Dataset Information

Localization of the Epileptogenic Zone Using Interictal MEG and Machine Learning in a Large Cohort of Drug-Resistant Epilepsy Patients.

ABSTRACT: Objective: Epilepsy surgery results in seizure freedom in the majority of drug-resistant patients. To improve surgery outcome we studied whether MEG metrics combined with machine learning can improve localization of the epileptogenic zone, thereby enhancing the chance of seizure freedom. Methods: Presurgical interictal MEG recordings of 94 patients (64 seizure-free >1y post-surgery) were analyzed to extract four metrics in source space: delta power, low-to-high-frequency power ratio, functional connectivity (phase lag index), and minimum spanning tree betweenness centrality. At the group level, we estimated the overlap of the resection area with the five highest values for each metric and determined whether this overlap differed between surgery outcomes. At the individual level, those metrics were used in machine learning classifiers (linear support vector machine (SVM) and random forest) to distinguish between resection and non-resection areas and between surgery outcome groups. Results: The highest values, for all metrics, overlapped with the resection area in more than half of the patients, but the overlap did not differ between surgery outcome groups. The classifiers distinguished the resection areas from non-resection areas with 59.94% accuracy (95% confidence interval: 59.67-60.22%) for SVM and 60.34% (59.98-60.71%) for random forest, but could not differentiate seizure-free from not seizure-free patients [43.77% accuracy (42.08-45.45%) for SVM and 49.03% (47.25-50.82%) for random forest]. Significance: All four metrics localized the resection area but did not distinguish between surgery outcome groups, demonstrating that metrics derived from interictal MEG correspond to expert consensus based on several presurgical evaluation modalities, but do not yet localize the epileptogenic zone. Metrics should be improved such that they correspond to the resection area in seizure-free patients but not in patients with persistent seizures. It is important to test such localization strategies at an individual level, for example by using machine learning or individualized models, since surgery is individually tailored.

SUBMITTER: Nissen IA

PROVIDER: S-EPMC6090046 | biostudies-literature | 2018

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Localization of the Epileptogenic Zone Using Interictal MEG and Machine Learning in a Large Cohort of Drug-Resistant Epilepsy Patients.

Nissen Ida A IA Stam Cornelis J CJ van Straaten Elisabeth C W ECW Wottschel Viktor V Reijneveld Jaap C JC Baayen Johannes C JC de Witt Hamer Philip C PC Idema Sander S Velis Demetrios N DN Hillebrand Arjan A

Frontiers in neurology 20180807

<b>Objective:</b> Epilepsy surgery results in seizure freedom in the majority of drug-resistant patients. To improve surgery outcome we studied whether MEG metrics combined with machine learning can improve localization of the epileptogenic zone, thereby enhancing the chance of seizure freedom. <b>Methods:</b> Presurgical interictal MEG recordings of 94 patients (64 seizure-free >1y post-surgery) were analyzed to extract four metrics in source space: delta power, low-to-high-frequency power rati ...[more]

PMID: 30131762

Similar Datasets

Project description:Electromagnetic source imaging (ESI) offers unique capability of imaging brain dynamics for studying brain functions and aiding the clinical management of brain disorders. Challenges exist in ESI due to the ill-posedness of the inverse problem and thus the need of modeling the underlying brain dynamics for regularizations. Advances in generative models provide opportunities for more accurate and realistic source modeling that could offer an alternative approach to ESI for modeling the underlying brain dynamics beyond equivalent physical source models. However, it is not straightforward to explicitly formulate the knowledge arising from these generative models within the conventional ESI framework. Here we investigate a novel source imaging framework based on mesoscale neuronal modeling and deep learning (DL) that can learn the sensor-source mapping relationship directly from MEG data for ESI. Two DL-based ESI models were trained based on data generated by neural mass models and either generic or personalized head models. The robustness of the two DL models was evaluated by systematic computer simulations and clinical validation in a cohort of 29 drug-resistant focal epilepsy patients who underwent intracranial EEG (iEEG) evaluation or surgical resection. Results estimated from pre-operative MEG interictal spikes were quantified using the overlap with resection regions and the distance to the seizure-onset zone (SOZ) defined by iEEG recordings. The DL-based ESI provided robust results when no personalized head geometry is considered, reaching a spatial dispersion of 21.90 ± 19.03 mm, sublobar concordance of 83 %, and sublobar sensitivity and specificity of 66 and 97 % respectively. When using personalized head geometry derived from individual patients' MRI in the training data, personalized DL-based ESI model can further improve the performance and reached a spatial dispersion of 8.19 ± 8.14 mm, sublobar concordance of 93 %, and sublobar sensitivity and specificity of 77 and 99 % respectively. When compared to the SOZ, the localization error of the personalized approach is 15.78 ± 5.54 mm, outperforming the conventional benchmarks. This work demonstrates that combining generative models and deep learning enables an accurate and robust imaging of epileptogenic zone from MEG recordings with strong sublobar precision, suggesting its added value to enhancing MEG source localization and imaging, and to epilepsy source localization and other clinical applications.

Project description:Aim: In neuroscience research, data are quite often characterized by an imbalanced distribution between the majority and minority classes, an issue that can limit or even worsen the prediction performance of machine learning methods. Different resampling procedures have been developed to face this problem and a lot of work has been done in comparing their effectiveness in different scenarios. Notably, the robustness of such techniques has been tested among a wide variety of different datasets, without considering the performance of each specific dataset. In this study, we compare the performances of different resampling procedures for the imbalanced domain in stereo-electroencephalography (SEEG) recordings of the patients with focal epilepsies who underwent surgery. Methods: We considered data obtained by network analysis of interictal SEEG recorded from 10 patients with drug-resistant focal epilepsies, for a supervised classification problem aimed at distinguishing between the epileptogenic and non-epileptogenic brain regions in interictal conditions. We investigated the effectiveness of five oversampling and five undersampling procedures, using 10 different machine learning classifiers. Moreover, six specific ensemble methods for the imbalanced domain were also tested. To compare the performances, Area under the ROC curve (AUC), F-measure, Geometric Mean, and Balanced Accuracy were considered. Results: Both the resampling procedures showed improved performances with respect to the original dataset. The oversampling procedure was found to be more sensitive to the type of classification method employed, with Adaptive Synthetic Sampling (ADASYN) exhibiting the best performances. All the undersampling approaches were more robust than the oversampling among the different classifiers, with Random Undersampling (RUS) exhibiting the best performance despite being the simplest and most basic classification method. Conclusions: The application of machine learning techniques that take into consideration the balance of features by resampling is beneficial and leads to more accurate localization of the epileptogenic zone from interictal periods. In addition, our results highlight the importance of the type of classification method that must be used together with the resampling to maximize the benefit to the outcome.

Project description:Objective:Electrical source imaging (ESI) is used increasingly to estimate the epileptogenic zone (EZ) in patients with epilepsy. Directed functional connectivity (DFC) coupled to ESI helps to better characterize epileptic networks, but studies on interictal activity have relied on high-density recordings. We investigated the accuracy of ESI and DFC for localizing the EZ, based on low-density clinical electroencephalography (EEG). Methods:We selected patients with the following: (a) focal epilepsy, (b) interictal spikes on standard EEG, (c) either a focal structural lesion concordant with the electroclinical semiology or good postoperative outcome. In 34 patients (20 temporal lobe epilepsy [TLE], 14 extra-TLE [ETLE]), we marked interictal spikes and estimated the cortical activity during each spike in 82 cortical regions using a patient-specific head model and distributed linear inverse solution. DFC between brain regions was computed using Granger-causal modeling followed by network topologic measures. The concordance with the presumed EZ at the sublobar level was computed using the epileptogenic lesion or the resected area in postoperative seizure-free patients. Results:ESI, summed outflow, and efficiency were concordant with the presumed EZ in 76% of the patients, whereas the clustering coefficient and betweenness centrality were concordant in 70% of patients. There was no significant difference between ESI and connectivity measures. In all measures, patients with TLE had a significantly higher (P < 0.05) concordance with the presumed EZ than patients with with ETLE. The brain volume accepted for concordance was significantly larger in TLE. Significance:ESI and DFC derived from low-density EEG can reliably estimate the EZ from interictal spikes. Connectivity measures were not superior to ESI for EZ localization during interictal spikes, but the current validation of the localization of connectivity measure is promising for other applications.

Project description:Background and purposePatients with MR imaging-negative epilepsy could have subtle FCD. Our aim was to determine if structural changes could be identified by using DTI in children with intractable epilepsy, from MR imaging-visible FCD and MR imaging-negative localization-related epilepsy, that were concordant with the epileptogenic zone as defined by using the MEG dipole cluster.Materials and methodsEight children with MR imaging-visible FCD and 16 with MR imaging-negative epilepsy underwent DTI and MEG. Twenty-six age-matched healthy children underwent DTI. Analysis was performed on controls across individual patients. Agreement between the location of DTI abnormalities and FCD and MEG dipole clusters was assessed.ResultsIn patients with MR imaging-visible FCD, abnormal FA, MD, λ(1), λ(2), and λ(3) were lobar concordant with the MEG dipole cluster in 4/8 (50.0%), 5/8 (62.5%), 3/8 (37.5%), 6/8 (75.0%), and 5/8 (62.5%), respectively. In patients with MR imaging-visible FCD, abnormal FA, MD, λ(1), λ(2), and λ(3) overlapped the x-, y-, and z-axes of the MEG dipole cluster in 1/8 (12.5%), 4/8 (50%), 4/8 (50%), 6/8 (75%), and 4/8 (50%), respectively, and with FCD in 1/8 (12.5%), 3/8 (37.5%), 0/8 (0%), 3/8 (37.5%), and 1/8 (12.5%), respectively. In patients with MR imaging-negative epilepsy, abnormal FA, MD, λ(1), λ(2), and λ(3) were lobar-concordant with the MEG dipole cluster in 11/16 (68.8%), 11/16 (68.8%), 8/16 (50.0%), 10/16 (62.5%), and 10/16 (62.5%), respectively, and overlapped the x-, y-, and z-axes of the MEG dipole cluster in 9/16 (56.3%), 10/16 (62.5%), 8/16 (50%), 8/16 (50%), and 8/16 (50%), respectively. There was no significant difference between abnormal DTI lobar concordance with the MEG dipole cluster in patients with MR imaging-visible FCD and MR imaging-negative epilepsy.ConclusionsWhite matter changes can be detected with DTI in children with MR imaging-visible FCD and MR imaging-negative epilepsy, which were concordant with the epileptogenic zone in more than half of the patients.

Project description:Drug-resistant focal epilepsy is a major clinical problem and surgery is under-used. Better non-invasive techniques for epileptogenic zone localization are needed when MRI shows no lesion or an extensive lesion. The problem is interictal and ictal localization before propagation from the epileptogenic zone. High-density EEG (HDEEG) and magnetoencephalography (MEG) offer millisecond-order temporal resolution to address this but co-acquisition is challenging, ictal MEG studies are rare, long-term prospective studies are lacking, and fundamental questions remain. Should HDEEG-MEG discharges be assessed independently [electroencephalographic source localization (ESL), magnetoencephalographic source localization (MSL)] or combined (EMSL) for source localization? Which phase of the discharge best characterizes the epileptogenic zone (defined by intracranial EEG and surgical resection relative to outcome)? Does this differ for interictal and ictal discharges? Does MEG detect mesial temporal lobe discharges? Thirteen patients (10 non-lesional, three extensive-lesional) underwent synchronized HDEEG-MEG (72-94 channel EEG, 306-sensor MEG). Source localization (standardized low-resolution tomographic analysis with MRI patient-individualized boundary-element method) was applied to averaged interictal epileptiform discharges (IED) and ictal discharges at three phases: 'early-phase' (first latency 90% explained variance), 'mid-phase' (first of 50% rising-phase, 50% mean global field power), 'late-phase' (negative peak). 'Earliest-solution' was the first of the three early-phase solutions (ESL, MSL, EMSL). Prospective follow-up was 3-21 (median 12) months before surgery, 14-39 (median 21) months after surgery. IEDs (n = 1474) were recorded, seen in: HDEEG only, 626 (42%); MEG only, 232 (16%); and both 616 (42%). Thirty-three seizures were captured, seen in: HDEEG only, seven (21%); MEG only, one (3%); and both 25 (76%). Intracranial EEG was done in nine patients. Engel scores were I (9/13, 69%), II (2/13,15%), and III (2/13). MEG detected baso-mesial temporal lobe epileptogenic zone sources. Epileptogenic zone OR [odds ratio(s)] were significantly higher for earliest-solution versus early-phase IED-surgical resection and earliest-solution versus all mid-phase and late-phase solutions. ESL outperformed EMSL for ictal-surgical resection [OR 3.54, 95% confidence interval (CI) 1.09-11.55, P = 0.036]. MSL outperformed EMSL for IED-intracranial EEG (OR 4.67, 95% CI 1.19-18.34, P = 0.027). ESL outperformed MSL for ictal-surgical resection (OR 3.73, 95% CI 1.16-12.03, P = 0.028) but was outperformed by MSL for IED-intracranial EEG (OR 0.18, 95% CI 0.05-0.73, P = 0.017). Thus, (i) HDEEG and MEG source solutions more accurately localize the epileptogenic zone at the earliest resolvable phase of interictal and ictal discharges, not mid-phase (as is common practice) or late peak-phase (when signal-to-noise ratios are maximal); (ii) from empirical observation of the differential timing of HDEEG and MEG discharges and based on the superiority of ESL plus MSL over either modality alone and over EMSL, concurrent HDEEG-MEG signals should be assessed independently, not combined; (iii) baso-mesial temporal lobe sources are detectable by MEG; and (iv) MEG is not 'more accurate' than HDEEG-emphasis is best placed on the earliest signal (whether HDEEG or MEG) amenable to source localization. Our findings challenge current practice and our reliance on invasive monitoring in these patients. 10.1093/brain/awz015_video1 awz015media1 6018582479001.

Project description:Electrical source imaging of interictal spikes observed in EEG recordings of patients with refractory epilepsy provides useful information to localize the epileptogenic focus during the presurgical evaluation. However, the selection of the time points or time epochs of the spikes in order to estimate the origin of the activity remains a challenge. In this study, we consider a Bayesian EEG source imaging technique for distributed sources, i.e. the multiple volumetric sparse priors (MSVP) approach. The approach allows to estimate the time courses of the intensity of the sources corresponding with a specific time epoch of the spike. Based on presurgical averaged interictal spikes in six patients who were successfully treated with surgery, we estimated the time courses of the source intensities for three different time epochs: (i) an epoch starting 50 ms before the spike peak and ending at 50% of the spike peak during the rising phase of the spike, (ii) an epoch starting 50 ms before the spike peak and ending at the spike peak and (iii) an epoch containing the full spike time period starting 50 ms before the spike peak and ending 230 ms after the spike peak. To identify the primary source of the spike activity, the source with the maximum energy from 50 ms before the spike peak till 50% of the spike peak was subsequently selected for each of the time windows. For comparison, the activity at the spike peaks and at 50% of the peaks was localized using the LORETA inversion technique and an ECD approach. Both patient-specific spherical forward models and patient-specific 5-layered finite difference models were considered to evaluate the influence of the forward model. Based on the resected zones in each of the patients, extracted from post-operative MR images, we compared the distances to the resection border of the estimated activity. Using the spherical models, the distances to the resection border for the MSVP approach and each of the different time epochs were in the same range as the LORETA and ECD techniques. We found distances smaller than 23 mm, with robust results for all the patients. For the finite difference models, we found that the distances to the resection border for the MSVP inversions of the full spike time epochs were generally smaller compared to the MSVP inversions of the time epochs before the spike peak. The results also suggest that the inversions using the finite difference models resulted in slightly smaller distances to the resection border compared to the spherical models. The results we obtained are promising because the MSVP approach allows to study the network of the estimated source-intensities and allows to characterize the spatial extent of the underlying sources.

Project description:Accurate localization of the epileptogenic zone (EZ) is a key factor to obtain good surgical outcome for refractory epilepsy patients. However, no technique, so far, can precisely locate the EZ, and there are barely any reports on the combined application of multiple technologies to improve the localization accuracy of the EZ. In this study, we aimed to explore the use of a multimodal method combining PET-MRI, fluid and white matter suppression (FLAWS)-a novel MRI sequence, and high-frequency oscillation (HFO) automated analysis to delineate EZ. We retrospectively collected 15 patients with refractory epilepsy who underwent surgery and used the above three methods to detect abnormal brain areas of all patients. We compared the PET-MRI, FLAWS, and HFO results with traditional methods to evaluate their diagnostic value. The sensitivities, specificities of locating the EZ, and marking extent removed versus not removed [RatioChann(ev)] of each method were compared with surgical outcome. We also tested the possibility of using different combinations to locate the EZ. The marked areas in every patient established using each method were also compared to determine the correlations among the three methods. The results showed that PET-MRI, FLAWS, and HFOs can provide more information about potential epileptic areas than traditional methods. When detecting the EZs, the sensitivities of PET-MRI, FLAWS, and HFOs were 68.75, 53.85, and 87.50%, and the specificities were 80.00, 33.33, and 100.00%. The RatioChann(ev) of HFO-marked contacts was significantly higher in patients with good outcome than those with poor outcome (p< 0.05). When intracranial electrodes covered all the abnormal areas indicated by neuroimaging with the overlapping EZs being completely removed referred to HFO analysis, patients could reach seizure-free (p < 0.01). The periphery of the lesion marked by neuroimaging may be epileptic, but not every lesion contributes to seizures. Therefore, approaches in multimodality can detect EZ more accurately, and HFO analysis may help in defining real epileptic areas that may be missed in the neuroimaging results. The implantation of intracranial electrodes guided by non-invasive PET-MRI and FLAWS findings as well as HFO analysis would be an optimized multimodal approach for locating EZ.

Dataset Information

Localization of the Epileptogenic Zone Using Interictal MEG and Machine Learning in a Large Cohort of Drug-Resistant Epilepsy Patients.

Publications

Localization of the Epileptogenic Zone Using Interictal MEG and Machine Learning in a Large Cohort of Drug-Resistant Epilepsy Patients.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets