Investigating Epileptic and Functional Networks in Patients with Drug Resistant Focal Epilepsy

National Institute Of Neurological Disorders And Stroke

Investigators

Linked publications & trials

Abstract

During the past year, we have made significant progress on this project. Despite the challenges of COVID-related restrictions on our activity, we were able to continue with data collection in patients undergoing clinical presurgical evaluations, most notably functional and structural neuroimaging, neuropsychological testing, and epilepsy surgery. Much of the progress made in the past year has focused on identification of subtle epileptogenic lesions using structural magnetic resonance imaging (MRI). Over the past several years, in collaboration with Souheil Inati, formerly of the Functional MRI Core Facility in NIMH, we developed a novel MRI post-processing method to aid with the detection of focal cortical dysplasias (FCDs). Identification of epileptogenic lesions has been shown to significantly improve seizure outcomes in patients undergoing epilepsy surgery. Epileptogenic lesions are often readily apparent on visual inspection. However, a significant proportion of patients with apparently nonlesional epilepsy undergoing surgery are found on pathologic examination to have focal cortical dysplasia. Several post-processing methods have been previously reported to aid in the detection of these often subtle lesions, but none have been made readily publicly available, and most of these require large numbers of healthy volunteers and/or patients as training data. We developed a novel approach based on a set of multiscale imaging filters adapted from the computer vision literature. We applied them to our multicontrast structural MR images, and used a machine learning approach based on the resulting features to create a representation of both normal and dysplastic cortex. Using this approach, we were able to identify FCDs with similar efficacy to previously described methods with a considerably smaller healthy volunteer and patient training set (Snyder et al. 2021). We are now working to provide a publicly available pipeline so that other centers can implement this approach, and are in the process of setting up collaborations to test the validity of our approach using data from other centers. In a related project, we have used the same set of image filters to develop a template-free brain tissue segmentation algorithm. Although there are numerous publicly available brain tissue segmentation procedures designed for use in structurally normal brains, many of these approaches fail to accurately segment brains with significant structural abnormalities. Accurate brain segmentation is important in our research because it serves as the basis for neurophysiologic source localization algorithms and for semi-automated creation of brain resection masks. In our patients, particularly in those with large regions of encephalomalacia, we have found that our approach appears to be more accurate than several commonly used software packages. A manuscript is currently in preparation describing this approach and its accuracy for brain tissue classification as well as for semi-automated creation of brain resection masks (Hussein RK, Abdollahi S, Inati SJ, Inati SK). While intermittent seizures are the most obvious manifestation of epilepsy, interictal functional disturbances are also well described, particularly in the region of seizure onsets. This has been observed in the form of interictal epileptiform activity (IEA) using electroencephalography (EEG) and magnetoencephalography (MEG), hypometabolism using fluorodeoxyglucose positron emission tomography (FDG-PET) imaging, and at times changes in perfusion and diffusion imaging using MRI. Therefore, in our cohort of patients we have been prospectively collecting arterial spin labeling (ASL) MRI in collaboration with Lalith Talagala in the NINDS NMR Center Core Facility. ASL MRI is a widely available, safe, noninvasive imaging modality used to study cerebral blood flow and perfusion. In our cohort of patients with temporal lobe epilepsy (TLE), we found that those with lesional epilepsy on MRI (almost all medial temporal sclerosis), show significant hypoperfusion in the medial temporal lobe, particularly in the parahippocampal gyrus, as well as the lateral temporal lobe, primarily driven by asymmetries in the anterior temporal neocortex, compared to both healthy volunteers and patients with non-lesional TLE. Non-lesional TLE patients demonstrated similar brain perfusion to healthy volunteers except for the presence of anterior temporal neocortical hypoperfusion. We are currently preparing a manuscript describing these findings (Rentzeperis R, Abdennadher M, Dembny K, Snyder K, Talagala L, Theodore WH, Zaghloul KA, Inati SK). The second goal of this project is related to exploring the relationship between propagation of epileptiform activity and underlying structural and functional networks in individual patients with epilepsy. Through our close ongoing collaboration with Kareem Zaghloul's lab in the NINDS Surgical Neurology Branch, we are exploring the relationship between two novel approaches to identifying the source of interictal and ictal epileptiform activity using invasive EEG recordings (Diamond et al., 2019, Diamond et al. 2021) and structural connectivity assessed using diffusion tensor imaging (DTI). We found that both structural connectivity and propagation of interictal epileptiform activity (IEA) are strongly distance dependent. However, in a significant number of patients, even when controlling for distance, propagation of IEA appears more likely to occur in structurally connected brain regions. A manuscript is also in preparation describing these findings (Yang BY, Diamond J, Goodyear A, Snyder K, Theodore WH, Zaghloul KA, Inati SK). As part of our ongoing epilepsy imaging study, we prospectively obtain language and resting state functional MRI (fMRI) data. Several of our close collaborators have made use this data over the past year to study both language and memory. Dr. Theodore's group compared the lateralization of language dominance obtained using resting state fMRI to the more traditionally used task-based approach. They found that although task-based and resting-state based language lateralization were concordant in most patients, language dominance is less lateralized on resting state compared to task-based fMRI (Rolinski et al. 2020). Leigh Sepeta, a collaborator from Children's National Medical Center, is investigating the use of language fMRI activations to predict memory outcomes following temporal lobe epilepsy surgery. She has found that functional activations in the both the mesial temporal and memory networks during language fMRI tasks appear to have utility in predicting memory outcomes following surgery. Her group is preparing a manuscript describing these findings. We have begun and continued several other collaborations involving both fMRI and DTI data in our patient cohort with groups within and outside of the NIH, and hope that these will be productive and described in next year's annual report.