Authors :
Presenting Author: Felix Zahnert, MD – A. A. Martinos Center for Biomedical Imaging
Peter Hadar, MD, MS – Massachusetts General Hospital; Harvard Medical School
Kwok-Shing Chan, PhD – A. A. Martinos Center for Biomedical Imaging
Hansol Lee, PhD – A. A. Martinos Center for Biomedical Imaging
Yixin Ma, PhD – A. A. Martinos Center for Biomedical Imaging
Aneri Bhatt, BS – A. A. Martinos Center for Biomedical Imaging
Yuting Chen, MSc – A. A. Martinos Center for Biomedical Imaging
Julianna Gerold, Bsc – A. A. Martinos Center for Biomedical Imaging
Sydney Cash, MD, PhD – Massachusetts General Hospital
Steven Stufflebeam, MD – A. A. Martinos Center for Biomedical Imaging
Lawrence Wald, PhD – A. A. Martinos Center for Biomedical Imaging
Susie Huang, MD, PhD – A. A. Martinos Center for Biomedical Imaging
Rationale:
Despite extensive multimodal imaging, a considerable proportion of presurgical patients with epilepsy remain non-lesional. These patients are less likely to achieve postoperative seizure freedom. It was our aim to evaluate the benefit of adjunctive microstructural imaging, including modelling of cell body signal fractions and radii, for the detection of subtle epileptogenic lesions in a presurgical setting.
Methods:
Patients with focal epilepsy who were scheduled for intracranial EEG monitoring (SEEG) at the Massachusetts General Hospital during the study period (March-July 2025; recruiting until 03/2026) were included prospectively. Magnetic resonance imaging was conducted using the ultra-high gradient Connectome 2.0 scanner. Diffusion-weighted images were acquired with an SMS-EPI sequence (8 shells, maximum b-value of 6000 s/mm²). Healthy control participants who had been scanned with the same sequence were included. The soma and neurite density imaging (SANDI) model was fit to the diffusion data. SANDI metrics were z-scored and Mahalanobis distances (D²) were computed using bilateral control data. Maps were projected back to native space for evaluation in the context of anatomical, PET and SEEG data. Spatial correlations of multimodal maps were investigated using a spin permutation framework.
Results:
Six patients (four female, age 21-61 [M=30]) and 33 healthy controls (22 female, age 19-69 [M=32]) were included. SEEG had been conducted in three patients, and 2/6 were MRI-negative (one with SEEG). FDG-PET showed focal hypometabolism in one MRI-negative patient. On clinical MRI, detected lesions comprised a left grey matter heterotopia, a left temporal encephalocele, right mesial temporal sclerosis, and a subtle bilateral occipital polymicrogyria.
Clusters of abnormal microstructure co-localized with electrode contacts involved at seizure onset in 2/3 patients who had received SEEG. In one of these patients, regional abnormality scores (D²) of microstructure were negatively correlated with FDG uptake, and this spatial correlation was specific to a right temporal seizure onset zone (SOZ). Altered microstructural metrics were the only imaging findings indicating potential right temporal involvement prior to SEEG implantation in this patient (Figure 1b).
In the remaining cases, clear microstructural abnormalities matching electroclinical focus hypotheses were detected, regardless of the absence (n=1) or presence (n=2) of clinical imaging abnormalities (Figures 1-2). The ipsilateral Pulvinar and Anterior Nucleus of the Thalamus, both targets for neuromodulation, showed abnormalities in 2/6 cases. Lower FDG-uptake and increased D² spatially converged in the likely SOZ in 2/6 cases.
Conclusions:
Microstructural modelling seems sensitive to signatures of epileptogenic tissue in focal epilepsies. Correlations of microstructural abnormalities with FDG uptake seem to increase their spatial specificity to the region of seizure onset in some cases, even in the absence of clinically visible hypometabolisms.
Funding: Felix Zahnert received funding from the Otfrid Foerster Grant of the German chapter of the ILAE (DGfE) and from the Clinician Scientist Program (SUCCESS) of the University of Marburg