Search

IntroductionOne 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.MethodsThe 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.ResultsOur 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.ConclusionsThis 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.HighlightsThis large, multi-centre, multi-scanner neuroimaging cohort captures the heterogeneity of histopathological subtypes and imaging features of patients with FCD.We developed a robust and interpretable surface-based algorithm which detects FCDs with a sensitivity of 67% and a specificity of 54%.The algorithm generates individual patient reports that “open the AI black-box” highlighting predicted lesion locations, alongside the imaging features and their relative saliency to the classifier.

PurposeChildren undergoing stereoelectroencephalography (SEEG)-guided epilepsy surgery represent a complex cohort. We aimed to determine whether the proportion of putative seizure onset zone (SOZ) contacts resected associates with seizure outcome in a cohort of children undergoing SEEG-guided resective epilepsy surgery.MethodsPatients who underwent SEEG-guided resective surgery over a six-year period were included. The proportion of SOZ contacts resected was determined by co-registration of pre- and post-operative imaging. Seizure outcomes were classified as seizure free (SF, Engel class I) or not seizure-free (NSF, Engel class II-IV) at last clinical follow-up.ResultsOf 94 patients undergoing SEEG, 29 underwent subsequent focal resection of whom 22 had sufficient imaging data to be included in the primary analysis (median age at surgery of 10 years, range 5-18). Fifteen (68.2%) were SF and 7 (31.8%) NSF at median follow-up of 19.5 months (range 12-46). On univariate analysis, histopathology, was the only significant factor associated with SF (pConclusionHistopathology is a significant predictor of surgical outcomes in children undergoing SEEG-guided resective epilepsy surgery. The percentage of SOZ contacts resected was not associated with SF. Factors such as spatial organisation of the epileptogenic zone, neurophysiological biomarkers and the prospective identification of pathological tissue may therefore play an important role.

Network control theory provides a framework by which neurophysiological dynamics of the brain can be modelled as a function of the structural connectome constructed from diffusion MRI. Average controllability describes the ability of a region to drive the brain to easy-to-reach neurophysiological states whilst modal controllability describes the ability of a region to drive the brain to difficult-to-reach states.In this study, we identify increases in mean average and modal controllability in children with drug-resistant epilepsy compared to healthy controls. Using simulations, we purport that these changes may be a result of increased thalamocortical connectivity. At the node level, we demonstrate decreased modal controllability in the thalamus and posterior cingulate regions. In those undergoing resective surgery, we also demonstrate increased modal controllability of the resected parcels, a finding specific to patients who were rendered seizure free following surgery.Changes in controllability are a manifestation of brain network dysfunction in epilepsy and may be a useful construct to understand the pathophysiology of this archetypical network disease. Understanding the mechanisms underlying these controllability changes may also facilitate the design of network-focussed interventions that seek to normalise network structure and function.

ImportanceStereoelectroencephalography (SEEG) is more frequently being used in the pre-surgical evaluation of children with focal epilepsy. Many factors affect the rate of identification of a definable seizure onset zone (SOZ) and subsequent seizure freedom following SEEG-guided epilepsy surgery, which have not been systematically examined in multi-centre studies.ObjectivesDetermine the rates and factors that predict (a) whether or not a definable putative SOZ was identified on SEEG and (b) subsequent seizure freedom following surgical intervention.DesignRetrospective cohort studySettingMulticentre study involving 6 of 7 UK Children’s Epilepsy Surgery Service centres that perform paediatric SEEG in the UK.ParticipantsAll children undergoing SEEG from 2014 - March 2019 were included. Demographic, non-invasive evaluation, SEEG and operative factors were collected retrospectively from patient records.Main OutcomesThe two main outcome measures were (a) whether or not a definable putative SOZ was identified on SEEG (binary yes/no outcome) and (b) subsequent seizure freedom following surgical intervention (Engel classification)FindingsOne hundred and thirty-five patients underwent 139 SEEG explorations using a total of 1767 electrodes. A definable SOZ was identified in 117 patients (85.7%); odds of successfully finding a SOZ were 6.4x greater for non-motor seizures compared to motor seizures (p=0.02) and 3.6x more if ≥ 4 seizures were recorded during SEEG (p=0.03). Of 100 patients undergoing surgical treatment, 47 (47.0%) had an Engel class I outcome at a median follow-up of 1.3 years; the only factor associated with outcome was indication for SEEG (p=0.03). SEEG was safe with one (0.7%) haematoma requiring surgical evacuation and no long-term neurological deficits as a result of the procedure.Conclusions and RelevanceThis large nationally representative cohort illustrates that, in these patients who may not have otherwise been offered resective surgery, SEEG-guided surgery can still achieve high rates of seizure freedom. Seizure semiology and the number of seizures recorded during SEEG are important factors in the identification of a definable SOZ and the indication for SEEG is an important factor in post-operative outcomes.

Drug-resistant focal epilepsy is often caused by focal cortical dysplasias (FCDs). The distribution of these lesions across the cerebral cortex and the impact of lesion location on clinical presentation and surgical outcome are largely unknown. We created a neuroimaging cohort of patients with individually mapped FCDs to determine factors associated with lesion location and predictors of postsurgical outcome.The Multi-centre Epilepsy Lesion Detection (MELD) project collated a retrospective cohort of 580 patients with epilepsy attributed to FCD from 20 epilepsy centres worldwide. MRI-based maps of individual FCDs with accompanying demographic, clinical and surgical information were collected. We mapped the distribution of FCDs, examined for associations between clinical factors and lesion location, and developed a predictive model of postsurgical seizure freedom.FCDs were non-uniformly distributed, concentrating in the superior frontal sulcus, frontal pole and temporal pole. Epilepsy onset was typically before age 10. Earlier epilepsy onset was associated with lesions in primary sensory areas while later epilepsy onset was associated with lesions in association cortices. Lesions in temporal and occipital lobes tended to be larger than frontal lobe lesions. Seizure freedom rates varied with FCD location, from around 30% in visual, motor and premotor areas to 75% in superior temporal and frontal gyri. The predictive model of postsurgical seizure freedom had a positive predictive value of 70% and negative predictive value of 61%.FCD location is an important determinant of its size, the age of epilepsy onset and the likelihood of seizure freedom post-surgery. Our atlas of lesion locations can be used to guide the radiological search for subtle lesions in individual patients. Our atlas of regional seizure freedom rates and associated predictive model can be used to estimate individual likelihoods of postsurgical seizure freedom. Data-driven atlases and predictive models are essential for evidence-based, precision medicine and risk counselling in neurology.

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 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.

The presence of interictal epileptiform discharges on electroencephalography (EEG) may indicate increased epileptic seizure risk and on invasive EEG are the signature of the irritative zone. In highly epileptogenic lesions – such as cortical tubers in tuberous sclerosis – these discharges can be recorded with intracranial stereotactic EEG as part of the evaluation for epilepsy surgery. Yet the network mechanisms that underwrite the generation and spread of these discharges remain poorly understood, limiting their current diagnostic use.Here, we investigate the dynamics of interictal epileptiform discharges using a combination of quantitative analysis of invasive EEG recordings and mesoscale neural mass modelling of cortical dynamics. We first characterise spatially organised local dynamics of discharges recorded from 36 separate tubers in 8 patients with tuberous sclerosis. We characterise these dynamics with a set of competing explanatory network models using dynamic causal modelling. Bayesian model comparison of plausible network architectures suggests that the recurrent coupling between neuronal populations within – and adjacent to – the tuber core explains the travelling wave dynamics observed in these patient recordings.Our results – based on interictal activity – unify competing theories about the pathological organisation of epileptic foci and surrounding cortex in patients with tuberous sclerosis. Coupled oscillator dynamics have previously been used to describe ictal activity, where fast travelling ictal discharges are commonly observed within the recruited seizure network. The interictal data analysed here add the insight that this functional architecture is already established in the interictal state. This links observations of interictal EEG abnormalities directly to pathological network coupling in epilepsy, with possible implications for epilepsy surgery approaches in tuberous sclerosis.Significance StatementInterictal epileptiform discharges (IEDs) are clinically important markers of an epileptic brain. Here we link local IED spread to network coupling through a combination of clinical recordings in paediatric patients with tuberous sclerosis complex, quantitative EEG analysis of interictal discharges spread, and Bayesian inference on coupled neural mass model parameters. We show that the kinds of interictal discharges seen in our patients require recurrent local network coupling extending beyond the putative seizure focus and that in fact only those recurrent coupled networks can support seizure-like and interictal dynamics when run in simulation. Our findings provide a novel integrated perspective on emergent epileptic dynamics in human patients.