Breath-Hold-Triggered BOLD fMRI in Drug-Resistant Nonlesional Focal Epilepsy—A Pilot Study

This single-center study was conducted at the Department of Neurology and Epileptology, University Hospital Tübingen, between November 2021 and May 2022. Participants were prospectively and consecutively recruited from all individuals undergoing presurgical evaluation. Inclusion criteria were: i) age 18–70 years; ii) diagnosis of drug-resistant focal epilepsy according to the criteria of the Task Force of the International League Against Epilepsy (ILAE) Commission on Therapeutic Strategies [1]; iii) nonlesional focal epilepsy, i.e., prior structural imaging according to the HARNESS-MRI protocol that failed to detect an epileptogenic lesion and iv) case review at the interdisciplinary epilepsy surgery conference. Exclusion criteria were: i) inability to give informed consent; ii) any history of psychogenic nonepileptic seizures; iii) any history of primary generalized onset seizures; iv) pregnancy; v) any absolute or relative contraindications against MRI, including but not limited to presence of implanted devices, prostheses, surgical clips, wires, sutures or mesh, dental implants or retainers, or any foreign metallic body and vi) any condition that may prevent the participant from lying down for an extended period (e.g., pain, claustrophobia or pulmonary disease). Overall, 24 individuals were screened, and 10 individuals agreed to participate in the study.

Electroclinical data were obtained from epilepsy surgery conference reports and medical chart review. All available data were considered within the multidimensional concept of the epileptogenic zone by Lüders et al. [24]. The irritative zone, i.e., the area of cortex that generates interictal spikes, and the seizure-onset zone, i.e., the area of the cortex that initiates clinical seizures, were respectively defined with routine scalp EEG, video EEG, magnetoencephalography (MEG) or stereo electroencephalography when available. The symptomatogenic zone, i.e., the area of cortex that produces the initial ictal signs, was defined by consideration of lateralizing and localizing seizure semiology provided by the patient history or video EEG observation. The functional deficit zone, i.e., the cortical area that is functioning abnormally in the interictal period, was defined by neurological examination, neuropsychological examination, and functional imaging (18-FDG-PET-MRI) where available. Interictal SPECT was not done. Proposed localization of the epileptogenic zone was determined via blinded case review by an epileptologist who was not provided with the results of bh-fMRI before the localization hypothesis was established. The epileptogenic zone was defined as “area of cortex that is necessary and sufficient for initiating seizures and whose removal (or disconnection) is necessary for complete abolition of seizures” [25], with an additional emphasis on ictal EEG and especially confirmation by stereo electroencephalography if available [26]. This theoretical cortical zone is a primarily surgically motivated abstract concept that cannot be directly visualized on imaging, including bh-fMRI, but which represents the integrated diagnostic outcome of the presurgical evaluation. Thus, any correlation between focal abnormalities seen on bh-fMRI and this localization hypothesis would be potentially useful in a clinical setting.

All imaging data were acquired on a 3 T Scanner (Magnetom Prisma fit, Siemens, Erlangen, Germany) using a standard 20 channel head coil. The imaging protocol consisted of a T2* weighted EPI sequence with the following imaging parameters: TR 3000 ms, TE 36 ms, matrix 96 * 96, slice thickness 3 mm, 40 slices, FOV 245 × 245 mm, voxel size 2.6 × 2.6 × 3 mm. The paradigm consisted of 7 cycles of 9 s breath-hold in expiration followed by 60 s of normal breathing, starting with 1min of normal breathing. During the fMRI acquisition, compliance was checked via monitoring breathing using the built-in physiological monitoring capability of the scanner. The instructions were presented visually using a wall-mounted display and a coil-mounted mirror. A test run, where the participant was placed on the scanner table but outside of the gantry, preceded the final measurement to ensure that the participant understood the visually presented task. All instructions were scanner-triggered and software-generated (Presentation 20.1, Neurobehavioral Systems, Berkeley, CA, USA).

The MRI data processing was performed on a dedicated workstation. Data were preprocessed using the software package statistical parametric mapping SPM12 (The Wellcome Dept. of Imaging Neuroscience, London, UK; www.​fil.​ion.​ucl.​ac.​uk/​spm). After conversion into nifti file format, slice timing correction was applied to the functional data to compensate for differences in image acquisition times. The data were then realigned to remove effects caused by patient movements. After unwarping to remove geometric distortions created in regions of tissue-air interface, the functional data were co-registered with the anatomic T1 3D dataset. The data were spatially normalized to the standard Montreal Neuroimaging Institute (MNI) template, and all data were written in MNI space for subsequent evaluation.

Further analysis was done by a custom script programmed in MATLAB R2021a (The MathWorks Inc, Natick, MA, USA; http://​www.​mathworks.​com). To improve signal quality and reduce noise, the signal was averaged using 116 volumes of interest provided by the automatic anatomic labeling atlas AAL 1 [27]. After detrending to remove signal drift during the measurement, the signal of each breath-hold period was averaged resulting in on signal time curve per volume of interest (VOI). Breath-hold periods were excluded if the cerebellar signal curve, which is assumed to be independent of supratentorial pathology [15], showed no stimulus-related positive signal curves. Furthermore, the signal time course of the cerebellar signal can be used as an independent surrogate parameter for arterial CO₂ fluctuations, which serves to confirm physiological hemodynamic response (i.e., sufficient vasodilation after breath hold) [13, 28], including patient compliance with the breath-hold maneuvers.

Concordance between the electroclinical localization hypothesis and the results of breath-hold fMRI was defined as such: i) full concordance corresponded to correct lobar localization and lateralization; ii) partial concordance corresponded to correct lateralization; iii) discordance corresponded to incorrect localization and lateralization. We note that it is currently unclear if breath-hold fMRI offers sufficient spatial resolution to consistently enable sublobar localization. Instead, we chose the broader concepts of lobar localization and lateralization as our endpoints, as they are of key importance in the presurgical evaluation of epilepsy and represent information with an immediate impact on surgical decision-making and a strong correlation to postoperative outcome [29‐32]. This study was designed as a proof-of-concept study to establish the feasibility and safety of this novel method and was therefore not sufficiently powered for statistical analysis. Cohort size was restricted to 10 participants by the local ethics committee.