This single-institute prospective study was approved by the institutional review board. Informed consent was obtained from all patients.
Patients
From November 2014 to March 2018, 86 consecutive patients with focal epilepsy were enrolled in this study. Inclusion criteria were as follows: (1) an International League Against Epilepsy (ILAE) outcome scale [11] score of 5 or less and (2) availability of both FDG-PET/CT and FDG-PET/MRI results. The exclusion criteria were as follows: (1) unknown EZ (n = 24), (2) no surgery (n = 26), and (3) surgery without resection, such as multiple subpial transection, corpus callosotomy, and vagus nerve stimulation (n = 5). As a result, a total of 31 patients (17 males, 14 females; median age, 31 years, age range, 8–58 years) were enrolled.
PET/CT system
A biograph mCT PET/CT system (Siemens Healthcare) was used according to a standard clinical protocol [12]. Patients were instructed to have no caloric intake for 4 h prior to FDG administration. All patients were normoglycaemic (blood glucose below 150 mg/dL) at the time of FDG injection. After the injection of a weight-based dose of FDG (4.0 MBq/kg), patients rested for 60 min in a dark quiet room to allow for tissue uptake. PET scans were performed for 10 min using 3D acquisition and time-of-flight technology. The crystal of the PET/CT system was lutetium oxyorthosilicate, and the imaging matrix was 400 × 400. Each dataset consisted of 90 transaxial PET images with a 2-mm slice thickness and a 256-mm field-of-view (voxel: 2 × 2 × 2 mm3). CT for attenuation correction was acquired for each patient with the same protocol (120 kVp, 50–135 mAs, detector 64 rows × 1.2 mm). Images were reconstructed using a time-of-flight and commercially available technique, the TrueX (Siemens) technique, with an all-pass filter (iteration, 8; subsets, 21).
PET/MRI system
All patients were scanned using PET/MRI within 30 min after the PET/CT exam using the Ingenuity TF PET/MRI system (Philips Healthcare). After the scout image was taken and a 3D-T1-weighted image was acquired to correct attenuation [10], patients underwent PET imaging with 3D-ordered subset expectation maximisation (3D-OSEM) and time-of-flight. The field-of-view for the PET imaging was 576 × 576 mm2 (recon. voxel: 2 × 2 × 2 mm3). After PET imaging, 3D fluid-attenuated inversion-recovery (FLAIR) images were obtained using a turbo-spin echo sequence with repetition time/inversion time/echo time: 4800/1650/293 ms; flip angle, 180°; turbo-spin echo factor, 182; bandwidth, 1187.1 Hz; field-of-view, 250 × 250 × 220 mm3; matrix, 252 × 250 × 220; thickness of image, 1 mm; number of excitations, 1; number of slices, 220; and spectral inversion-recovery was applied for fat suppression; SENSE factor, 2. The scan time in our protocol was approximately 60 min.
Observer testing
Five board-certified radiologists namely one PET/MRI specialist, one nuclear medicine specialist, and three neuro-specialists (22, 21, 20, 13, and 7 years of experience, respectively) who were blinded to patient information conducted the observer tests. Each observer attended two reading sessions held at least 1 month apart to minimise learning effects [13]. They read either FDG-PET/CT or FDG-PET/MRI in the first session and vice versa in the second session. After the second session, we performed an additional standalone MRI session to verify how FDG-PET affected the diagnostic performance. All images of all 31 patients were presented in a randomised order every session. In the standalone MRI session, both conventional and dedicated-epilepsy MR protocols were used, which included axial and paracoronal FLAIR images reconstructed in the long axis and perpendicular to the hippocampi.
Sensitivity
Firstly, the observers were instructed to observe the laterality of FDG uptake using axial as well as coronal PET images. Secondly, the observers determined the anatomical parts of the EZ using all three types of images, i.e., PET, CT, and MRI, as well as fused PET/MRI or PET/CT. EZ detection was confirmed when both the laterality and the anatomical part of the EZ were detected. The sensitivity of each system was calculated from the EZ detection results from both PET systems plus standalone MRI. The operative report was used as a reference standard when readers disagreed.
Quantitative visual assessment
The observers also evaluated the boundary of the lesion using a 4-point visual score. We used the modified Paldino et al method [9]. The various scores were defined as follows:
-
Score 4: clear detection of EZ in terms of both laterality and border
-
Score 3: clear detection of EZ laterality, but a border that was a little obscure
-
Score 2: possible detection of EZ laterality and an unclear border
-
Score 1: poor quality for detection of EZ laterality and border
The observers were able to easily adjust the window level and window width, as well as the degree of transparency using the workstation (SYNAPSE VINCENT; Fujifilm Medical).
Statistical analysis
The sensitivity of FDG-PET/MRI, FDG-PET/CT, and standalone MRI for the EZ detection was compared using the McNemar test with Bonferroni correction. The intraclass correlation coefficient (ICC) was calculated and interpreted as follows: excellent agreement, ICC > 0.8; good agreement, ICC > 0.6; moderate agreement, ICC > 0.4; and poor agreement, ICC ≤ 0.4, as proposed earlier among the five observers [14]. The 4-point visual score was also compared for the three systems using Dunn’s multiple comparisons test. All statistical analyses were performed using a commercial software program (GraphPad Prism 7.0; GraphPad Software). A p value < 0.05 was considered statistically significant.