SPECT and PET in nonlesional epilepsy

Synchronized neuronal discharge activity during epileptic seizures increases the cerebral blood flow in its vicinity by up to threefold [10, 11]. Regional cerebral blood flow (rCBF) can be measured by SPECT using two different tracers: ^99mTc-hexamethyl-propylene-amine-axime (HMPAO) or ^99mTc-ethylene-cysteine-diethylester (ECD) [12, 13]. There have been comparisons of both tracers with the slight benefit of brain to extracerebral contrast in HMPAO [14]. However, our own experience clearly shows that ECD is superior to HMPAO for co-registrations and postprocessing (unpublished data). Unfortunately, the choice of tracer is not only based on comparing data but also on availability, which is rather limited for ECD in Europe at present.

The characteristics of the two different tracers (HMPAO and ECD) show a rapid brain uptake within 20–60 s of injection [15, 16]. Imaging can be safely arranged within 2 h after ictal injection, once the patient has passed the postictal phase. As a result, both tracers show a snapshot of rCBF shortly after injection [17]. Hence, the injection should be performed as early as possible during the seizure [18‐20]. Seizure duration should be longer than 30 s [20].

Interictal changes of rCBF measured by HMPAO-SPECT did not show any significant correlation with the seizure focus [12]. Measuring rCBF with ictal injection, on the contrary, yielded highly significant results of hyperperfusion in the seizure-onset zone. Specificity and sensitivity increased in comparison with interictal perfusion patterns [21].

A further increase of sensitivity and specificity was achieved by adding postprocessing and subtracting interictal from ictal SPECT followed by co-registration on MRI (SISCOM; [22]).

Since the development of SISCOM, further techniques including statistical parametric mapping have been developed and compared [23‐26]. In general, adding statistical analysis to the postprocessing seems to further increase the sensitivity and specificity. As a result, postprocessing of ictal SPECT is regarded the gold standard.

Various tracers are available for imaging in PET. In diagnostic use for epilepsy, the most common tracer is ¹⁸F‑fluorodeoxyglucose (FDG). For presurgical evaluation in temporal lobe epilepsy, FDG-PET is standard; it indicates glucose uptake in the neurons of the brain as an indirect measure of energy consumption [27, 28]. Most often, changes in glucose metabolism indicate hypometabolism reflecting reduced functionality. This is regarded as an indirect measure of neuronal dysfunction. Particularly in temporal lobe epilepsy, glucose hypometabolism in FDG-PET might be widespread, indicating either the epileptogenic network or the functional deficit zone according to the concept of the epileptic zone [29]. In nonlesional temporal lobe epilepsy, FDG-PET may lateralize the epilepsy well. Furthermore, the distribution of the extent of hypometabolism within the temporal lobe and also beyond the borders of the temporal lobe may predict the surgical outcome [30].

In extratemporal lobe epilepsy, and particularly in focal cortical dysplasia, FDG-PET shows focal reduced or even absent glucose metabolism [31‐33].

Besides FDG, ¹¹C‑flumazenil (FMZ) is a further tracer applied in the presurgical diagnosis of the epileptic focus. Flumazenil binds to γ‑aminobutyric acid A/central benzodiazepine receptor (GABA_A/cBZR) in the brain, reflecting indirectly neuronal density [34]. Interestingly, within the CA1 region of the hippocampus, the reduction in GABA_A/cBZR binding was even lower than explained by the reduced neuronal density. Disturbances in neocortical neuronal density are shown as a very focal reduction in FMZ binding on PET [35]. It has also been shown that migration disturbances such as focal cortical dysplasia may show FMZ binding in the white matter areas, particularly in group analysis [36]. More recently, ¹⁸F‑flumazenil has become available and is more suitable for clinical application due to the longer half-life of ¹⁸F [37].

A particular specialized tracer is α‑¹¹C-methyl-L-tryptophan (AMT), which is applied for patients with tuberous sclerosis. The seizure onset zone is in these cases often difficult to diagnose, in particular if several tubers are present. However, in most cases only one tuber is epileptogenic. Here, AMT-PET may help to identify the epileptogenic tuber [38]. Interestingly, a recent study showed a group of five microRNAs that were correlated with tuber epileptogenicity and AMT uptake [39].

Several studies reported on patient cohorts that included either patients with discordant results from MRI and EEG or with nonlesional MRI findings in extratemporal lobe epilepsy. In one of the largest prospective studies, SISCOM contributed to the hypotheses of the epileptic foci in 74 out of 130 patients undergoing ictal SPECT injection [40]. Concordance of SISCOM activation to the side of surgery was 82% in those patients who underwent surgical resection. Ten out of 28 patients had no defined MRI abnormality. Neuropathology in this group showed gliosis, hippocampus sclerosis, or no abnormality, whereas in those who had a possible subtle lesion, focal cortical dysplasias type IIa and IIb were predominantly found. In a similar prospective study, also with a mixed population with discordant or nonlesional MRI findings, SISCOM was applied prior to implantation of invasive electrodes. In 75% of patients the electrode placement was changed according to the SISCOM result [41]. When comparing magnetic source imaging, ictal SPECT and FDG-PET, concordant SISCOM had the highest predictive value (odds ratio [OR]: 9.9) for excellent surgical outcome [42]. Again, most of these patients had extratemporal lobe epilepsy with a nonlocalizing MRI or EEG. Another study compared high-density electric source imaging (HD-ESI) with MRI, PET, and SISCOM [9]. Of 190 included patients, 58 were investigated with all four modalities. The combination of concordant MRI and HD-ESI showed the highest predictive value (92%) for postoperative seizure freedom. The majority of the cases were lesional but eight cases were nonlesional. Using PET and SISCOM together did not add further information if MRI and HD-ESI findings were discordant; however, if only PET and SPECT were taken into consideration, the findings were concordant, and the area of hypo/hyperactivity was resected, the OR of achieving seizure freedom was 6.5. In general, the sensitivity of SISCOM varied from 53% to 85% [9, 40, 41, 43, 44]. Interestingly, all studies showed postsurgical outcome in patients with nonlesional MRI findings comparable to patients with lesional MRI findings undergoing epilepsy surgery. Neuropathology in those nonlesional cases showed in up to 75% features of FCD type IIa or IIb [40, 45].

The impact of PET in nonlesional epilepsy cases has been shown in numerous studies. One recent study of a cohort of 106 patients, of whom 43% were nonlesional, indicated encouragement to proceed with intracranial epilepsy monitoring if concordance of FDG-PET with the clinical consensus of epileptic foci was given. Concordance of the clinical consensus with the SISCOM results, by contrast, predicted a good surgical outcome [46]. The combination of MRI and PET in hybrid scanners increases the detection rate of abnormalities. In a group of 33 nonlesional patients with histologically proven extratemporal focal cortical dysplasia, the combination of PET, MRI, morphometric analysis program (MAP, [47]), and statistical parametric mapping-PET (SPM-PET, [48]) had the highest detection rate of 97% [49]. No case of type I FCD was detected by MAP. Examination with PET-MRI was most sensitive for detecting FCDs type II and IIa. Again, 26 out of 33 patients achieved seizure freedom for more than 1.5 years after resective surgery. Another study using hybrid PET/MRI showed in 40 nonlesional patients a morphological lesion in 18 patients and glucose hypometabolism in 29 patients. Out of 60 included patients, a change in the management plan due to the FDG-PET/MRI results was reported for one-third [50]. This mirrors our own experience, that co-registration of PET and MRI increases the diagnostic yield.

In 53 patients undergoing intracranial EEG monitoring, the preoperative FDG-PET showed 56% sensitivity for the same region as intracranial EEG. By contrast, ictal SPECT had a sensitivity of 87% [44]. A recent single-center experience including 40 patients with nonlesional MRI findings showed that concordance with localized FDG-PET results was one of the predictors of good surgical outcome [51]. The importance of concordant results in nonlesional MRI with functional neuroimaging including FDG-PET was shown to improve postsurgical outcome with up to 75% seizure freedom. Thereby, concordance between two or more investigations (interictal/ictal EEG, MRI, ictal SPECT) with localizing PET were predictive of an outcome of seizure freedom [52]. Postprocessing with SPM or a scanner software increased the pick-up rates for PET abnormalities. This was shown in several studies, which are summarized in a recent review [30].

In our own clinical experience and also reported in some of the papers quoted above, nonlesional MRI-negative epilepsy may change to lesional epilepsy after pinpointing on the basis of PET or SPECT results. As outlined in Fig. 1, very subtle signs of focal cortical dysplasia were identified after a SISCOM-positive lesion and an FDG-PET-negative cortical area pointed out the abnormality in a 28-year-old female patient with frontal lobe epilepsy. Further invasive studies confirmed the seizure-onset zone, and resection has been followed by seizure freedom for more than 2 years at the time of writing. Figure 2 shows another example of localizing a lesion with ictal SPECT and FDG-PET as well as FMZ-PET. Additional findings with MRI identified a subtle cortical dysplasia and again invasive recording confirmed the epileptic-onset zone. After high seizure frequency of weekly focal-to-bilateral tonic–clonic seizures, the patient has been seizure free for 3 months since surgical resection. This is in line with the above quoted reports that nonlesional epilepsy might change to FCD on the basis of histopathology results in a number of cases. Hence, the nonlesional or MRI-negative criterion depends on the quality of MRI, MRI reporting, re-reporting, postprocessing, and possibly the combination with additional investigations such as PET and/or ictal SPECT. Currently, nonlesional epilepsies are most often defined by MRI criteria only [7]. It is advisable to have a second look at nonlesional MRI images once additional information becomes available.

Focal cortical dysplasias are one of the three most common histopathologies in epilepsy surgery [53]. Surgical outcome in refractory epilepsy caused by FCD is excellent [54]. Hence, detection of such pathologies with functional neuroimaging including PET and SPECT may lead to excellent surgical outcome in previously nonlesional cases. Furthermore, the group with no clear histopathology findings showed a fair chance of achieving long-term seizure freedom (53%) after epilepsy surgery [54], which is still much higher compared to seizure freedom in the groups with best medical treatment. Therefore, identification of the epileptogenic-onset zone and its consecutive resection may contribute to fair surgical results in truly nonlesional patients. The importance of functional neuro-imaging with PET and SPECT in this difficult patient group is attributed to its additional functional information.