Value of ultra-high field MRI in patients with suspected focal epilepsy and negative 3 T MRI (EpiUltraStudy): protocol for a prospective, longitudinal therapeutic study

This is a multicenter prospective, longitudinal, therapeutic study assessing the value of UHF MRI for lesion detection and therapeutic impact of this improved lesion detection, in patients with a suspected focal cause of drug-resistant epilepsy and unrevealing clinical 3 T MRI, who are candidates for resective epilepsy surgery (Fig. 1).

The diagnosis of focal epilepsy and a suspected region of interest (ROI) as the focal epileptogenic zone is based on thorough examinations including medical history with seizure semiology, a video-EEG, neuropsychological examination, and 3 T MRI according to a dedicated epilepsy protocol [15, 26], including morphometric analysis program (MAP18) postprocessing [27, 28]. In selected cases, a PET scan and/or subtraction ictal SPECT co-registered to MRI (SISCOM) and/or EEG triggered fMRI and/or MEG are performed. In a subset of these patients, with discordant results from non-invasive examinations, intracranial EEG recordings by implanted surface or intracerebral depth electrodes are necessary. Subsequently, a multidisciplinary team, consisting of neurosurgeons, neurologists, clinical neurophysiologists, neuropsychologists, and neuroradiologists, assesses eligibility for epilepsy surgery. All clinical 3 T MRI scans are reviewed by a radiologist specialized in epilepsy imaging, and are considered “negative” only after review of the scan in the multidisciplinary team, where the 3 T scan and other non-invasive examination are considered.

For the present study, we include pediatric patients (children, aged 12–17 years) and adults (aged 18 and above), with drug-resistant focal epilepsy (ongoing seizures despite at least two adequate and well-tolerated anti-epileptic drugs trials [29]), who are under evaluation for resective epilepsy surgery. All patients have had a clinical 3 T MRI without an identified explanatory epileptogenic lesion. Exclusion criteria are intellectual disability (IQ < 70), incapacity to sign informed consent, pregnancy, and criteria specific for UHF MRI (see Table 1).

MRI acquisition will be performed at Scannexus BV, a specialized UHF MRI research facility, using an actively shielded Siemens Magnetom “classic” 7 T with whole-body gradients and 32-channel NOVA head coil, and a Siemens Magnetom 9.4 T with head-gradient set and 31-channel receive/16-channel parallel RF transmit head coil.

Development of a clinically applicable 7 T and 9.4 T MRI scan protocol for detection of structural abnormalities is part of this study. We created a format for a proposed UHF scan protocol (see Table 2). All sequences will acquire images of the whole brain. This scan protocol allows for the use of normative control data, acquired by Haast et al. [30]. Two dielectric CaTiO₃ pads with demineralized water are used to improve the inhomogeneous MRI signal in the temporal lobes. As a preparation, we scanned some controls at 7 T to assess image quality (see Figs. 2 and 3).

Patients are included after obtaining written informed consent; subsequently, MRI will be performed. In a first run, both 3 T and UHF images will be independently evaluated by two neuroradiologists experienced in epilepsy-imaging, and a neurosurgeon or neurologist specialized in epilepsy surgery, without prior knowledge of previously obtained non-invasive examinations and the potential epileptogenic zone (= ROI). In a second run, first the 3 T, subsequently, the UHF MRI scans will be evaluated with prior knowledge of the ROI by the same investigators. The images will be reviewed for image quality and potential epileptogenic lesions using a structured UHF MRI scoring form, allowing objective comparison between observers and scans. Lesion size/volume, number, lateralization, localization, shape, and signal intensity detected on the different sequences are marked. The mesiotemporal structures (hippocampus, amygdala) are evaluated for abnormalities and asymmetries in volume and signal intensity. Evaluation for FCDs is done by screening for features of blurring cortical-subcortical junction/cortical thickening/transmantle sign/hyperintensity of grey and white matter/abnormal gyral-sulcal pattern/segmental or lobar hypoplasia/hypoplasia with loss of regional white matter volume. Additionally, we evaluate features of an intracortical black line as a possible marker of a FCD and magnetic susceptibility effects due to tissue irregularities [31]. Additionally, susceptibility effects are used to detect the presence of abnormal vasculature adjacent to areas of cortical dysplasia or polymicrogyria [32]. A diffusion sequence (by acquiring diffusion tensor imaging, DTI) is firstly performed to aid the detection of intra-cortical radial and tangential diffusion abnormalities, suggestive for FCD types 1a and 1b, and secondly to detect a difference in intra-cortical radial and tangential diffusion between FCD types 1a and 1b. Using post-processing, additional FA and ADC maps will be generated. Abnormalities in blood-oxygenation level fluctuations are detected using resting-state fMRI and arterial spin labeling (ASL). Furthermore, we will make use of voxel-based morphometry (VBM) structural post-processing technique. The aim of this technique is to quantitatively analyze morphology and signal intensity on both 3 T and UHF scans by using algorithms for automated FCD detection as qualitative visual inspection of the MRI will fail to detect tiny aberrances [33].

Results of the UHF MRI scans will be presented and discussed in a monthly multidisciplinary epilepsy surgery conference. We will evaluate the association between the presence and absence of an abnormal finding, with information of other non-invasive examinations. Additionally, we will assess the potential improvement in detection rate after adding prior knowledge of the ROI to the results of reviewing, by comparing the first with the second run, along with intra-observer and inter-observer variability.

After obtaining UHF MRI results, patients can undergo either (1) resective epilepsy surgery, (2) implantation of intracerebral depth electrodes for intracranial EEG recordings (S-EEG), (3) S-EEG guided radiofrequency thermo coagulation (RFTC), or (4) proceed to palliative treatment options (e.g., vagus nerve stimulation). Surgical resection is the first choice due to its efficacy and will be performed when possible, with regard to lesion location, extent, and expected lesion etiology. Depth electrode implantation will be used when additional information on the possible epileptogenic zone is required. S-EEG RFTC will be performed when resective surgery is not recommended, for example, in case of deep-seated periventricular heterotopias. Palliative treatment options are explored when other curative treatments are excluded. Samples of surgically treated patients will be examined by routine histopathological analysis and classified according to ILAE guidelines [20]. In case of resective epilepsy surgery, UHF MRI–based diagnosis will be compared to histopathological diagnosis. Postoperative seizure outcome following ILAE and Engel classification [34] will be assessed by the treating neurologist as standard care minimally 1 year after surgery.

The primary outcome parameter of this study is the number of patients with detected “epileptogenic” structural and functional brain abnormalities on 7 T and 9.4 T UHF neuroimaging, the so-called “diagnostic gain” in patients with 3 T MRI-negative drug-resistant focal epilepsy. The second primary outcome parameter is postoperative seizure outcome (ILAE/Engel class at 12 months) in operated UHF MRI-positive versus MRI-negative patients “i.e., therapeutic gain.”

As secondary outcome parameters, detection rate of UHF resting-state fMRI for cortical abnormalities in spontaneous blood oxygenation level-dependent (BOLD) signal fluctuations as a marker of an epileptogenic lesion, concordance of UHF MRI imaging with histopathological diagnosis of surgical specimen, intra- and inter-observer agreement for UHF MRI evaluation, and association between ROIs in the non- or semi-invasive work-up modalities and UHF MRI data will be assessed.

Recruitment of eligible study participants will take place at (1) both locations of the Academic Center for Epileptology (ACE) Kempenhaeghe/Maastricht University Medical Center + (MUMC +), Heeze/Maastricht, The Netherlands, and (2) the RWTH Aachen University Hospital, Section of Epileptology and Department of Neurosurgery Aachen, Germany. Based on feasibility and number of patients, we expect 30 patients from Kempenhaeghe/MUMC + , among which eight children 12 to 18 years, to qualify for this study annually. Recruitment from Aachen is expected to be 10 patients annually. Taking into account, a probable informed consent rate of 75% results in an annual inclusion of 30 patients.

The study participants’ timeline is outlined below; an overview is also given in Fig. 1.

Data obtained by UHF MRI (part 4. Measurements) will be discussed in the multidisciplinary epilepsy surgery workgroup and, when consensus is reached about possible epileptogenicity of a novel lesion, stereo-EEG, respective, or minimally invasive surgery will be considered.

Results from the epilepsy surgery workgroup, surgery, histopathology, and postoperative seizure outcome at 1 year will be recorded from digital patient files. If patients do not undergo epilepsy surgery, conclusions, and recommendations from the workgroup, next to possible other treatment suggestions and seizure follow-up will be recorded.

The primary study outcome is the lesion detection rate of UHF MRI. This outcome will be estimated without formal null-hypothesis testing. We previously conducted a study in which we included 19 adult patients where MEG directed to an epileptogenic lesion [35]. All patients were 3 T negative; on 7 T, six patients showed possible epileptogenic lesions. However, as there is no similar research based on other examination modalities (SPECT, PET, etc.) available for UHF MRI imaging, no data to perform a sample size calculation is available [23, 24]. Therefore, we take a pragmatic approach to the determination of sample size. Rules of thumb for linear regression analysis state that at least 10 observations are needed for each variable that is entered in the regression model. In addition to univariate analysis, we aim to perform multivariable analysis in which we will correct for up to five potential confounding variables. Hence, we will need to include at least 60 patients for this study. With a sample of this size, the margin of error (i.e., half-width of the 95% confidence interval) of the detection rate will likely be below 0.20.

The clinical trial sponsor (MUMC +) and principal investigators (OEMGS and AJC) have overall responsibility for the execution of this study including safety. Individual investigators will be responsible for reporting AEs and serious AEs (SAEs) to the principal investigators. SAEs are defined as AEs resulting in death, life-threatening events, prolonged hospital stays, or significant disability. Up till now, there are no reported (S)AEs associated with the clinical use of UHF MRI in literature. UHF MRI is, with regard to safety, highly comparable to standard MRI scanning at 3 T for clinical health care when safety precautions are taken into account [36, 37]. The MRI safety relevant inclusion/exclusion criteria for 7 T and 9.4 T are set by the local Safety Review Board (SRB), which acts as an expert advisory panel on behalf of the local Medical Ethical Committee/Institutional Review Board (IRB). The MRI-questionnaire determines whether a patient is eligible and safe to undergo the scan.

Claustrophobia can be a possible concern of UHF MRI, due to a longer magnet bore. However, this is accounted for by excluding patients with known claustrophobia. Also, all subjects already underwent a clinical 3 T MRI scan; any occurrence of adverse events during this scan would lead to exclusion of the subject. One important frequently reported side-effect at UHF is vertigo when moving in/out the bore, in some cases resulting in nausea or small involuntary eye motion, i.e., nystagmus [38]. However, this poses no health threat to patients [36]. In case patients experience uncontrollable nausea, scanning will be stopped immediately. All events or complaints from signing informed consent until 1 day after completion of the UHF MRI scan will be considered an adverse event, and are reviewed by the principal investigator to decide if there is a causal link and, when applicable, appropriate action will be undertaken. Follow-up in patients will continue until minimally 1 year after epilepsy surgery. Due to the nature of the investigated pathology (epilepsy), seizures and subsequent sequelae between completion of the UHF MRI scan and 1-year follow-up will not be considered an adverse event related to study procedures. SAEs will be reported through the web portal to ToetsingOnline to the IRB (METC azM/UM). Liability and subject study-related insurance are provided.

Baseline characteristics of the cohort will be described in detail as mean and standard deviation or median and interquartile range for continuous variables (depending on their normality of distribution) and as count and percentage for categorical variables. Missing values are unlikely to occur, but in case of incomplete MRI data, we will use stochastic regression imputation to facilitate the inclusion of all study participants in the analyses.

UHF MRI data, such as characteristics of and, if applicable, localization of abnormalities, will be summarized using descriptive statistics. Primary outcomes diagnostic and therapeutic gain will be expressed as point estimates with 95% confidence intervals. Explorative analysis is used for quantitative MRI measurements (diffusion, fMRI, ASL). Radiological assessment will be done independently, in case of different results consensus will be sought. Intra-observer and inter-observer agreement in assessing MRI images will be quantified using Cohen’s kappa for categorical MRI parameters, and intra-class correlation coefficient (ICC) for continuous MRI parameters. Associations between qualitative findings and histopathological abnormalities will be computed using Spearman rank correlation, point bi-serial correlation, or Fisher’s Exact test, depending on the distribution of the findings.

This study was classified as intermediate risk by the local data monitoring committee (Clinical Trial Center Maastricht (CTCM)). Monitoring visits by the CTCM include a review of consent and study procedures according to study protocols, source data and audit trail verification, and the review of (serious) AE reporting. Monitoring is independent and performed at least once a year and can be unannounced. Monitor evaluations are reported to the local IRB.

This research protocol has been approved by the local IRB (Medisch-ethische toetsingscommissie academisch ziekenhuis Maastricht/Maastricht University (METC azM/UM)) and has been assigned the following protocol ID: NL66929.068.18. Approval of the IRB of Kempenhaeghe Epilepsy Center (ID METC KH 19.14) and the RWTH Aachen University medical faculty was gained (CTC-A 20–302). This study has also been registered at the Netherlands National Trial Register (ID: NTR7536), which is acknowledged by the WHO and International Committee of Medical Journal Editors (ICMJE). Results will be recorded using audit trails to increase reproducibility. Study protocol and results will be submitted to peer-reviewed journals and presented at international conferences.

The added diagnostic value of UHF, compared to lower field strengths, has been demonstrated for distinct pathologies, such as vascular malformations [42], hippocampal sclerosis [43], brain tumors [44], and degenerative brain diseases [45]. Over the last years, some studies demonstrating the diagnostic gain of 7 T MRI over clinical 1.5 T or 3 T MRI in patients with drug-resistant focal epilepsy have been published [23, 24, 46‐48]. We recently published a systematic review discussing UHF MRI in human epilepsy [49].

UHF MRI offers a higher signal/contrast-to-noise ratio, enabling higher spatial resolution with better depiction of micro-anatomical structures, therefore improving detection rate of epileptogenic lesions [50]. Studies comparing diagnostic yield of 3 T versus 1.5 T MRI in patients with focal epilepsy have shown this [9, 51]. An additional factor contributing to the diagnostic gain might derive from the use of a dedicated protocol. Interestingly, very recently, a 7-T epilepsy task force published a consensus paper with recommendations on the use of 7 T in clinical practice, based on multicenter and multinational experience with 7 T [52]. The proposed scan protocol for this study includes the considered “minimum scan requirements” and recommendations of this task force, with the addition of what we consider promising sequences, also based on findings of our systematic literature review [15, 49, 52]. Due to the high spatial resolution and sensitivity to the magnetic susceptibility properties of tissues, GRE (T2*) imaging allows better evaluation of the different components of the cortex [52]. Furthermore, GRE contrast increases with increasing magnetic field, hence the particular interest of its value in this study. FLAIR imaging, emphasizing signal changes of the cortex and along the cortical-white matter interface, and at 7 T might uncover even slight signal hyperintensities not otherwise visible [49, 52]. Several MRI features characterize FCD, but with great variety in conspicuity [53]. Especially in mild malformations of cortical development and FCD type I, changes can be subtle and indistinguishable from signal averaging and partial volume effects [54]. Very high isotropic resolution diffusion sequence (close to 1 mm isotropic) achievable in the whole human brain at UHF[55] has been shown to be sensitive to the intra-cortical radial and tangential diffusion direction [56]. This may enhance detectability of FCD type I, and differential abnormalities in radial and tangential diffusion could help stratification into FCD type Ia and type Ib.

Besides structural imaging, growing evidence demonstrates that connectomics, such as functional connectivity, could serve as a marker for pathological functions and networks, especially in a network-disease like epilepsy [57]. Other studies have shown early results that functional connectivity is homotopically correlated with resting-state default-mode networks, visualized by resting-state fMRI [58]. These imaging techniques can couple functional abnormalities to structural lesions, which has been demonstrated in a study by Gupta et al., where abnormalities in spontaneous blood oxygenation level-dependent fluctuations in the perilesional area of FCDs were found [59]. Another recent study confirmed this interesting new direction [60], while functional ASL allows additional measurement of baseline cerebral blood flow [61]. It has been shown that vascular abnormalities are associated with the underlying dysplastic cortex and even (pre)ictal neurovascular and metabolic coupling surrounding a seizure focus [32, 46, 62]. Besides further enhancing diffusion and susceptibility imaging at UHF, abnormalities in BOLD fMRI has been shown to correlate with epileptic foci, especially FCD [59, 63, 64]. Inclusion of these “non-structural” sequences in our scan protocol aims at exploring the additional value of these sequences.

The next step to improve diagnostic yield is the use of a structural post-processing technique, voxel-based morphometry, with its own specific diagnostic properties [65, 66]. The add-on application of 9.4 T MRI in the EpiUltraStudy is intended to provide the next step in examining the possible role of 9.4 T MRI for lesion detection in MRI-negative patients with epilepsy. Since previous 7 T studies have shown a ± 30% increase in detection rate [23, 24, 47], applying 9.4 T MRI could further improve signal-to-noise by up to a factor of two over 7 T[67], and should lead to an additional increase in detection rate.

However, improving imaging only by using more modern MRI-equipment might not be sufficient. An “a priori” hypothesis on the probable location of the epileptogenic zone is crucial to zoom in around a predefined ROI in order to increase detection rate of subtle and potential epileptogenic lesions [11, 12]. This a priori hypothesis can be formulated by application of a variety of non-invasive examinations, for example source localization by pre-surgical MEG-guided 7 T MRI analysis, which results in a clear gain in number of detected abnormalities [35]. PET, SPECT, and EEG-fMRI are other non-invasive modalities and are applied next to clinical semiology, video-EEG, and neuropsychological tests in the pre-surgical work-up of epilepsy surgery candidates to determine a plausible lateralization and/or localization of the epileptogenic focus. If the results of the aforementioned modalities lead to a robust hypothesis of the seizure-onset zone, a ROI can be defined to be examined using UHF MRI.

The EpiUltraStudy will contribute to improve preoperative counseling of the patient with significant and direct benefits for epilepsy surgical candidates. The epileptogenic character of novel detected abnormalities needs to be confirmed in selected cases, after which these patients will be offered epilepsy surgery. On the one hand, it enables the neurologist and neurosurgeon to inform patients on the possible cause of their epilepsy. On the other hand, inclusion of UHF MRI in preoperative workup will lead to a more targeted operation planning, a higher chance of complete lesionectomy, and potentially a minimally invasive surgical treatment. Altogether, the acquired information will lead to an increased understanding of patho(physio)logy in epilepsy and probably will improve the quality of life of these patients. Especially in children, the achievement of seizure freedom, and consequently tapering of the anti-epileptic drugs, leads to improvements in different domains like cognitive function, development, and quality of life [68, 69]. Besides, as children pose the largest group for congenital abnormalities leading to epilepsy, it is expected that the largest part of structural abnormalities, detected on UHF MRI, will be found in children and patients with childhood onset of their epilepsy.

Sample size calculation for a study assessing diagnostic gain and improvement of postoperative seizure outcome is not trivial as there is no comparable research based on other modalities (SPECT, PET, etc.). Thus, no power calculation based on prior studies can be performed. Therefore, we took a pragmatic approach with sample size calculation based on observations needed for each variable entered in a regression model. Final study power has to be established.

The main technical challenges for UHF MRI acquisition are inhomogeneities in the main magnetic field (B₀) and radiofrequency transmit field (B₁) and their stronger interaction with any implants. The current proposed scan protocol might lead to increased sensitivity to motion artifacts due to longer scanning times in the chosen highest resolution protocols, as the patient has to lie still for a longer period of time. Our total scan time of ± 70 min is slightly above acquisition times of other 7 T studies, where acquisition time was not an issue [23, 31, 35, 47]. Additionally, clinical 3 T MRI is performed at an earlier stage than UHF MRI in the epilepsy surgery workup. The results of additional examinations possibly lead to novel insights on the ROI, influencing reviewing of UHF MRI scans. A potential bias may occur because not all patients receive all other non-invasive examinations, but at the clinician’s discretion. Furthermore, inexperience in clinical reading/interpretation of 7 T and 9.4 T images can be an issue, as reviewers might experience a reviewing learning curve. Also, UHF might produce artifacts or geometric disturbances not seen on conventional MRI, potentially resulting in detection of non-relevant lesions (false positives). To address this issue, several UHF pilot scans will be performed to gain experience at reviewing UHF MRI. This includes patients with positive 3 T MRI for a small lesion, along with non-epileptic volunteers.