The Imaging of Localization Related Symptomatic Epilepsies: The Value of Arterial Spin Labelling Based Magnetic Resonance Perfusion

Symptomatic epilepsy is defined as epilepsy having gross anatomic or pathologic abnormalities secondary to anacquired or genetic cause (1). Its natural course and response to medical treatment are determined by its etiology. Approximately 25% of symptomatic epilepsy cases are refractory (2). In such cases, accurate localization of the epileptogenic zone (EZ) is one of the strongest prognostic factors since it allows complete surgical resection and effective seizure control (3). Currently, localization is achieved by a combination of seizure semiology and electroencephalography (EEG) along with structural magnetic resonance imaging (sMRI). In cases with negative sMRI or discordant findings, interictal 18-flurodeoxyglucose positron emission tomography (PET) and ictal single-photon emission computed tomography (SPECT) can be used for localization. However, these modalities involve ionizing radiation with limited resolution. Although invasive EEG is highly accurate, it has a limited sampling area (4). A more circumscribed localization with non-invasive imaging can help guide the placement of these electrodes (5).

Arterial spin labelling (ASL)-based magnetic resonance (MR) perfusion is a technique that has potential to be used as a non-invasive and non-irradiating supplementary method of localization. In this article, we review clinical insights and advantages gained from its use in the evaluation of epilepsy.

Arterial spin labelling uses an inversion pulse to tag flowing blood in the neck. Following a short post-labelling delay (PLD) to allow for intracranial transit, these ‘labelled’ spins are imaged during exchange into the brain parenchyma. Subtracting a paired set of ‘labelled’ and unlabelled ‘control’ images will yield parametric maps of absolute cerebral blood flow (CBF) that may be used for qualitative and quantitative evaluation. In effect, ASL uses blood as an endogenous contrast material, making it non-invasive and repeatable. Thus, it has significant advantages, particularly for pediatric population (6).

Scans are performed on a 3T GE Discovery 750w scanner (GE Healthcare, Milwaukee, USA) using a three dimensional-pseudocontinuous ASL sequenceprior to contrast administration per recommendations by the International Society for Magnetic Resonance in Medicine (5). Post-processing includes CBF quantification and fusion with relevant sMRI sequences.

Two important concepts are necessary to understand the aim and utility of ASL in epilepsy: the EZ and the principle of neurovascular coupling.

The Epileptogenic Zone

The part of the cortex from which seizures originate is referred to as the ictal onset zone (IOZ). Some lesions such as arteriovenous malformations (AVMs) and tumors have a perilesional EZ that does not necessarily overlap with the IOZ. It is capable of generating seizures by itself even after removing the inciting lesion. Spread of epileptic activity to adjacent areas will manifest as semiology of the seizure. This is designated as the symptomatogenic zone. A functional deficit zone (FDZ) corresponds to areas with altered interictal function while the irritative zone generates interictal spikes that are secondary to inhibitory neural networks (Fig. 1). EZ represents the minimum amount of tissue that must be surgically resected or disconnected. Its accurate delineation is the holy grail of presurgical evaluation since it impacts post-operative seizure freedom and neurocognitive outcomes (5, 7, 8, 9). Evidence has emerged that FDZ identified by neuropsychological evaluation or regions of PET hypometabolism may expand horizons of EZ, particularly in temporal lobe epilepsy (TLE), thus requiring additional resection (9, 10, 11). EZ is thus a theoretical construct that can only be estimated based on combined evidence from seizure semiology, electrophysiological tests, and imaging. Alterations in perfusion as reflected on ASL may occur not only in EZ, but also in FDZ and symptomatogenic zones. ASL has a reasonable overall accuracy of 70% in localizing the seizure focus, which is lower than the overall accuracy of 95% by electroclinical localization (12). Thus, interpretation of ASL in isolation is limited. Electroclinical evidence must be considered to reach a more accurate estimation of EZ location.

Fig. 1 Various zones around epileptogenic lesion. Successful surgical outcome is dependent on resection of EZ. EZ = epileptogenic zone Click for larger imageDownload as PowerPoint slide

We will describe the utility of ASL in the evaluation of five basic substrates for epilepsy: 1) mesial temporal sclerosis, 2) malformations of cortical development (MCDs), 3) neoplasms, 4) vascular malformations, and 5) infections, inflammations, and gliosis. In addition, we will describe the application of ASL in sMRI-negative epilepsy. Where studies on ASL specific to the entity are lacking, we will describe comparable modalities such as PET and dynamic susceptibility contrast (DSC) imaging.

Hippocampal sclerosis is the most common cause of intractable epilepsy (2). Approximately 20–30% of patients with MTLE have no sMRI identifiable abnormality (20). While visual inspection of ASL CBF maps may identify obvious areas of altered perfusion (Fig. 2), quantitative methods are needed for cases with ambiguous lateralization. Mean CBF values derived by regions of interest over the right (R) and left (L) mesial temporal regions may be used to calculate Asymmetry Index (AI) using the following formula: AI = 100 × (L − R) / (R − L), with negative values indicating left lateralization (21). However, caution is needed since normal subjects may show physiological left temporal lobe hyperperfusion. Seizure frequency and severity inversely are correlated with ipsilateral hippocampal CBF, but not with the magnitude of AI possibly due to secondary involvement of contralateral temporal lobe (22, 23). Since these changes occur in sMRI-negative cases as well, ASL can improve the overall sensitivity of MRI. It has shown a good concordance with both PET and electrical source imaging (24, 25, 26). Another important caveat is peri-ictal hippocampal changes in conditions such as status epilepticus that may occur regardless of the site of ictal onset. Interval imaging may be necessary to look for reversibility or evolution into sclerosis (27, 28).

Fig. 2 Hippocampal sclerosis. Peri-ictal imaging in 14-year-old female child who presented with new-onset hypomotor complex partial seizures showing swelling and T2 hyperintensity of right hippocampus (arrow, A) with hyperperfusion on ASL fusion images (arrow, B). Unlike this particular case, chronic hippocampal sclerosis with secondary atrophy commonly show hypoperfusion on ASL. ASL = arterial spin lebelling Click for larger imageDownload as PowerPoint slide

Malformations of cortical development comprise between 10–50% of pediatric and 4–25% of adult surgical epilepsy patients (29). Focal cortical dysplasias (FCDs) are probably the most common cause of intractable extratemporal epilepsy (30). On nuclear imaging, FCDs commonly show interictal hypometabolism and peri-ictal hypermetabolism. These areas typically appear larger than sMRI-visible lesion (Fig. 3) (31). While this has conventionally been attributed to altered synaptic/epileptic activity, Wintermark et al. (32) have suggested that these findings might be related to higher microvessel density measured by immunohistochemical staining of CD34 positive endothelial cells on surgical resection specimens. They also found significantly higher vascular density in the adjacent normal appearing cortex and hypothesized that interictally hypervascular FCDs might require larger resection margin (32). ASL is also a viable substitute for nuclear imaging to localize the offending tuber in tuberous sclerosis (33).

Fig. 3 Focal cortical dysplasia. Interictal imaging in 24-year-old female with occasional breakthrough seizures showing right frontal Type II FCD on FLAIR (arrow, A) with hypoperfusion on corresponding ASL fusion images (arrow, B). In another case of 11-year-old female child with refractory extratemporal complex partial seizures, peri-ictal imaging showing right parietal FCD on T1 SPGR (arrow, C) and focal hyperperfusion on ASL grey-scale source images (arrow, D) consistent with active seizure focus. FCD = focal cortical dysplasia, FLAIR = fluid attenuation inversion recovery, SPGR = spoiled gradient recalled acquisition Click for larger imageDownload as PowerPoint slide

Hemimegalencephaly is a prime example of pathology in which perfusion and metabolism may be uncoupled. Involved areas may show variable metabolic activities on PET with a paradoxically opposite perfusion pattern on SPECT that suggests a role of either vascular proliferative abnormalities or subclinical seizures (34). ASL has shown similar variability in perfusion of these lesions (Fig. 4) (35, 36).

Fig. 4 Hemimegalencephaly. Note subtle enlargement of left cerebral hemisphere on FLAIR (most obvious in regions between arrows in A) in 16-year-old male with global developmental delay and refractory seizures. Interictal ASL fusion images (B) showing areas of cortical hypoperfusion that allow more clear delineation of extent of involvement (between arrows in B). Click for larger imageDownload as PowerPoint slide

Polymicrogyric cortex shows variable patterns of metabolic activity compared to normal cortex on PET (37). Band heterotopia may show similar or higher metabolic activity compared to overlying cortex appearing as a ‘double cortex’ similar to structural imaging. In both these entities, altered perfusion might be related to synaptic activity of neurons regardless of whether it results in clinical seizures (38, 39). ASL shows perfusion patterns concordant with PET (Fig. 5). ASL also allows ruling out of contralateral perfusion anomalies related to microstructural or network alterations since these anomalies could potentially impact post-surgical outcomes (31, 40).

Fig. 5 Polymicrogyria. Left perisylvian polymicrogyria seen on T1 SPGR (arrows, A) in 25-year-old male with extratemporal complex partial seizures. Focal areas of hyperperfusion are seen on ASL fusion images of polymicrogyric cortex (arrows, B) compared to normal perfusion pattern of contralateral side (C) in this peri-ictal scan. Click for larger imageDownload as PowerPoint slide

Tumors associated with drug resistant epilepsy for a duration of usually two years or longer are termed long-term epilepsy associated tumors, with gangliogliomas and dysembryoplastic neuroepithelial tumors being the commonest (41). Early resection is preferred for both seizure control and tumor remission (42). The utility of ASL in predicting tumor grade has been well established (43). However, when the primary goal of surgery is seizure control, EZ, not just the tumor, should be resected. It has been shown that EEG and ictal SPECT are useful in delineating EZ around tumors (44). Peri-ictal ASL has similar potential. It may reduce the need for a separate study. Alterations of perfusion may also represent ictal activity propagation to a symptomatogenic zone (Fig. 6).

Fig. 6 Dysembryoplastic neuroepithelial tumor. T2-weighted images (A) showing dysembryoplastic neuroepithelial tumor involving right peri-Rolandic region in 24-year-old male. While lesion per se is hypoperfused, adjacent hyperperfusion involves lateral right sensorimotor cortex (arrow, B) on this peri-ictal scan despite sMRI showing lack of perilesional changes, concordant with presentation of simple partial seizures of left hand despite lesion was located in foot region. sMRI = structural magnetic resonance imaging Click for larger imageDownload as PowerPoint slide

Seizures in vascular malformations are caused by neuronal toxicity and gliosis secondary to hemosiderin deposition during overt or subclinical hemorrhage. Chronic venous congestion, impaired peri-nidal cerebrovascular reserve, and steal phenomena are additional factors in the epileptogenicity of AVMs (45).

Arterial spin labelling allows characterization of both vascular malformation and its secondary effects on the brain parenchyma. Venous signal on ASL is a sensitive marker of shunting in occult malformations (46). Using a single compartment model, semi-quantitiative estimation of shunt fraction can be calculated. Multi-PLD ASL allows estimation of transit times. It is useful for monitoring patients after shunt reduction interventions or radiosurgery (47). Peri-nidal hyperperfusion and steal phenomena can also be identified and quantified (Fig. 7). Areas of peri-ictal hyperperfusion distant from the AVM nidus have been found to be consistent with secondary EZs (48).

Fig. 7 Arteriovenous malformation. Unruptured AVM involving left frontal lobe as seen on MR angiography (black arrow, A) in 34-year-old male with secondary generalized seizures. Shunting as depicted by intravascular signal within AVM nidus (black arrow, B) and draining superior sagittal sinus (black arrow, C) is seen on ASL color cerebral blood flow maps. Note hypoperfusion involving rest of left middle cerebral artery territory (best seen at white arrows in B, C). Catheter angiography confirmed high flow shunting with attenuation of ipsilateral parenchymal capillary blush (D) compared to contralateral normal side (E). Steal phenomena like these are known causes of seizures in AVMs. AVM = arteriovenous malformation Click for larger imageDownload as PowerPoint slide

Epileptogenic cavernomas may demonstrate peri-ictal hyperperfusion of the adjacent cortex (49). Developmental venous anomalies may also be associated with epilepsy, even when they are isolated (50). Although variable peri-lesional parenchymal perfusion changes may be seen, hyperperfusion must be interpreted with caution since it might be related to micro-shunting within the parenchymal microvasculature, not necessarily a peri-ictal change (51). In Sturge-Weber syndrome, DSC perfusion studies have demonstrated that white matter perfusion is increased in early stages with progressive hypoperfusion with increasing severity and duration of epilepsy (52).

An estimated 20–40% of surgical epilepsy patients have no identifiable causative lesion on sMRI. This adversely affects prognosis and presents difficulties for surgical resection (69). The most common underlying pathologies identified are FCDs, gliosis, hamartomas, and hippocampal sclerosis, with the rest being negative for any histopathological abnormality (70, 71). Invasive EEG monitoring before surgical resection can be aided by prior non-invasive lateralization or localization. PET has shown a high concordance of lateralization with intracranial EEG (44.3–100%), but a more variable localization value (11–69% concordance) (72). PET positivity itself is a good prognosticator for post-surgical outcomes (20). ASL has shown excellent and high concordances with PET in lateralization and localization of EZ. It is a promising utility in the evaluation of MR-negative epilepsies (Fig. 11) (73).

Fig. 11 sMRI negative temporal epilepsy. 15-year-old boy with temporal neocortical epilepsy. Hippocampi appeared normal on T2-weighted image (A). Right temporal hypoperfusion on ASL (arrows, B) suggested right temporal EZ. Confirmatory positron emission tomography showing excellent concordance with ASL (arrows, C). Click for larger imageDownload as PowerPoint slide

Imaging in Acute Settings

Failure to terminate seizures can lead to permanent neurological damage termed status epilepticus (or epilepsia partialis continua for partial seizures). A less easily recognized ‘non-convulsive’ status may also occur (74). sMRI may disclose a range of findings, including gyral swelling and secondary hippocampal or thalamic involvement with variable diffusion restriction, persistence of which indicates poorer outcomes. ASL demonstrates focal or diffuse hyperemia with high sensitivity and concordance with EEG (49). Distribution of abnormalities may encompass the EZ as well as areas of seizure propagation (Fig. 12) (49, 68). ASL also allows serial follow-up and prognostication (74, 75).

Fig. 12 Epilepsia partialis continua. 13-year-old boy presented with continuous right hemispheric partial seizures. FLAIR showed right frontal FCD (arrow, A) lacking significant changes in rest of ipsilateral hemisphere (not shown). Ictal ASL showed intense hyperperfusion of FCD along with secondary hyperperfusion in rest of ipsilateral cerebrum (B, C). This likely represents spread of seizure activity to symptomatogenic zone. Click for larger imageDownload as PowerPoint slide

Accurate identification of the EZ is crucial in presurgical evaluation of epilepsy since it influences prognosis. ASL is a non-contrast MR perfusion technique that demonstrates alterations in perfusion related to abnormal synaptic activity, akin to nuclear imaging. This not only allows localization of the epileptogenic focus, but also provides functional characterization of abnormalities identified by sMRI. ASL is a useful supplementary investigation. It has potential to make MRI a one-stop-shop in epilepsy evaluation.