficulties after surgery in her dominant left posterior superior temporal gyrus. Patient 22, an 11-year-old female, had approximately 2 weeks of expected transient right arm and leg weakness and mild speech deficits after a dominant SMA resection.
Surgical Pathology
Surgical pathology was positive in all 11 operated cases. For the 10 ASL-positive cases, analysis of the resected tissue revealed 7 cases of FCD,17 1 case of mild malformation of cortical development with oligodendroglial hyperplasia (MOGHE),18 and 2 cases of hippocampal sclerosis. For the single ASL false-negative case, the patient was found to have FCD type IIa.
Microvessel Density
Twelve specimens were used for microvessel density analysis, including 6 from patients with FCD (5 with hypoperfused and 1 with hyperperfused ASL), 1 with MOGHE, 1 with hippocampal sclerosis, and 4 internal negative controls. The mean number of microvessels/mm2 for the 6 FCD specimens (70.8 ± 9.93/mm2) was nonsignificantly different from the number in the internal negative controls (67.8 ± 4.35/mm2). Five FCD specimens from hypoperfused ASL cases had on average 71.4 ± 10.99 microvessels/mm2, while the single hyperperfused ASL FCD case had 68.0 microvessels/mm2.
Discussion
Cases of pediatric focal epilepsy that are poorly defined on imaging have a worse prognosis and require multiple imaging tests.19 It would be advantageous to reduce the need for CT- and radioactive tracer–based imaging for these patients.20 In this study, we chose to study ASL perfusion only in PDCs for two reasons. First, epilepsy surgery is already quite successful for cases with well-defined lesions such as brain tumors and cavernous malformations. We previously reported 100% and 86% 2-year seizure freedom for well-defined cases depending on whether intraoperative MRI was used or not, respectively, versus 50% and 57% for PDCs.2 Second, ASL has already been established in its ability to detect brain tumors.21,22 For both reasons, we wanted to explore the use of ASL MRI in all types of PDCs from visible malformations, which can extend beyond what is seen on MRI, to the completely MRI-negative cases.
We found perfusion abnormalities in 17/25 cases (68%) total, specifically in 17/20 MRI-positive PDCs (85%), which were reasonably well concordant with the other presurgical evaluation modalities. Ten of these ASL-positive cases underwent resection of brain regions concordant with the ASL abnormality, and positive surgical pathology and good seizure outcome confirmed the presurgical hypothesis. The sensitivity and specificity of ASL to detect the EZ were 90.9% and 50.0%, respectively.
To our knowledge, this is the largest prospective study in children to compare ASL perfusion to all common presurgical electrophysiological and neuroimaging modalities, as well as surgical pathology and long-term postoperative seizure outcomes. The majority of published ASL studies in epilepsy have been conducted in adults. In a combined group of 16 adult and pediatric patients with mixed pathologies, Pendse et al. found ASL hypoperfusion in all patients and showed that the ASL results corresponded to EEG, MRI, and FDG-PET findings, but these authors did not report seizure outcomes.23 Sierra-Marcos et al. studied ASL in lesional and nonlesional adult neocortical epilepsy and found perfusion abnormalities in 15/25 patients (60%) as well as a very good concordance with FDG-PET and structural MRI.24 However, in their study only 4/15 ASL-positive patients underwent surgery, and 3 of these 4 patients became seizure free with a mean follow-up of 40.75 months.24 In a recent study by Lee et al., 24/36 MRI-negative children (66.7%) had perfusion abnormalities, and the authors reported 52.8% concordance with the hypothesized EZ, but no patients underwent surgery.25 Prior studies of ASL in pediatric focal epilepsy have consisted of case reports or small series of lesional cases. Blauwblomme et al. described ASL hypoperfusion in 9 children with FCD, 5 of whom underwent surgery.9 However, the authors did not comment on surgical outcomes, and the mean follow-up was less than 1 year.
Together, our results and those of Blauwblomme et al. suggest that ASL may be particularly useful for FCD in children.9 We included all children with PDCs, and 7/10 operated patients had FCD. To our knowledge, our study is also the first to compare the ability of ASL to localize the EZ in MRI-positive versus MRI-negative PDCs in children. Perfusion abnormalities were not observed in any of the MRI-negative cases on prospective qualitative analysis. A possible explanation for this finding is that the spatial resolution and signal-to-noise ratio of ASL limit the detection of very small/subtle perfusion abnormalities and/or some EZs may be insufficiently hypoperfused to allow detection.
Like SPECT, the majority of ASL studies in epilepsy describe interictal hypoperfusion and ictal hyperperfusion. However, Wintermark et al. reported ASL hyperperfusion in 2 very young FCD patients (21 days and 4 months of age) during the interictal period and found increased microvessel counts in surgical specimens.26 These authors speculated that seizures could be due to increased microvasculature/perfusion.26 Although the authors reported that same-day EEGs were negative and no clinical seizures were seen in the MRI, their exact electrophysiological status was uncertain because concomitant EEG was not performed and subtle seizures can be difficult to recognize in very young infants. Another recent report described ASL hyperperfusion in 3 neonates, but each had either status epilepticus or very frequent seizures within 2–4.5 hours of the ASL MRI, and again concomitant EEG was not performed. Blauwblomme et al. did not find hyperperfusion in any of the 9 children with FCD and found no difference in microvessel counts.9 Pendse et al. found only hypoperfusion in 16 mixed-age cases.23 Sierra-Marcos et al. found hyperperfusion in only 4/15 patients with ASL abnormalities; 2 of these 4 patients had very frequent seizures, and 2 had brain tumors.24 Boscolo Galazzo et al. found postictal hyperperfusion in only 2/12 adult patients.27 In our study, only 2/18 children showed ASL hyperperfusion, and, like Blauwblomme et al., we did not find significant differences in microvessel counts.9 Practically speaking, it does not matter whether ASL perfusion reveals hypoperfusion or hyperperfusion, because both can indicate abnormal function and a potential EZ.28
We found that perfusion abnormalities were concordant with the final surgically proven EZ in 10/11 cases. There was 1 false-negative case, a patient in whom the ASL MRI was negative but in whom SEEG eventually revealed a seizure generator in the left SMA, which was confirmed postsurgically by identifying FCD type IIa and postsurgical seizure freedom. Retrospective quantitative analysis revealed this surgically proven EZ, but we were biased to know exactly where to look. There were also 7/25 cases (28%) with false-positive results with qualitative analysis, 3 of which were disproven with intracranial recording. In the other 4 cases, the ASL perfusion abnormalities simply did not correspond well enough with other components of the presurgical evaluation data. This false-positive rate was similar to that with SPECT, but higher than that for other presurgical evaluation tests. Mastin et al. reported 24% and 43% false positives for interictal and postictal SPECT relative to the site of surgery, respectively, and only a 12% false-positive rate for interictal PET.29 In our experience, FDG-PET coregistered with MRI has been useful in highlighting occult pathology beyond the visible lesion. The utility of SPECT, on the other hand, is more variable as it is difficult to time radionucleotide injection with seizure onset. Both modalities can in certain circumstances reveal the EZ even in MRI-negative cases. However, as presented in the current study, ASL does not appear to be helpful in MRI-negative cases and can certainly not replace either PET or SPECT at this time.
In the present study, only 2/5 patients with positive localizing intracranial recording findings had concordant ASL results. These 5 cases requiring intracranial recording were the most difficult cases to diagnose and treat as they did not have obvious lone focal signal abnormalities. Since the current study shows that ASL is more useful in colocalizing more obvious MRI signal abnormalities in PDCs, it is not surprising that the lowest concordance rates in our study were those between ASL and intracranial recording.
Study Limitations
Study limitations include the fact that the epileptic state of each patient during ASL acquisition was unknown because EEG was not performed concurrently, and the patient’s last seizure before the ASL MRI was not recorded.30 Furthermore, a technical limitation of our ASL protocol is that the simple retrospective motion correction scheme used neglects the small difference in contrast between the control and label images, which can introduce a small bias in registration outcome.30 In addition, despite this being, to our knowledge, the largest study of its kind in children, the sample sizes are still relatively small, particularly for MRI-negative cases and specimen groups used for microvessel comparisons. Also, our microvessel counting/architecture results were not compared to those of a true-negative control, such as age-matched autopsy specimens, as these were not available. Another limitation of our study is the fact that 8/25 (32%) MR images were done with sedation in young or uncooperative patients (6 of whom were ≤ 5 years old), which raises the question as to not only the impact of anesthesia on our results, but also the risk-benefit ratio of additional anesthesia time. The ASL sequence itself adds only approximately 7 minutes to the scanning time. There is no evidence that general anesthesia significantly affects cerebral perfusion if blood pressure is controlled within a normal range during the MRI.31 Anesthesia, which should have a global effect, should not have had an impact on the findings of our study, which was performed to detect focal asymmetry in the brain. We feel the potential for ASL to add support to a subtle structural abnormality outweighs the risk of adding 7 minutes of anesthesia time to the treatment of patients who already require sedation for their MRI. An additional potential limitation of our study is that the FreeSurfer segmentation method can be suboptimal for very young children (i.e., < 3 years old).32 However, in our study the 1 patient younger than 3 years (a 2-year-old) had a normal head shape and no previous neurosurgical intervention. Each patient’s segmentation was manually reviewed and judged to be of sufficient quality for analysis, and thus this was likely an insignificant problem in our study.
Conclusions
In summary, our findings suggest that ASL perfusion is a useful functional imaging technique for MRI-positive PDCs in children. Future optimized quantitative methods may improve the diagnostic yield of this technique, such as outlier-detection algorithms to see where each patient’s ASL CBF data deviate from normative ASL data sets in the age-matched healthy population, but this possibility has yet to be explored. While a positive lesion on MRI is the most important predictor for seizure freedom, we feel the real value of ASL at this time is not for the obvious lesions like FCD type IIb, which are often very visible on MRI, and definitely not for the completely negative MRI cases. Instead, the real value of ASL at present is for those patients in whom a subtle signal abnormality is seen on MRI, which could be either a “red herring,” such as a technical artifact or a benign anatomical variation such as an enlarged perivascular space, or actually a “tip-of-the-iceberg” pathological entity that needs confirmation with another imaging method. ASL MRI appears to be an effective noninvasive and efficient method of confirming that such signal abnormalities are truly pathological, when colocalized perfusion changes are found. We continue to perform ASL MRI as part of the presurgical evaluation for PDCs of focal epilepsy. ASL has not replaced another imaging tool at this point, and we do not foresee it doing so in the near future, at least in its current state. We feel that any benefit the ASL MRI might provide in these most complex cases is worth the additional 7-minute workflow in the overall scheme of the patient’s presurgical evaluation. Thus, due to the relative convenience, noninvasive nature, and high confirmatory rate of ASL in MRI-positive PDCs, we recommend that ASL be performed as part of the presurgical evaluation MRI in all children with PDCs of focal epilepsy.
Acknowledgments
We thank Guillaume Gilbert, Philips Canada, for his technical assistance and Nassima Addour for her aid in preparing the manuscript. This study was funded by the Department of Pediatric Surgery, Montreal Children’s Hospital, and the Montreal Children’s Hospital Research Institute.
Disclosures
Dr. Gilbert reports being an employee of Philips Healthcare. Dr. Moreau reports receiving clinical or research support for the study described (includes equipment or material) from the Canada First Research Excellence Fund awarded to McGill University for the Healthy Brains, Healthy Lives initiative; the Fonds de Recherche du Québec-Santé; and the Foundation of Stars.
Author Contributions
Conception and design: Dudley, Wintermark, Baillet. Acquisition of data: Dudley, Lam, Tomaszewski, Moreau, Farmer, Atkinson, Saint-Martin, Bernhardt. Analysis and interpretation of data: Dudley, Lam, Tomaszewski, Moreau, Guiot, Albrecht, Saint-Martin, Bernhardt, Baillet. Drafting the article: Dudley, Lam, Tomaszewski, Guiot, Albrecht, Saint-Martin. Critically revising the article: Dudley, Lam, Tomaszewski, Gilbert, Bernhardt, Baillet. Reviewed submitted version of manuscript: all authors. Approved the final version of the manuscript on behalf of all authors: Dudley. Statistical analysis: Dudley. Administrative/technical/material support: Dudley, Gilbert, Bernhardt, Baillet. Study supervision: Dudley, Bernhardt, Baillet.
References
- 1↑
Duncan JS, Winston GP, Koepp MJ, Ourselin S. Brain imaging in the assessment for epilepsy surgery. Lancet Neurol. 2016;15(4):420–433.
- 2↑
Warsi NM, Lasry O, Farah A, et al. 3-T intraoperative MRI (iMRI) for pediatric epilepsy surgery. Childs Nerv Syst. 2016;32(12):2415–2422.
- 3
Eryurt B, Oner AY, Ucar M, et al. Presurgical evaluation of mesial temporal lobe epilepsy with multiple advanced MR techniques at 3T. J Neuroradiol. 2015;42(5):283–290.
- 4
Guo X, Xu S, Wang G, et al. Asymmetry of cerebral blood flow measured with three-dimensional pseudocontinuous arterial spin-labeling MR imaging in temporal lobe epilepsy with and without mesial temporal sclerosis. J Magn Reson Imaging. 2015;42(5):1386–1397.
- 5
Lim YM, Cho YW, Shamim S, et al. Usefulness of pulsed arterial spin labeling MR imaging in mesial temporal lobe epilepsy. Epilepsy Res. 2008;82(2-3):183–189.
- 6
Liu HL, Kochunov P, Hou J, et al. Perfusion-weighted imaging of interictal hypoperfusion in temporal lobe epilepsy using FAIR-HASTE: comparison with H2 15O PET measurements. Magn Reson Med. 2001;45(3):431–435.
- 7
Oner AY, Eryurt B, Ucar M, et al. pASL versus DSC perfusion MRI in lateralizing temporal lobe epilepsy. Acta Radiol. 2015;56(4):477–481.
- 8
Wolf RL, Alsop DC, Levy-Reis I, et al. Detection of mesial temporal lobe hypoperfusion in patients with temporal lobe epilepsy by use of arterial spin labeled perfusion MR imaging. AJNR Am J Neuroradiol. 2001;22(7):1334–1341
- 9↑
Blauwblomme T, Boddaert N, Chémaly N, et al. Arterial spin labeling MRI: a step forward in non-invasive delineation of focal cortical dysplasia in children. Epilepsy Res. 2014;108(10):1932–1939.
- 10↑
Alsop DC, Detre JA, Golay X, et al. Recommended implementation of arterial spin-labeled perfusion MRI for clinical applications: a consensus of the ISMRM perfusion study group and the European consortium for ASL in dementia. Magn Reson Med. 2015;73(1):102–116.
- 11↑
Boudes E, Gilbert G, Leppert IR, et al. Measurement of brain perfusion in newborns: pulsed arterial spin labeling (PASL) versus pseudo-continuous arterial spin labeling (pCASL). Neuroimage Clin. 2014;6:126–133.
- 12↑
Shaikh H, Lechpammer M, Jensen FE, et al. Increased brain perfusion persists over the first month of life in term asphyxiated newborns treated with hypothermia: does it reflect activated angiogenesis? Transl Stroke Res. 2015;6(3):224–233.
- 14↑
Jenkinson M, Bannister P, Brady M, Smith S. Improved optimization for the robust and accurate linear registration and motion correction of brain images. Neuroimage. 2002;17(2):825–841.
- 15↑
Desikan RS, Ségonne F, Fischl B, et al. An automated labeling system for subdividing the human cerebral cortex on MRI scans into gyral based regions of interest. Neuroimage. 2006;31(3):968–980.
- 16↑
Baillet S. Magnetoencephalography for brain electrophysiology and imaging. Nat Neurosci. 2017;20(3):327–339.
- 17↑
Blümcke I, Thom M, Aronica E, et al. The clinicopathologic spectrum of focal cortical dysplasias: a consensus classification proposed by an ad hoc Task Force of the ILAE Diagnostic Methods Commission. Epilepsia. 2011;52(1):158–174.
- 18↑
Schurr J, Coras R, Rössler K, et al. Mild malformation of cortical development with oligodendroglial hyperplasia in frontal lobe epilepsy: a new clinico-pathological entity. Brain Pathol. 2017;27(1):26–35.
- 19↑
Bulacio JC, Jehi L, Wong C, et al. Long-term seizure outcome after resective surgery in patients evaluated with intracranial electrodes. Epilepsia. 2012;53(10):1722–1730.
- 20↑
Pearce MS, Salotti JA, Little MP, et al. Radiation exposure from CT scans in childhood and subsequent risk of leukaemia and brain tumours: a retrospective cohort study. Lancet. 2012;380(9840):499–505.
- 21↑
Dangouloff-Ros V, Deroulers C, Foissac F, et al. Arterial spin labeling to predict brain tumor grading in children: correlations between histopathologic vascular density and perfusion MR imaging. Radiology. 2016;281(2):553–566.
- 22↑
Noguchi T, Yoshiura T, Hiwatashi A, et al. Perfusion imaging of brain tumors using arterial spin-labeling: correlation with histopathologic vascular density. AJNR Am J Neuroradiol. 2008;29(4):688–693.
- 23↑
Pendse N, Wissmeyer M, Altrichter S, et al. Interictal arterial spin-labeling MRI perfusion in intractable epilepsy. J Neuroradiol. 2010;37(1):60–63.
- 24↑
Sierra-Marcos A, Carreño M, Setoain X, et al. Accuracy of arterial spin labeling magnetic resonance imaging (MRI) perfusion in detecting the epileptogenic zone in patients with drug-resistant neocortical epilepsy: comparison with electrophysiological data, structural MRI, SISCOM and FDG-PET. Eur J Neurol. 2016;23(1):160–167.
- 25↑
Lee SM, Kwon S, Lee YJ. Diagnostic usefulness of arterial spin labeling in MR negative children with new onset seizures. Seizure. 2019;65:151–158.
- 26↑
Wintermark P, Lechpammer M, Warfield SK, et al. Perfusion imaging of focal cortical dysplasia using arterial spin labeling: correlation with histopathological vascular density. J Child Neurol. 2013;28(11):1474–1482.
- 27↑
Boscolo Galazzo I, Mattoli MV, Pizzini FB, et al. Cerebral metabolism and perfusion in MR-negative individuals with refractory focal epilepsy assessed by simultaneous acquisition of (18)F-FDG PET and arterial spin labeling. Neuroimage Clin. 2016;11:648–657.
- 28↑
Pollock JM, Tan H, Kraft RA, et al. Arterial spin-labeled MR perfusion imaging: clinical applications. Magn Reson Imaging Clin N Am. 2009;17(2):315–338.
- 29↑
Mastin ST, Drane WE, Gilmore RL, et al. Prospective localization of epileptogenic foci: comparison of PET and SPECT with site of surgery and clinical outcome. Radiology. 1996;199(2):375–380.
- 30↑
Gaxiola-Valdez I, Singh S, Perera T, et al. Seizure onset zone localization using postictal hypoperfusion detected by arterial spin labelling MRI. Brain. 2017;140(11):2895–2911.
- 31↑
Rhondali O, André C, Pouyau A, et al. Sevoflurane anesthesia and brain perfusion. Paediatr Anaesth. 2015;25(2):180–185.
- 32↑
Levman J, MacDonald P, Lim AR, et al. A pediatric structural MRI analysis of healthy brain development from newborns to young adults. Hum Brain Mapp. 2017;38(12):5931–5942.
