PISCOM: a new procedure for epilepsy combining ictal SPECT and interictal PET

We performed a retrospective review of all patients with medically refractory epilepsy who had undergone presurgical evaluation and subsequent surgical resection at our centre from 2008 to 2013. The inclusion criteria were: (a) diagnosis of medically refractory epilepsy, (b) patient had undergone brain MRI, ictal SPECT, interictal SPECT and FDG PET studies, (c) surgical resection of the suspected EZ, and (d) favourable outcome (Engel class I or II) for at least 2 years after surgery. The patients had been previously evaluated in the clinical work-up to optimally localize the EZ using clinical and neuropsychological examinations, EEG monitoring, and anatomical and functional neuroimaging studies, as described in a previously [8]. On the basis of the inclusion criteria, 23 patients with a mean age of 31 years (range 4–61 years) were retrospectively identified (10 male and 13 female, 18 adults and 5 children). Informed written consent was obtained from all patients or their parents or legal guardians, and all procedures were approved by the hospital Ethics Committee.

Patient candidacy for surgery was established based on the standard evaluation protocol of the epilepsy unit according to the clinical, electrophysiological, and neuroimaging data obtained. Lobar or selective cortical resection was performed by neurosurgeons from the epilepsy unit. Histological analysis of all the brain tissue resected showed that 9 of the 23 patients (39%) had mesial temporal sclerosis, 9 (39%) had focal cortical dysplasia, two had mesial temporal sclerosis associated with focal cortical dysplasia, one had gliosis, one had low-grade glioma and one had grade I ganglioglioma (Table 1).

Postoperative seizure outcome was classified according to the Engel classification. Only patients with a favourable outcome (Engel class I or II) for a period of at least 2 years after surgery were included in the study. All 23 patients were classified as Engel class I with a median follow-up time of 3.41 years (range 2–5 years; Table 1).

The EZ was considered the gold standard based on surgical resection, histology results and favourable surgical outcome. The EZ was located in the temporal lobe in 15 of the 23 patients (65%), with 10 patients (43%) having lesions in the medial temporal lobe, 3 (13%) in the lateral neocortical temporal lobe and 2 (9%) in both the medial temporal and the lateral neocortical temporal lobe. Of the remaining 8 patients, the EZ was located in the frontal lobe in 6 (26%), the insula in 1 (4%) and the parietal lobe in 1 (4%; Table 1).

In all of 18 adult patients, MRI was performed with a 3-T unit (Tim Trio; Siemens, Erlangen, Germany) and in the 5 paediatric patients with a 1.5-T unit (Signa Exite; GE Healthcare, Milwaukee, WI) with a specific epilepsy protocol. All the MRI studies were interpreted visually by an expert neuroradiologist not blinded to the clinical data or the results of other tests.

Ictal SPECT was performed as part of patient admission for video-EEG monitoring in the epilepsy unit of our centre. Seizure onset was defined as the time of earliest indication of auras or the beginning of rhythmic ictal discharges detected by continuous video-EEG monitoring. The radiotracer was administered intravenously at seizure onset. Interictal SPECT was performed during the week after the patient had been free of seizures for more than 24 h. Ictal and interictal SPECT images were acquired within 2 h of injection of approximately 925 MBq of ^99mTc-hexamethylpropyleneamine oxime (^99mTc-HMPAO) radiotracer following the same protocol using a dual-head SPECT imaging system (Infinia™ Hawkeye™ 4; GE Healthcare Milwaukee, WI).

Interictal FDG PET images were acquired in 3D mode using a PET/CT system (Biograph; Siemens, Erlangen, Germany) after intravenous injection of approximately 5 MBq/kg of ¹⁸F-FDG. In selected cases, EEG monitoring was performed during FDG PET image acquisition to confirm interictal state. SPECT and PET studies were performed within an overall mean period of 86 days (range 7–201 days). No changes in medication or clinical epilepsy features were documented during this period of time.

SPECT images were reconstructed using iterative algorithms including scatter, attenuation and point spread function (PSF) corrections with the aim of obtaining SPECT images with resolution and degradation corrections more similar to those of FDG PET. Perfusion SPECT subtraction according to SISCOM methodology was performed using FocusDET, a previously developed software toolbox for SISCOM analysis [26].

With the aim of subtracting interictal PET images from ictal SPECT images, FDG PET images were reprocessed (rFDG PET) to assimilate features of the PET and SPECT images as follows: (1) FDG PET images were resampled to obtain the same matrix and voxel size as those of interictal perfusion SPECT, (2) FDG PET images were filtered to achieve similar smoothing between the SPECT and PET images (the FDG PET filter was derived from experimental PSF images of the SPECT and the PET acquisition systems), and (3) intensity normalization was applied to FDG PET images to correct for differences in the total number of photons detected (Fig. 1). As the loss of resolution that the real image suffered in the acquisition was due to the PSF, the FDG PET filter in step 2 was derived from experimental PSFs obtained from the SPECT and PET systems by using a radioactive point source. Fourier transforms of SPECT and PET PSFs were then calculated and both distributions were fitted to a gaussian distribution to avoid problems derived from high-frequency noise. Finally, the filter was obtained as the inverse Fourier transform of the quotient of the gaussian associated with SPECT and that associated with PET.

The PISCOM method was performed using FocusDET, following the same steps as the SISCOM analysis but replacing the interictal SPECT image with the rFDG PET image.

Systematic differences in the inherent biodistributions of ^99mTc-HMPAO and ¹⁸F-FDG were compared between the interictal SPECT and rFDG PET studies using the statistical parametric mapping (SPM8) software (Wellcome Department of Imaging Neuroscience, Institute of Neurology, London) [27]. A paired t test was performed (two sample t test: group 1, 23 interictal SPECT images; group −1, 23 rFDG PET images) applying two contrast levels [1 –1] and [−1 1]. If any significant differences related to the biodistribution of the radiotracer were detected at a subcortical or cerebellar level, a mask based on these findings was created to exclude these areas from the visual and volume of interest (VOI)-based analysis of the PISCOM images.

The PISCOM and SISCOM images were interpreted separately by two experts in nuclear medicine blinded to the clinical data and the results of other tests. PISCOM and SISCOM studies were superimposed on a basic T1 MRI sequence only to provide anatomical reference. In addition, the automated anatomical labelling (AAL) template was used for better anatomical localization of the findings. Abnormal findings were defined by visual evaluation of brain areas with activity greater than two standard deviations above the mean activity. The SOZ was defined as the cortical area with the greatest subtraction values that could not be explained by propagation, residual count data or artefacts. The presence and location of the SOZ and propagation activity with both modalities was reported. Both imaging modalities in each patient were presented randomly to ensure that the review of one study would not influence that of the other. In cases of discrepancy between reviewers, a third nuclear medicine specialist was consulted to achieve consensus.

The PISCOM and SISCOM results were compared with each other and with the EZ location. Successful identification of the SOZ had to match the PISCOM or SISCOM findings with the known EZ location at a sublobar level of accuracy; hemisphere or simply lobar concordance was considered as unsuccessful localization. Nevertheless, a focus located in the temporal pole was considered successful identification of the SOZ regardless of whether the EZ was located in the medial temporal lobe or the lateral neocortical temporal lobe. When PISCOM or SISCOM failed to identify any focus, the study were classified as “negative”. In summary, each PISCOM or SISCOM study was classified as “successful”, “unsuccessful” or “negative”.

To perform a more objective comparison between the PISCOM and SISCOM results, a supplementary analysis of all 15 concordant successful studies was carried out. The extension of each SOZ was quantified, as well as the amount of indeterminate activity that could hamper the identification of the SOZ with both modalities. First, the maximum value of the subtraction image in each SOZ identified by PISCOM and SISCOM was determined. VOIs were then obtained in each PISCOM and SISCOM study by intensity thresholding (60% of the maximum SOZ value) and clustering (minimum 100 voxels). Thus, each PISCOM and SISCOM study was segmented into several VOIs, one corresponding to the SOZ and the remainder corresponding to VOIs that matched the threshold criteria which were considered as indeterminate activity (i.e. propagation, background or artefact). Specific algorithms were developed for this task. The number of VOIs in each PISCOM and SISCOM study was determined to evaluate the amount of indeterminate activity revealed by the two modalities. In addition, the number of voxels of each VOI corresponding to the SOZ delimited by PISCOM and SISCOM was quantified to compare the SOZ voxel extension in each patient. The percentages of voxels shared by the SOZ from the two modalities were calculated.

The proportions of the PISCOM and SISCOM results that were “successful”, “unsuccessful” and “negative” were calculated and compared (McNemar test of symmetry). The global agreement between the results of the two techniques was calculated. The proportions of successful identification of the SOZ by PISCOM and SISCOM are reported with their 95% confidence intervals (CI). Variables derived from the VOI-based analysis (number of VOIs and SOZ voxel extension) were compared using a paired t test (p < 0.05 was considered significant). Statistical analyses were performed using SPSS Statistics, version 17.0 (SPSS Inc., Chicago).

The paired t test comparing interictal SPECT and rFDG PET (contrast [1 –1]) showed no significant differences in cortical activity. However, the SPECT group showed higher activity than the rFDG PET group in the cerebellum, midbrain and basal ganglia (p < 0.05, corrected for multiple comparisons using the family-wise error rate; Fig. 2). The paired t test comparing rFDG PET and interictal SPECT (contrast [−1 1]) showed no significant cortical or subcortical differences.

PISCOM and SISCOM showed equivalent results in 17 of 23 patients with good concordance (global agreement 74%, 95% CI 56–87%; Fig. 3). PISCOM successfully identified the SOZ in 19 patients (83%) and SISCOM in 17 patients (74%). However, there were no significant differences in the proportions of the PISCOM and SISCOM results that were “successful”, “unsuccessful” and “negative” (p = 0.368).

The two techniques showed concordant successful SOZ localization in 15 of the 23 patients (65%). Of the two remaining patients with concordant but unsuccessful localization, the SOZ was located at a lobar level in one (P14), and in the other (P21) PISCOM and SISCOM failed to identify any focus. Temporal lobe epilepsy was successfully identified in 13 of 15 patients (87%) by both PISCOM and SISCOM. Of the eight patients with extratemporal epilepsies, the SOZ was successfully identified by PISCOM in six (75%) and by SISCOM in four (50%). The PISCOM and SISCOM results were discordant in 6 patients (26%). Together, PISCOM or SISCOM successfully identified the SOZ in 21 patients (91%; Table 1 and Fig. 3). Propagation activity was visually observed by PISCOM in 14 patients (61%) and by SISCOM in 15 patients (65%), and in nine of these patients coinciding propagation was observed at a lobar level of concordance. In 19 patients (83%) MRI successfully identified a lesion consistent with the EZ location. In three of the remaining patients (P10, P12 and P15), studies were reported as nonlesional or negative, and in one patient (P05) MRI detected a right parietal cavernoma nonconcordant with the SOZ localization; PISCOM and SISCOM successfully identified the SOZ in all four of these patients (Table 1).

SISCOM showed a higher number of VOIs from indeterminate activity in the 15 concordant successful studies (Table 2) with statistically significant differences compared with PISCOM (mean total numbers of VOIs: 11.3 for SISCOM and 3.5 for PISCOM; p = 0.001). Table 2 shows the results of the VOI-based analysis, and the results in an example patient are shown in Fig. 4a.

SOZ voxel extension delimited by PISCOM and SISCOM was equivalent, with no significant differences between PISCOM and SISCOM SOZ volumes (mean SOZ VOI volume: 12.58 cm³ with PISCOM, 13.01 cm³ with SISCOM; p = 0.67). SISCOM and PISCOM studies in all 15 patients with concordant successful results showed some degree of spatial overlap between the two SOZ VOIs with a mean 54% of shared voxels (range 19–84%, Table 2). The results in example patients are shown in Fig. 4b.

This study was an initial proof-of-concept study in the development of a new image processing technique that combines the valuable perfusion and metabolic information provided by ictal SPECT and interictal FDG PET studies. Our results suggest that perfusion and glucose metabolism in the cerebral cortex can be correlated and that PISCOM could be a valid technique for identification of the SOZ, although further studies are needed to validate these results.