CHAPTER

Functional Neuroimaging

Jorge Oldan, Terence Wong, and Jeffrey Petrella

Functional imaging, referring here to nuclear techniques such as PET and SPECT, as well as functional magnetic resonance imaging (fMRI), is primarily useful in the surgical treatment of epilepsy. PET and SPECT are primarily used to localize potential candidates for resection of an epileptogenic focus, whereas fMRI is principally used to map crucial areas and aid in surgical planning.

While 70% of epilepsy patients can be treated medically, others may require surgical resection of the epileptogenic focus. Surgery is successful (defined as resolution of seizures over a period of 2 years after resection) in 60% to 90% of cases of temporal lobe epilepsy (TLE) and 22% to 65% of cases of extratemporal epilepsy (1). Usually, MRI is the favored modality for detecting structural abnormalities such as hippocampal or mesial temporal sclerosis (where patients also have the best postsurgical prognosis) (1). However, if MRI is normal (as in 20% to 25% of cases (2)), or localizes the focus to a different location from EEG, nuclear medicine techniques, PET and SPECT, may be used to guide placement of, or in some cases, replace subdural electrodes (1–3) for further study. For defining an epileptogenic focus (EF), ictal SPECT is the most sensitive study, followed by interictal PET and interictal SPECT (1).

PET

Introduction and Technique

PET uses tracers labeled with positron emitters to visualize and quantify cerebral metabolism. The usual tracer is 2-deoxy-2-[18F] fluoroglucose, or FDG, a glucose analogue. Formerly used more extensively, it has been largely supplanted by high-quality MRI. In addition to localizing the EF and suggesting prognosis (particularly for suggesting seizure freedom with a focal abnormality) (4) it can detect subtle anomalies such as focal cortical dysplasia (2). As PET may often not match MRI and EEG, all positive regions should be considered for electrode placement (4).

Unfortunately, tracer availability is somewhat limited by the half-lives of the atoms they are attached to. While SPECT tracers can be obtained from a generator good for a week, PET tracers used in clinical epilepsy imaging must be made by a cyclotron. The most commonly used positron emitter, F-18, has a half-life of 108 minutes, and others have even shorter half-lives (20 minutes for C-11). While F-18-labeled tracers can be bought and transported, C-11 tracers must be manufactured by an on-site cyclotron (2).

PET has superior spatial resolution (about 4–10 mm) (1) than SPECT, but as brain glucose uptake is slower (over 30 to 40 minutes) (5) and since FDG-PET reflects metabolism during the uptake phase, ictal studies are difficult to obtain. PET scans are thus usually interictal (6). The patient is monitored with continuous electroencephalography (cEEG) during the scan (and/or the 45- to 60-minute uptake between injection and scan) to avoid accidentally performing an ictal PET study (1,4). Ideally, the patient should be seizure-free for 24 hours, although 2 to 12 hours is considered acceptable in some centers (1).

FIGURE 21.1 Positive PET scan with hypometabolism in the left temporal lobe, corresponding to hypometabolic region. Upper left: PET image, upper right: MR image, lower left: fusion image using “hot iron” color map, lower left: fusion image using “rainbow” colormap. Note the additional decreased metabolism in the ipsilateral basal ganglia.

Findings may be related to semiology: a study of complex partial seizures found that hypometabolism was limited to the epileptogenic zone in focal limbic seizures, included the limbic cortex in widespread limbic seizures, included the extralimbic frontal lobe with posturing, and included cerebellum and parietal lobes with secondarily generalized seizures (3). Interictal hypometabolism in the putamen is associated with ictal dystonic posturing (4).

Temporal Lobe

The most common focal site of seizure onset is in the temporal lobe. It may be mesial, localized in the hippocampus and due to sclerosis, or lateral, usually neocortical in origin, due to a focal lesion or other cause. Mesial and lateral sites are difficult to distinguish on PET due to spatial resolution (4,5), volume averaging, and widespread temporal lobe hypometabolism in refractory mesial epilepsy (4).

Hypometabolism usually extends outside the lesion itself, although it is less severe extratemporally (4,5). Common sites are the ipsilateral temporal lobe, and less often the frontal (particularly insular and inferior frontal (4)), parietal, and occasionally contralateral temporal lobes (5), as well as, commonly, the dorsal thalamus (4). Mesial TLE specifically may involve the mesial and anterolateral temporal cortex, as well as the basal ganglia and thalamus (1,4), and more rarely the occipital lobe (4). Refractory patients have more widespread disease; conversely, nonrefractory TLE patients may have a normal PET (4,5). The inferolateral temporal lobe is a common site in MRI lesion negative disease (1).

Over time, the volume and severity of hypometabolism in mesial TLE increase (4). Executive function deficits and depression may be associated with orbitofrontal and prefrontal cortex abnormalities, with more frequent seizures causing worse impairment and larger abnormalities on PET; seizure control may decrease the size of these areas, and they may improve after surgery (3,8). Bilateral temporal lobe hypometabolism is more commonly associated with bilateral, diffuse, or extratemporal seizure onset and bilateral or diffuse MRI abnormalities (3,5). These patients have a lower probability of seizure remission after surgery (3,5) or medical treatment (3), and more cognitive deficits (3).

Extratemporal Lobe

PET is less sensitive in the extratemporal region, particularly if MRI is normal (4). Extratemporal seizures tend to spread rapidly, so a rapid injection for ictal SPECT becomes more important (5), and detecting the precise focus is more difficult. However, PET can still be useful in guiding subdural electrode placement (5). The border with normal tissue is more gradual (4,5); if there is no lesion, the area may be quite small (4). Frontal lobe epilepsy frequently has a normal PET scan (3). If not, hypometabolism may be quite widespread, including ipsilateral temporal and parietal lobes and occasionally the thalami and basal ganglia (4,5). Parietal lobe foci appear to be more difficult to locate (3).

Aids to Interpretation

Potential aids include coregistration with MRI and voxel-based comparisons with normative data (2). These are particularly useful in extratemporal epilepsy (8).

Coregistering or fusing PET data with a structural MRI can help identify subtle findings such as cortical dysplasias, either by themselves or in association with tumors, and improve delineation of the epileptogenic zone for surgical planning (9). In general, the PET and MRI have to be registered by commercial software (most conventional PACS systems cannot do this well), which frequently allows for display of a fused image as well. A rainbow-color map protocol using a 15% difference corresponding to a color change and SUV calculation with differences of 10% considered significant has shown some effectiveness in the detection of focal cortical dysplasia (6). The resolution is superior to SPECT, but partial volume correction remains useful (8).

In statistical parametric mapping (SPM), each pixel is treated as a z-score using the mean and standard deviation for that patient’s overall study. Initial findings have shown successful lateralization and correlation with seizure duration, with the anterolateral temporal region being more important in predicting response to surgery (3). SPM may also be more useful in identifying bilateral abnormalities (3), particularly if asymmetric (4). Some authors suggest using a volume threshold to avoid focusing on very tiny, statistically insignificant foci (4).

Validity

Temporal Lobe

In the temporal lobe, PET sensitivity ranges from 80% to 90% (1,2,6), increasing to 90% with quantitative analysis (comparing to standard normal) in some trials (1,5). However, its sensitivity in MRI-normal cases may be as low as 44% (8). Concordance with the ultimate site of resection producing relief is high at 96% (10). Some preliminary work suggests flumazenil (FMZ) may be more effective at determining the area to be resected (5). As this is a C11-tracer, availability is limited to centers having a cyclotron.

Postsurgical outcomes are better with more severe abnormalities restricted to the temporal lobe (3,4) (especially mesial (3,5)), a single focus of hypometabolism (4) ipsilateral to the resection site (3,4,6,8), a greater degree of asymmetry (3,8), a greater portion of the hypometabolic area resected (2–4,6), white matter changes on MRI (3) and no or ipsilateral (versus contralateral) thalamic asymmetry (3). Conversely, extratemporal and bilateral temporal hypometabolism make postsurgical seizure control less likely (4). A lack of asymmetry may indicate extratemporal or bilateral foci (3). Essentially, the more likely it is that the temporal lobe is the problem, the more likely removing a portion of the temporal lobe is to help.

In some cases of refractory mesial TLE, use of intracranial electrodes may be avoided if there is unilaterally predominant hypometabolism when extracranial EEG electrodes show ictal onset from the ipsilateral temporal lobe, MRI is normal, nonspecific, or localized to the ipsilateral temporal lobe, and other data are not discordant (4). However, simply because the most severe abnormality on PET is in a temporal lobe does not indicate that all seizures arise from that location; thus, no other contradictory data should be present. If the EEG is nonlocalizing, intracranial electrodes may still be necessary (4).

Extratemporal

PET sensitivity for extratemporal epilepsy is lower than for temporal lobe epilepsy. It is somewhat higher at 70% to 90% than the sensitivity of 30% to 80% for MRI (6) (although concordance may be somewhat lower at 68% (10)). In the case of frontal lobe epilepsy, sensitivity is lower at 45% to 73%, and even lower at 36% for MRI lesion-negative patients; performance may be somewhat higher in children (5).

Patients with extratemporal cortical hypometabolism ipsilateral to the site of surgery generally had a lower frequency of successful surgery than those with temporal cortical hypometabolism (45%–60% vs. 75%) (3,5,8). PET is of more utility in guiding subdural electrode placement than in detecting the epileptic foci (EF), due to the relatively rapid spread (5). In addition, work using high-resolution registration has suggested that small mismatches between hypometabolic and epileptogenic areas are common in extratemporal epilepsy. PET may also find additional foci (4).

Cortical Dysplasias

Other Uses

PET is useful in neonates with intractable seizures of unknown etiology. A focal abnormality suggests a focal developmental abnormality, whereas diffuse abnormalities suggest a metabolic cause (11).

PET has been shown to detect abnormalities not seen on MRI in cryptogenic infantile spasms (West syndrome). Surgical removal of abnormal foci of cortical dysplasia can stop seizures and sometimes reverse developmental arrest (3,4,11), particularly if PET and EEG are concordant and/or PET anomalies normalize on medical therapy (4).

PET in Lennox-Gastaut syndrome may suggest EF (3). Cortical tubers are hypometabolic on FDG-PET (to about 30% to 50% vs. the other side) (1), and show increased uptake with 11C-AMT—in two-thirds of epileptogenic cases (5). In Sturge-Weber, PET shows hypometabolism ipsilateral to the nevus and demonstrates the extent of involvement in the cortex below, guiding resection and detecting contralateral hypometabolism when hemispherectomy is considered (3–5). Patients with cutaneous but no intracranial anomalies have a normal PET (4). Patients with GLUT1 transporter deficiency show generalized reduction in FDG activity, but relative preservation in the basal ganglia (4). PET is generally not used in primary generalized epilepsy to exclude surgical candidates, as such patients are usually diagnosed from history and EEG; if performed, interictal FDG studies are usually normal, as is sometimes the case with secondarily generalized epilepsy (4).

New Directions

Simultaneous acquisition in a PET-MRI system may improve coregistration, as well as helping to determine motion correction, or obtain multifunctional imaging such as combining PET ligand uptake with processes such as arterial spin labeling, MR spectroscopy, or blood oxygenation-level-dependent imaging (12).

While FDG is the most common tracer used, other tracers can be used to investigate other pathways, such as 11C-FMZ for GABA-A receptors or 11C-alpha-methyltryptophan (AMT), a serotonin precursor (1). Many of these are limited by use of 11C as a radionuclide, whose half-life of 20 minutes not only allows for a smaller radiation dose but also limits administration to institutions with a cyclotron.

Much like FDG, FMZ binding is significantly lower in the EF (1,16), the degree of which may correlate with seizure frequency (3). 11C-flumazenil may have a more restricted area of hypometabolism than FDG (1,9,11), and may be more sensitive for contralateral abnormalities (9) and mesial temporal sclerosis (11). Changes have been seen in the thalamus in temporal lobe epilepsy and contralateral cerebellum in frontal lobe epilepsy as well (3). Larger regions of FMZ uptake may correlate with seizure frequency (1), and periventricular uptake before surgery may correlate with a poor outcome (1). It may be more accurate in extratemporal, in particular, frontal, epilepsy as well (1,3).

11C-AMT has shown to be specific for seizure foci, particularly in tuberous sclerosis (8), as described previously, where epileptogenic tubers have a higher uptake (1). It may also be useful in identification of nonresected epileptogenic cortex (5) or developmental malformations (11).

Among the many other tracers in investigative use are 11C-carfentanil and 11C-methylnaltrindole for opioid receptors and 18-MPPF, a selective 5-HT1A agonist (3). Opioid-receptor tracer uptake may be more localized than FDG and may increase postictally. Synaptic opioid levels may correlate with seizure activity (3). 18-MPPF, a selective 5-HT1A agonist, may be more sensitive and accurate (3) in MRI lesion-negative temporal lobe cases.

SPECT

Introduction and Technique

SPECT evaluates regional cerebral perfusion (rCP), which is generally coupled to metabolism. The seizure zone is hyperperfused during epileptic activity because of local neuronal hyperactivity (8), whereas it is hypoperfused between seizures (6). SPECT, unlike PET, allows for ictal imaging (13). Spatial resolution is generally not as good as PET (about 8–10 mm) (6).

Of the three FDA-approved radiopharmaceuticals, one (IMP, or isopropyl-iodo-amphetamine) is primarily used in Japan. The other two, both using Tc-99m as the radionuclide, are HMPAO (hexamethylpropylamine oxime, Ceretec, GE Healthcare, Chalfont St Giles, UK) and ECD (ethyl cysteinate dimer; Neurolite, Bristol-Myers Squibb, North Billerica, MA). The Tc-99m radionuclide has a longer half-life (6 hours) and is obtainable from generators, allowing for easier availability relative to PET. In general, the compound is bound to a specific area of brain on its first pass through and remains there (being converted to a polar metabolite) for several hours. The patient can thus be injected during a seizure, and the SPECT scan be done hours (up to 2 (13) to 4(1)) later when the patient has recovered and is cooperative. One important difference between the two compounds is that HMPAO must be used within 20 to 30 minutes of its preparation, whereas ECD is stable for 4 to 6 hours. ECD is also rapidly excreted renally, decreasing radiation burden and allowing for administration of a larger amount of radiopharmaceutical with the same overall dose (5). In the experience of the authors, ECD produces a much better quality of image with much higher sensitivity.

In general, the patient is admitted for video EEG (vEEG) monitoring with ictal and interictal scans planned during the same admission. The patient (or their parents) is consented. The tracer is injected in the telemetry suite, with the radiopharmaceutical in a lead-shielded syringe containing the prescribed dose. A vial in a lead container with more of the radiopharmaceutical is available as well, so a sufficient amount of tracer to replace the decayed tracer can be withdrawn if necessary. A sheet is used for displaying an appropriate volume to be withdrawn to account for radioactive decay (5). Note that the compound must be discarded after 6 hours if the patient does not have a seizure, so frequently the radiopharmacy will have to send up to two doses during the day (13).

The exact times of tracer injection and EEG onset of seizure should be noted so one can determine if the injection is ictal, peri-ictal, or postictal (5). Seizures should last at least 5 to 10 seconds, and (ideally) the SPECT tracer should be injected less than 20 seconds (8) after the seizure starts, although 45 (13) to 90 seconds (10) is acceptable. Injection after the seizure—postictal SPECT—is less sensitive (13), as focal hyperperfusion becomes hypoperfusion. This postictal switch is generally complete at 100 seconds after the seizure ends (8). While ictal scans are generally considered superior, a postictal scan may retain some value, at least in temporal lobe lesions. One study showed a pattern of mesial hyperperfusion accompanied by lateral hypoperfusion having a 97% PPV; a postictal injection retains a 70% to 90% sensitivity in mesial TLE (5).

Imaging itself usually takes about 20 to 30 minutes (13). Children 1 to 4 years old (or older if uncooperative) are generally sedated, whereas children less than 1 year of age are fed and wrapped before the scan (5). In general, the patient should be sedated on the ward before arriving in the imaging department (5).

The interictal scan, conversely, is usually performed when the patient has been seizure-free for at least 30 minutes (usually on a different day). The patient is usually injected in the imaging department for this scan, with sedation if necessary for a pediatric patient. The interictal scan is less useful by itself, but aids the localization of the EF together with the ictal scan by visual assessment or subtraction (5).

Images can be reconstructed at any angle (13); however, data should be reconstructed after reorientation parallel to the anterior–posterior commissure line with another set of data reconstructed parallel to the long axis of the temporal lobe (5). A continuous color scale is recommended, to avoid overestimation of tiny asymmetries. A high-resolution or fan-beam collimator should be used (5).

Whole-body effective dose equivalent from a 20 mCi dose of Tc-99m HMPAO is 6.9 mSv, whereas that from 20 mCi of Tc-99m ECD is 5.7 mSv. This is about half the dose for a CT scan of the abdomen and pelvis, and 3 to 3.5 times that of a head CT scan (7).

Interpretation

FIGURE 21.2 Ictal (upper rows-”EARLY”) and interictal (lower rows-”LATE”) SPECT, with focus of increased activity seen in left temporal lobe on ictal imaging, which is not present on interictal imaging. In this case where SPECT images are being interpreted without reference to an outside image, ictal and interictal series are lined up in each of the three reference planes, with manual adjustment of each series for better correspondence.

Patterns of hyperperfusion differ with time of injection, and include (ideally) a large and intense hyperperfused area at the lesion itself (in early injection), a less intensely hyperemic area at the actual lesion with the most intense uptake at a distance (at later injection times), and a complicated, multilobulated pattern (also at later injection times), seen in frontal lobe seizures with fast propagation. In frontal lobe lesions, the largest cluster of ictal hyperperfusion usually includes other brain regions as well as the ictal-onset zone (8).

Patterns of propagation are somewhat distinct for extratemporal epilepsies as well. Occipital seizures quickly spread to both temporal lobes, making early injection important. An anterior spread is often associated with sensory and motor manifestations (5). Frontal lobe epilepsy often demonstrates hyperperfusion in ipsilateral basal nuclei and contralateral cerebellar hypoperfusion (5). A larger area of hyperperfusion than the tumor itself may be seen in dysembryoplastic neuroepithelial tumour (DNET) (5). Focally increased uptake on ictal SPECT in tuberous sclerosis corresponds to sustained rhythmic fast activity on an ictal EEG (5). This may be useful in cases of multiple tubers (5).

Aids to Interpretation