18f-fdg Pet/mr in Focal Epilepsy: a New Step for Improving the Detection of Epileptogenic Lesions

Purpose: Hybrid PET/MR is a promising tool in focal drug-resistant epilepsy, however the additional value for the detection of epileptogenic lesions and surgical decision-making remains to be established. Methods: We retrospectively compared 18 F-FDG PET/MR images with those obtained by a previous 18 F-FDG PET co-registered with MRI (PET+MR) in 25 consecutive patients (16 females, 13-60 year-old) investigated for focal drug-resistant epilepsy. Visual analysis was performed by two readers blinded from imaging modalities, asked to assess the technical characteristics (co-registration, quality of images), condence in results, location of PET abnormalities and presence of a structural lesion on MRI. The clinical impact on surgical strategy and outcome was assessed independently. Results: The location of epilepsy was temporal in 9 patients and extra-temporal in 16 others. MRI was initially considered negative in 21 of them. PET alone demonstrated metabolic abnormalities in 19 cases (76%), and the coregistration with MRI allowed the detection of 4 additional structural lesions. PET/MR was considered better performing than PET+MR in 56% of patients. The increase in sensitivity was 13% and new structural lesions (mainly focal cortical dysplasias) were detected in 6 patients (24%). Change of surgical decision-making was substantial for 40% of patients, consisting in avoiding invasive monitoring in 6 patients and modifying the planning in 4 others. Seizure-free outcome was obtained in 13/14 patients who underwent a cortical resection. Conclusion: Hybrid PET/MR improves the detection of epileptogenic lesions, allowing to optimize the presurgical work-up and to increase the proportion of successful surgery even in the more complex cases.


Introduction
The detection of epileptogenic lesions is crucial for identifying the best candidates for surgery in drug-resistant focal epilepsy. Among the different types of lesions causing epilepsy, focal cortical dysplasias (FCDs) represent one of the most frequent both in children and in adults [1]. Surgical resection of the dysplastic cortex allows a favorable outcome [2][3], however FCDs can be di cult to identify by imaging even though using optimal MRI protocols [4][5][6][7]. 18 F-FDG PET has proved to be useful in negative MRI cases, and the additional value of the co-registration of PET and MRI has been demonstrated [8][9][10][11][12][13]. More recently, new hybrid PET/MR imaging has been developed, allowing simultaneous acquisition of brain metabolism images provided by 18 F-FDG PET and morphological/functional images by MRI under the same clinical conditions. This new technique aimed to minimize the gap related to suboptimal co-registration when PET and MRI images were obtained separately on different machines. In addition, PET/MR decreased the radiation exposure compared with PET-CT, a notable advantage, especially in children. The role of PET/MR has been lately assessed in focal epilepsy and some studies have shown an increased its accuracy compared with PET-CT and MRI performed separately for the localisation of the epileptogenic zone (EZ) and the detection of structural lesions [14][15][16][17][18]. However, the additional value for the surgical decision-making of this new hybrid technique remains to be established.
We aimed to assess the role of PET/MR in a population of patients investigated for intractable focal epilepsy and candidates for surgery. We compared the results of PET/MR images with those obtained by a previous PET coregistered with MRI (PET + MR) in the same patients. The rst analysis consisted in a systematic comparison of both imaging data including technical characteristics, sensitivity, and speci city. The second analysis aimed to determine the clinical impact of the PET/MR based on 1) the detection of new structural lesions compared with the reference images; 2) the potential change in decision-making and surgical management; 3) the results of the new surgical strategy.
We hypothesized that the PET/MR may provide a better quality of co-registration and images, a higher degree of con dence for the visual analysis, an increased rate of lesion detection, and nally a higher number of patients referred for surgery and achieving a favorable outcome.

Patients And Methods
The study was conducted retrospectively in a single institution (SHFJ, CEA, Orsay, France) including patients referred for intractable focal epilepsy between 2017 and 2020.
Inclusion criteria.
1) All consecutive patients who underwent an 18 F-FDG PET/MR at SHFJ for intractable focal epilepsy, having a previous 18 F-FDG PET (in the same institution or not) available for review and MRI including 3D-T1 and FLAIR (Fluid Attenuation Inversion Recovery) sequences for the co-registration of PET and MRI; 2) Clinical data available (history of epilepsy, localisation of epileptic focus, result of surgery when performed); 3) Informed consent given by patients or parents for children. The study was approved by the local institutional review board (SHFJ-Codir 06-24-2019).
Presurgical work-up and surgery.
Presurgical evaluation was conducted in the Epilepsy Unit (Department of Neurosurgery, GHU Paris-Sainte-Anne), including video-EEG recordings, 3T morphological and functional MRI, and neuropsychological assessment for all patients. In addition, stereo-electroencephalography (SEEG) was performed in nine patients. SEEG electrode implantation was based on the integration of electroclinical and imaging data including PET results [19]. Cortical resection was performed using microsurgical techniques, after a multidisciplinary confrontation of the presurgical data. Histological analyses of the cortical specimen and classi cation of FCDs were based on international recommendations [20]. Radiofrequency thermocoagulations (RFTC) consisted in performing stereotactically coalescent thermal lesions to destroy a brain target as previously described [21]. Seizure-outcome assessment was based on Engel's classi cation [22]. Imaging data. 18 F-FDG PET/MR. All acquisitions were performed on the SIGNA PET-MR machine (GE Healthcare, Milwaukee, Wisconsin, USA) based on a 3T 750w MR scanner, a whole-body MRI system with a 70 cm patient bore. The PET detector ring comprised 28 identical water-cooled modules, with a total of 20160 Lutetium-based scintillator (LBS) crystals (720 per module, crystal dimension: 3.95 X 5.3 X 25 mm 3 ) read by arrays of Hamamatsu SiPM (Silicon PhotoMultiplier) devices (HPK S12044). Other main characteristics were as follows: PET bore diameter: 60 cm, axial FOV: 25 cm, sensitivity: 22.9 cps/kBq, trans-axial spatial resolution: 4.3 mm at 1cm [23]. Attenuation correction (AC) was based on fast zero-echo-time (ZTE) MRI derived from proton-density-weighted ZTE images by applying tissue segmentation and assigning continuous attenuation values to the bone [24]. 18 F-FDG was injected intravenously 30 min before image acquisition, at a mean dose of 3MBq/kg of body-weight, in an awake and resting state, in a quiet, dimly lit environment, and carefully monitored for head movements and ictal events.
The MRI protocol was similar to the one used for patients investigated for drug-resistant epilepsy, including anatomical sequences in 3D gradient echo T1 (1 mm isotropic, 256 x 256), coronal 2D T2 FSE (2.8 mm, 640 x 448) in hippocampal plan for temporal lobe epilepsy (TLE) and AC-PC plan for extratemporal lobe epilepsy (ETLE), and 3D FLAIR (1.4 mm, 224 x 224) sequences as well as other sequences not used for the present study: 2D BOLD (blood oxygen level dependent) echo-planar imaging (EPI) and diffusion tensor imaging (DTI, 40 directories). The duration of the exam was 30 min.
Reference Imaging. For 14 patients, previous 18 F-FDG PET examination was performed in the same centre as the PET/MR (SHFJ), either on a PET-CT (Siemens Biograph 6) in 10 patients or on a 3D head-dedicated PET camera (Siemens Exact HR+) in four patients. For the 11 other patients, it was performed on different PET cameras, the characteristics of which are detailed in Table 1. 18 F-FDG was injected intravenously at a mean dose of 3MBq/kg of body-weight in the same conditions as for the PET/MR. Image acquisition started 30 min after the injection and ended 15-20 min later. Table 1 Technical characteristics of the different PET cameras used in the study. Anonymisation of the data Data was de-identi ed using the Dicom dedicated tool from GE Healthcare. A random number was generated for each subject imaging session as last name, and "MRI" and "PET" as rst name to separate both modalities from the PET/MR or PET and MRI examinations for separately acquired data.

Results
Patient population.
Epilepsy onset ranged from 2 to 35 years (mean: 10 years, SD: 7.1) and duration of epilepsy from 4 to 36 years (mean: 16 years, SD: 9.1). Epilepsy was drug-resistant in all patients, with a high seizure frequency in most cases (daily in 16, weekly or monthly in the others). The treatment consisted of the association of two antiepileptic drugs (AED) for 14 patients and of three AED for 11 others. The presumed location of the epileptic focus was temporal (9 cases), frontal (9 cases), central (6 cases) and parietal (1 case). The lateralisation was on the right hemisphere in 10 cases, on the left in 13 cases and undetermined in 2. The MRI was initially considered negative in 21 patients and positive in the four others (hippocampal sclerosis in three and FCD in one).

PET/MR and PET + MR acquisition
The delay between the PET/MR and the previous PET ranged from 0 to 10 years (mean: 2.9, SD: 2.3). In two patients, the PET/MR was performed immediately after the PET-CT (and 45 min after 18 F-FDG injection), to assess potential differences between the two exams. For the other patients, the reason for performing the PET/MR was a lack of su cient information provided by the previous PET + MR ( during the precedent night, but none occurring less than six hours before injection. One patient had a short seizure 20 min after injection. Another patient suffered from diabetes with relatively high glycemia (1.6g/l) before the two exams, resulting in poor quality of images.
PET/MR versus PET + MR analysis (   The PET/MR obtained higher scores than the PET + MR for each item and for the two readers. The best improvement was noted for the quality of images (reader 1) and the co-registration (reader 2), while a higher degree of con dence was obtained for both readers. Considering the sum of scores and the change of category (PET and/or MR negative becoming positive), the PET/MR was stated better performing than the PET + MR in 14 cases (56%) for both readers (Fig. 1), whereas the PET + MR was considered better in one case by reader 1 and four cases for reader 2 (mainly due to the poor quality of the MRI due to motion artefacts). Notably, the delay between both exams was similar whether the PET/MR was considered better than the PET + MR or not (mean: 3 years each). The extent and location of hypometabolism was somewhat different for the two readers with a better detection in di cult regions (central, parietal and insula) and small areas of hypometabolism for reader 2, traducing the role of experience for the results of analysis.
Interestingly, the PET/MR allowed the detection of new structural lesions by the two readers. For reader 1, the positive MRI rate increased from four to eight (16-32%). For reader 2, this rate was high in two modalities (10 positive MRI after the co-registration with PET contrasting with four for MRI stand-alone), however the PET/MR allowed the detection of four additional structural lesions, corresponding to FCD type 2 proven by the histological examination in three of them ( Fig. 2). Finally, the sensitivity was higher for the PET/MR compared with the PET + MR for both readers (76% versus 59% respectively for the reader 1; 95% versus 86% for the reader 2). However, the speci city was slightly lower due to few true negative ndings. After consensus, the sensitivity reached 96% and the speci city 100%.
Clinical impact of the PET/MR (Table 3).  All patients of this series were candidates for surgery, but for most of them invasive monitoring (SEEG) was discussed on the basis of the classical presurgical work-up. In addition, visual assessment of the PET was considered negative in six patients, and the MRI was negative in 21 patients. After co-registration of the PET and MRI, four new structural lesions were identi ed. After the PET/MR, only two PET exams remain negative and six additional new lesions were detected on the MRI (24%). Considering the data provided by the PET/MR, the initial planning of SEEG in 17 patients was reduced to 11. Among the six patients in whom the invasive monitoring was avoided, ve had a subtle structural lesion found on the PET/MR, corresponding to FCD type 2 con rmed by histology, while no speci c lesion was found in the latter patient. All patients operated on without SEEG achieved a favorable outcome after the cortical resection. For the 11 other patients where the SEEG remained indicated, the planning of electrode implantation was modi ed in four of them. Therefore, the change of surgical strategy was substantial for 10 patients (40%). Of note, surgery was not indicated in only two patients, due to a bene t/risk ratio considered too low for this counselling. For the other patients, most of those who underwent a corticectomy achieved a seizure-free outcome (Engel class I in 12/13 cases (92%), with a minimal 6-month follow-up (up to 4 years, mean 1.5). In addition, one patient was also seizure-free after the RFTC.

Discussion
In this study, we demonstrate the additional value of the PET/MR and its in uence on the surgical management of epileptic patients examined for intractable focal epilepsy. Based on a series of 25 patients with well-de ned EZ and successful surgery for most of them, we can precisely assess the role of the new hybrid PET/MR compared to the reference imaging. We emphasized that the studied population accounted for the most di cult patients in epilepsy surgery including a high rate of extratemporal location and initially negative MRI cases.
First of all, we show that the PET/MR offers better accuracy than the PET + MR, in terms of co-registration and quality of images, and therefore a higher degree of con dence for the reader, whatever his/her training and experience. Notably the sensitivity was increased of 13-17%, despite a slight decrease of speci city due to the small number of negative studies. However, the speci city was improved after consensus. Previous studies have established the equivalent properties of the PET/MR compared with the PET-CT by visual or quanti ed analysis [16,[25][26][27]. Other studies have shown the higher sensitivity of the PET/MR to detect epileptogenic lesions compared with PET and MRI stand-alone [15,[17][18], but to our knowledge, the additional value of the PET/MR compared with the PET + MR has not been demonstrated to date.
Importantly, we found that the PET/MR allowed the detection of structural lesions in a high proportion of patients (24 %) in whom the MRI was initially considered negative, despite optimal examination including the co-registration with the PET. This multimodal approach has been previously reported as increasing the detection of FCDs [8][9][10][11] and considered as the reference method that should be recommended as a standard procedure for MRI negative ETLE [12]. However, according to our data, the PET/MR reaches a new step for the detection of such epileptogenic lesions compared to the reference imaging. Reasons for this improvement may be related to 1) a better co-registration of the PET and MRI provided by the PET/MR; 2) a better detection of small lesions due to the simultaneous acquisition of high-resolution 3D-T1 MRI sequences (especially when located at the bottom of the sulcus, that is frequent for small FCDs); 3) a better analysis of FLAIR images guided by PET ndings. Despite advances in fusion software, it may be di cult to generate optimal co-registration of PET and MR images acquired separately on different machines, at different times and under different conditions, including post-ictal state. In addition, visual analysis of 3D-T1 and FLAIR images is greatly facilitated by the simultaneous acquisition of the PET. It should also be noti ed that in healthy volunteers, the time of day may in uence the brain morphometric measures, including gray matter, white matter and cerebrospinal uid, with a highly signi cant reduction (more than one percent per 12hr day) in apparent brain volume, from morning to evening [28]. Moreover, these variations were found predominantly in frontal and temporal lobes, which are frequently involved in epileptic patients. This observation argues for the simultaneous acquisition of PET and MRI for a better accuracy of the co-registration, especially when searching for millimetric FCDs.
Worthy of note, despite initial concerns about the correction of attenuation with PET/MR, equivalent properties with PET-CT have been demonstrated afterward [16,26]. Interestingly, using ZTE-AC, the overall 18 F-FDG uptake bias has proved to be 25% lower than that of atlas-AC [24]. This progress was achieved by the improved registration and assignment of correct attenuation values for the skull. Especially in central structures, insula and cingulate areas, a signi cant improvement with ZTE-AC was found in comparison to clinical atlas-AC. This could be particularly relevant in our patient population in which such location was frequently involved. On the other hand, ZTE-AC was found to misclassify some of the nasal and mastoid areas as bone, which caused overestimation of 18 F-FDG uptake in the temporal lobe and cerebellum [24]. However, such inconvenience was of less importance since temporal hypometabolism is usually easy to detect, and a potential overestimation is unlikely to provide a false diagnosis.
Another potentially important advantage of the PET/MR is to compare the metabolic activity with other functional MR images such Arterial Spin Labeling (ASL), that could be highly in uenced by the physiologic and pathologic conditions, especially the interictal and post-ictal states. This comparison was not tested in the present study but good correlations between the two modalities have been reported, providing complementary information in determining the EZ [25,29].
A last (but not least) point is the one stop-shop concept of the hybrid PET/MR, which is more convenient for patients and clinicians. Indeed, for the patient, there is only one scan, which is more comfortable and less stressful, especially for children and patients with high seizure frequency. For the clinician, it is also more convenient to obtain both high quality PET and MR for a simultaneous reading.
Finally, considering the new information provided by the PET/MR in this study, the surgical management was modi ed in 40% of the patients. Invasive monitoring (SEEG) was avoided in six patients, without loss of chance since all were seizure-free after the cortical resection. Furthermore, all patients having FCD type 2 proven by histology (8 cases) were operated on without SEEG and achieved a favorable outcome. These results con rm those previously reported by our team in patients operated on for FCDs [9,11] and support the view that SEEG may be avoided on the basis of PET/MR ndings. For the patients in whom SEEG remained indicated, the planning of electrode implantation was modi ed in a third of the cases, allowing a better de nition of the EZ. Overall, the PET/MR was considered to improve the surgical decision-making and outcome in nearly half of the patients.
Interestingly, the utility of the 18 F-FDG PET alone was previously estimated at 53% in a large series of epileptic patients investigated for surgery [30]. In addition, improvement of seizure-free outcome has been observed after changing the imaging protocol including the PET in a pediatric population [31]. Moreover, the PET/MR was reported as providing better accuracy of electrode implantation during SEEG [32]. It can be suggested that the PET/MR may still increase the positive impact already provided by the PET alone in epilepsy surgery, as demonstrated in our population.
The nal goal of a new technique is to prove its value for improving diagnosis and curative treatments for patients.
Considering epilepsy surgery, seizure outcome and long-term results including stopping the antiepileptic medication are highly dependent on nding an epileptogenic lesion in the cortical specimen removed during the operation [33]. In ETLE and negative MRI cases (so called "non lesional epilepsy"), seizure-free outcome is commonly reported lower than in TLE and well-identi ed lesion on MRI. It was reported ranging from 15-38% in previous large series of non-lesional ETLE in which the role of PET was not considered [34][35][36] and 27-46% in a previous meta-analysis [37]. In addition, non-lesional epilepsy was associated with a two to three times lower rate of seizure-free outcome than lesional epilepsy [38]. Contrasting with these results, the chance of being seizure-free increased to 67% when FCD type 2 was found by histological examination in the cortical specimen, as demonstrated in a recent large multicentric study [33]. This dramatic improvement of outcome highlights the role of imaging in identifying such epileptogenic lesion preoperatively.
Accordingly, we found a high rate of seizure-free outcome in a previous cohort of patients operated on for intractable epilepsy due to FCD type 2. Based on a presurgical work-up including 18 F-FDG PET in all cases, we did not nd signi cant difference between MRI-positive and MRI-negative cases (Engel class I in 94% and 88% respectively) [4].
Further improvement should be expected with the PET/MR as suggested by the present series, including more successful surgery performed without SEEG. Notably, the better sensitivity of this hybrid imaging leads to revisit the proportion of truly non-lesional epilepsy patients that could be referred for surgery, questioning on the presence of a structural abnormality at the microscopic level.

LIMITATIONS
The rst limitation of this study is the relatively small size of population, due to the di culties for obtaining images in both modalities and su cient clinical information to perform meaningful correlations. However, the well-localised EZ and the high rate of patients with favorable outcome after surgery enhance the con dence in our results. Secondly, the PET from the PET/MR provided optimised images compared to the previous PET, which were performed on different cameras with various spatial resolution and reconstruction methods. Nevertheless, it cannot explain the increase of sensitivity on its own, since a better co-registration was noted as frequently as a better quality of images by the two readers. In addition, the structural lesions newly detected on MRI cannot be attributed to a better quality of MRI or PET images independently but are rather due to the improvement of the fusion of both. Finally, the delay between the two examinations may play or role for the discrepancy, however we did not nd any relationship between this delay and the improvement provided by the PET/MR. Furthermore, there was no major clinical change in terms of seizure frequency, AEDs and delay between the last seizure and the 18 F-FDG injection. It should be emphasized that optimal comparative studies are rare, other studies based on consecutive examinations having also some limitations related to the different delay after the 18 F-FDG injection. Furthermore, the main limitations in previous studies consisted in the variable (and sometimes weak) assessment of the clinical usefulness, that we believe having demonstrated in our study.

Conclusion
We demonstrated the clinical utility of the PET/MR in a series of patients investigated for focal drug-resistant epilepsy, in whom the simultaneous acquisition of the PET and MRI improved the detection of epileptogenic lesions, allowing a higher proportion of successful surgery and simplifying the presurgical work-up, even in the more complex cases.
Combining "the best of two worlds", we postulate that these two imaging modalities acquired in a single session on a hybrid system would be helpful for the pre-surgical evaluation of epileptic patients and recommended in case of previous negative imaging ndings.

AKNOWLEDGEMENTS
The authors thank Claude Comtat for his methodological advices and all the SHFJ team, especially Philippe Gervais, Thierry Le Kieffre, and Christine Baron for carrying out the PET and PET/MR. They also thank Dr Catherine Chiron who performed the PET in children and the Sainte-Anne's team (Elisabeth Landré and Bertrand Devaux) for the patient care.

COMPLIANCE WITH ETHICAL STANDARDS
Funding: This work was performed on a platform member of France Life Imaging network (grant ANR-11-INBS-0006).
Con icts of interest/Competing interests: Anthime Flaus declares that he has no con ict of interest; Charles Mellerio Flaus declares that he has no con ict of interest; Sebastian Rodrigo declares that he has no con ict of interest; Vincent Brulon declares that he has no con ict of interest: Vincent Lebon declares that he has no con ict of interest; Francine Chassoux declares that she has no con ict of interest.