Interpretable surface-based detection of focal cortical dysplasias: a MELD study

Introduction: One outstanding challenge for machine learning in diagnostic biomedical imaging is algorithm interpretability. A key application is the identification of subtle epileptogenic focal cortical dysplasias (FCDs) from structural MRI. FCDs are difficult to visualise on structural MRI but are often amenable to surgical resection. We aimed to develop an open-source, interpretable, surface-based machine-learning algorithm to automatically identify FCDs on heterogeneous structural MRI data from epilepsy surgery centres worldwide. Methods: The Multi-centre Epilepsy Lesion Detection (MELD) Project collated and harmonised a retrospective MRI cohort of 1015 participants, 618 patients with focal FCD-related epilepsy and 397 controls, from 22 epilepsy centres worldwide. We created a neural network for FCD detection based on 33 surface-based features. The network was trained and cross-validated on 50% of the total cohort and tested on the remaining 50% as well as on 2 independent test sites. Multidimensional feature analysis and integrated gradient saliencies were used to interrogate network performance. Results: Our pipeline outputs individual patient reports, which identify the location of predicted lesions, alongside their imaging features and relative saliency to the classifier. Overall, after including a border-zone around lesions, the developed MELD FCD surface-based algorithm had a sensitivity of 67% and a specificity of 54% on the withheld test cohort, and a sensitivity of 85% on a restricted subcohort of seizure free patients with FCD type IIB who had T1 and FLAIR MRI data. Conclusions: This multicentre, multinational study with open access protocols and code has developed a robust and interpretable machine-learning algorithm for automated detection of focal cortical dysplasias, giving physicians greater confidence in the identification of subtle MRI lesions.

Factors influencing the detection of treatable epileptogenic lesions on MRI. A randomized prospective study

Background To prospectively analyze factors associated with detecting epileptogenic lesions on MRI within the work-sharing process of neurologists, epileptologists, radiologists and neuroradiologists. Methods We assembled four sets of six MRI scans, each set representing five typical epileptogenic lesions (hippocampal sclerosis or limbic encephalitis; focal cortical dysplasias; periventricular nodular or other heterotopias; long-term epilepsy associated tumors; gliotic scar, hemosiderin or cavernoma), and non - lesional epilepsy. At professional conferences, we invited neurologists, epileptologists, radiologists, and neuroradiologists to read two out of four MRI sets, one of which was presented with a clinical focus hypothesis. Participants were randomly assigned to MRI sets. Effects of examiners’ specialty, duration of training and professional experience on detection rate of epileptogenic lesions were investigated. Results Fourty-eight neurologists, 22 epileptologists, 20 radiologists and 21 neuroradiologists read 1323 MRI scans. Overall, 613 of 1101 (55.7%) epileptogenic lesions were detected. Long-term epilepsy associated tumors (182/221, 82.4%) were found more frequently than gliotic scar, hemosiderin or cavernoma (157/220, 71.4%), hippocampal sclerosis or limbic encephalitis (141/220, 64.1%), nodular heterotopia (68/220, 30.9%) and focal cortical dysplasias (65/220, 29.5%, p < 0.001). Provision of a focus hypothesis improved the detection of hippocampal sclerosis or limbic encephalitis (86/110, 78.2% vs 55/110, 50%, p < 0.001) and focal cortical dysplasias (40/110, 36.4% vs 25/110, 22.7%, p = 0.037). Neuroradiologists and epileptologists were more likely than radiologists and neurologists to be amongst the most successful readers. In multivariable analysis, type of epileptogenic lesion, specialty of MRI reader, and provision of focus hypothesis predicted correct identification of epileptogenic lesions. Conclusions Epileptogenic lesions are often not recognized on MRI even by expert readers. Their detection can be improved by providing a focus hypothesis. These results stress the need for training in the MRI characteristics of epilepsy - specific pathology, and, most importantly, interdisciplinary communication between neurologists/epileptologists and (neuro)radiologists to improve detection of epileptogenic lesions.

Morphological and Calcium Signaling Alterations of Neuroglial Cells in Cerebellar Cortical Dysplasia Induced by Carmustine

Cortical dysplasias are alterations in the organization of the layers of the brain cortex due to problems in neuronal migration during development. The neuronal component has been widely studied in experimental models of cortical dysplasias. In contrast, little is known about how glia are affected. In the cerebellum, Bergmann glia (BG) are essential for neuronal migration during development, and in adult they mediate the control of fine movements through glutamatergic transmission. The aim of this study was to characterize the morphology and intracellular calcium dynamics of BG and astrocytes from mouse cerebellum and their modifications in a model of cortical dysplasia induced by carmustine (BCNU). Carmustine-treated mice were affected in their motor coordination and balance. Cerebellar dysplasias and heterotopias were more frequently found in lobule X. Morphology of BG cells and astrocytes was affected, as were their spontaneous [Ca2+]i transients in slice preparation and in vitro.

Surgical treatment of epilepsy at the Russian Children’s Clinical Hospital: substrates and imaging aspects

Introduction. MRI revealing the epileptogenic foci is the central part of the presurgical evaluation of children with focal-resistant epilepsy. Besides that, MRI seems to be a reliable instrument of treatment optimization and good prognosis. Aim. To identify epileptogenic lesions in children with resistant focal epilepsy by high resolution (HR) MRI and comparing obtained data with postsurgical morphology. Materials and methods. We assessed the results of neuroimaging, EEG, clinical neurological status, and postsurgical morphological data in 65 children with focal epilepsy who had undergone antiepileptic surgery in Russian State Children Hospital, neurosurgery department (Moscow, Russia) from 2016 to 2020. Results. The broad spectrum of epileptogenic lesions was revealed by HR MRI, including focal and diffuse cortical dysplasias (44.6%), cortical post hypoxic gliosis (27.7%), glial tumors (7.7%), Rasmussen encephalitis (10.8%), tuberous sclerosis, and Sturge-Weber syndrome (3%), small angiodysplasias (4.6%) and other. Focal cerebral dysplasia had an obvious superiority as the main etiological factor in focal epilepsy and was approved in 33.8% of all patients. In all cases, epilepsy is characterized by frequent and resistant seizures, inflicting motor arrest, and psycho-emotional deterioration. Conclusion. HR MRI and multidisciplinary investigation have to be unchangeable standards in the presurgical evaluation of children with focal epilepsy. This approach’s effectiveness has reassuring confirmation by high level (95.2%) of total coincidence MRI and morphological results.

The Putative Role of mTOR Inhibitors in Non-tuberous Sclerosis Complex-Related Epilepsy

Epilepsy affects ~5 out of every 10,000 children per year. Up to one-third of these children have medically refractory epilepsy, with limited to no options for improved seizure control. mTOR, a ubiquitous 289 kDa serine/threonine kinase in the phosphatidylinositol 3-kinase (PI3K)-related kinases (PIKK) family, is dysregulated in a number of human diseases, including tuberous sclerosis complex (TSC) and epilepsy. In cell models of epilepsy and TSC, rapamycin, an mTOR inhibitor, has been shown to decrease seizure frequency and duration, and positively affect cell growth and morphology. Rapamycin has also been shown to prevent or improve epilepsy and prolong survival in animal models of TSC. To date, clinical studies looking at the effects of mTOR inhibitors on the reduction of seizures have mainly focused on patients with TSC. Everolimus (Novartis Pharmaceuticals), a chemically modified rapamycin derivative, has been shown to reduce seizure frequency with reasonable safety and tolerability. Mutations in mTOR or the mTOR pathway have been found in hemimegalencephaly (HME) and focal cortical dysplasias (FCDs), both of which are highly correlated with medically refractory epilepsy. Given the evidence to date, a logical next step is to investigate the role of mTOR inhibitors in the treatment of children with medically refractory non-TSC epilepsy, particularly those children who have also failed resective surgery.

MRI diagnosis of cortical dysplasia in the immature brain

Introduction. Cortical dysplasias (CDs) encompass a wide variety of disorders that in most cases lead to epilepsy, especially in infants and young children. MRI diagnosis of CDs is a major part of presurgical examination of pediatric patients with resistant focal epilepsy.Aim. To identify MR markers of CD in the immature brain and develop an MRI protocol for early diagnosis of CDs.Materials and methods. Children aged <2 y.o. (total 128) diagnosed with focal epilepsy were examined over 2017-2019. All MRI scans were performed using the GE 3 T system (General Electric, USA) in the standard MR sequences including T2WI FSE, T1 SE, FLAIR, DWI, SWAN, and FSPGR BRAVO supported with anesthesiological assistance. Аll patients were divided into 3 groups according to the degree of brain maturity; of those, 28 patients had MR signs of CD.Results. The rate of detection of small-size cortical malformations, such as nodular heterotopies or focal cortical dysplasias was significantly higher in groups of patients whose brains (according to MR images) were at the infantile or adult phases of myelination. In children with the isointensive phase myelination, only large cortical dysplasias could be identified. In the first phase, the focal malformations had low amplitude signals in T2-weighted images and high amplitude signals in T1, unlike those in adult patients. In the isointensive phase, the quality of visualization was significantly reduced and provided poor diagnostic information.Conclusion. The results confirm the diagnostic significance of early (before age of 5 months) MRI testing in cases with suspected CD-associated focal epilepsy. However, at the period between 5 and 12 months of age, MR imaging was ineffective for CD diagnosing. Later, in the period from 12 to 15 months, the MRI ability to identify the CDs gradually increased. We consider the standard T2 weighted images with high TR values, the most effective MR modality for diagnosing CDs in young children.

Planning sEEG implantation using automated lesion detection: retrospective feasibility study

ObjectiveA retrospective, cross-sectional study to evaluate the feasibility and potential benefits of incorporating deep-learning on structural MRI into planning stereoelectroencephalography (sEEG) implantation in paediatric patients with diagnostically complex drug-resistant epilepsy. This study aims to assess the degree of co-localisation between automated lesion detection and the seizure onset zone (SOZ) as assessed by sEEG.MethodsA neural network classifier was applied to cortical features from MRI data from three cohorts. 1) The network was trained and cross-validated using 34 patients with visible focal cortical dysplasias (FCDs). 2) Specificity was assessed in 20 paediatric healthy controls. 3) Feasibility for incorporation into sEEG implantation plans was evaluated in 38 sEEG patients. Coordinates of sEEG contacts were coregistered with classifier-predicted lesions. sEEG contacts in seizure onset and irritative tissue were identified by clinical neurophysiologists. A distance of <10mm between SOZ contacts and classifier-predicted lesions was considered co-localisation.ResultsIn patients with radiologically-defined lesions, classifier sensitivity was 74% (25/34 lesions detected). No clusters were detected in the controls (specificity 100%). Of 34 sEEG patients, 21 patients had a focal cortical SOZ. Of these there was co-localisation between classifier output and SOZ contacts in 62%. The algorithm detected 7/8 histopathologically-confirmed FCDs (86%).ConclusionsThere was a high degree of co-localisation between automated lesion detection and sEEG. We have created a framework for incorporation of deep-learning based MRI lesion detection into sEEG implantation planning. Our findings demonstrate that automated MRI analysis could be used to plan optimal electrode trajectories.