DeepIED: An epileptic discharge detector for EEG-fMRI based on deep learning.
Ontology highlight
ABSTRACT:
SUBMITTER: Hao Y
PROVIDER: S-EPMC5752096 | biostudies-literature |
REPOSITORIES: biostudies-literature
Similar Datasets
Project description:Electroencephalogram (EEG) is one of the main diagnostic tests for epilepsy. The detection of epileptic activity is usually performed by a human expert and is based on finding specific patterns in the multi-channel electroencephalogram. This is a difficult and time-consuming task, therefore various attempts are made to automate it using both conventional and Deep Learning (DL) techniques. Unfortunately, authors do not often provide sufficiently detailed and complete information to be able to reproduce their results. Our work is intended to fill this gap. Using a carefully selected 79 neonatal EEG recordings we developed a complete framework for seizure detection using DL approch. We share a ready to use R and Python codes which allow: (a) read raw European Data Format files, (b) read data files containing the seizure annotations made by human experts, (c) extract train, validation and test data, (d) create an appropriate Convolutional Neural Network (CNN) model, (e) train the model, (f) check the quality of the neural classifier, (g) save all learning results.
| S-EPMC9338048 | biostudies-literature
Project description:ObjectiveThe concurrent recording of electroencephalography (EEG) and functional magnetic resonance imaging (fMRI) is a technique that has received much attention due to its potential for combined high temporal and spatial resolution. However, the ballistocardiogram (BCG), a large-amplitude artifact caused by cardiac induced movement contaminates the EEG during EEG-fMRI recordings. Removal of BCG in software has generally made use of linear decompositions of the corrupted EEG. This is not ideal as the BCG signal propagates in a manner which is non-linearly dependent on the electrocardiogram (ECG). In this paper, we present a novel method for BCG artifact suppression using recurrent neural networks (RNNs).MethodsEEG signals were recovered by training RNNs on the nonlinear mappings between ECG and the BCG corrupted EEG. We evaluated our model's performance against the commonly used Optimal Basis Set (OBS) method at the level of individual subjects, and investigated generalization across subjects.ResultsWe show that our algorithm can generate larger average power reduction of the BCG at critical frequencies, while simultaneously improving task relevant EEG based classification.ConclusionThe presented deep learning architecture can be used to reduce BCG related artifacts in EEG-fMRI recordings.SignificanceWe present a deep learning approach that can be used to suppress the BCG artifact in EEG-fMRI without the use of additional hardware. This method may have scope to be combined with current hardware methods, operate in real-time and be used for direct modeling of the BCG.
| S-EPMC7808341 | biostudies-literature
Project description:PurposeIn simultaneous scalp electroencephalography (EEG) and functional magnetic resonance imaging (fMRI), blood oxygen level dependent (BOLD) changes occurring before the spike have been sometimes described but could not be explained. To characterize the origin of this prespike BOLD signal change, we looked for electrographic changes in stereo-EEG (SEEG) possibly preceding the scalp spike in patients that showed early BOLD response in EEG/fMRI.MethodsWe studied four patients with drug-resistant focal epilepsy who underwent EEG/fMRI, showed a prespike BOLD response, and were then studied with depth electrodes for presurgical localization of the epileptic generator. Early BOLD responses in the region of the spike field were analyzed using models with hemodynamic response functions (HRFs) peaking from -9 to +9 s around the spike. SEEG recordings in the period and location corresponding to the early HRF responses were analyzed to detect if electrographic changes were present in the SEEG before the scalp abnormality.Key findingsOne of the four patients presented a SEEG interictal discharge in the period corresponding to the early BOLD response. In the other three, no electrographic changes were detected in the SEEG in the period corresponding to early BOLD changes.SignificanceAlthough the early BOLD activity may sometimes be explained by a synchronized neural discharge detectable with SEEG but not visible on the scalp EEG, in most cases the early BOLD response reflects a metabolic phenomenon that does not appear to result from a synchronized neuronal discharge. Prespike metabolic responses can result from synchronized or nonsynchronized neuronal activity, or from nonneuronal mechanisms including glia.
| S-EPMC3771927 | biostudies-literature
Project description:Background: Potentially curative epilepsy surgery can be offered if a single, discrete epileptogenic zone (EZ) can be identified. For individuals in whom there is no clear concordance between clinical localization, scalp EEG, and imaging data, intracranial EEG (icEEG) may be needed to confirm a predefined hypothesis regarding irritative zone (IZ), seizure onset zone (SOZ), and EZ prior to surgery. However, icEEG has limited spatial sampling and may fail to reveal the full extent of epileptogenic network if predefined hypothesis is not correct. Simultaneous icEEG-fMRI has been safely acquired in humans and allows exploration of neuronal activity at the whole-brain level related to interictal epileptiform discharges (IED) captured intracranially. Methods: We report icEEG-fMRI in eight patients with refractory focal epilepsy who had resective surgery and good postsurgical outcome. Surgical resection volume in seizure-free patients post-surgically reflects confirmed identification of the EZ. IEDs on icEEG were classified according to their topographic distribution and localization (Focal, Regional, Widespread, and Non-contiguous). We also divided IEDs by their location within the surgical resection volume [primary IZ (IZ1) IED] or outside [secondary IZ (IZ2) IED]. The distribution of fMRI blood oxygen level-dependent (BOLD) changes associated with individual IED classes were assessed over the whole brain using a general linear model. The concordance of resulting BOLD map was evaluated by comparing localization of BOLD clusters with surgical resection volume. Additionally, we compared the concordance of BOLD maps and presence of BOLD clusters in remote brain areas: precuneus, cuneus, cingulate, medial frontal, and thalamus for different IED classes. Results: A total of 38 different topographic IED classes were identified across the 8 patients: Focal (22) and non-focal (16, Regional = 9, Widespread = 2, Non-contiguous = 5). Twenty-nine IEDs originated from IZ1 and 9 from IZ2. All IED classes were associated with BOLD changes. BOLD maps were concordant with the surgical resection volume for 27/38 (71%) IED classes, showing statistical global maximum BOLD cluster or another cluster in the surgical resection volume. The concordance of BOLD maps with surgical resection volume was greater (p < 0.05) for non-focal (87.5%, 14/16) as compared to Focal (59%, 13/22) IED classes. Additionally, BOLD clusters in remote cortical and deep brain areas were present in 84% (32/38) of BOLD maps, more commonly (15/16; 93%) for non-focal IED-related BOLD maps. Conclusions: Simultaneous icEEG-fMRI can reveal BOLD changes at the whole-brain level for a wide range of IEDs on icEEG. BOLD clusters within surgical resection volume and remote brain areas were more commonly seen for non-focal IED classes, suggesting that a wider hemodynamic network is at play.
| S-EPMC8490636 | biostudies-literature
Project description:To assess the extent of brain involvement during focal epileptic activity, we studied patterns of cortical and subcortical metabolic changes coinciding with interictal epileptic discharges (IEDs) using group analysis of simultaneous electroencephalography and functional magnetic resonance imaging (EEG-fMRI) scans in patients with focal epilepsy.We selected patients with temporal lobe epilepsy (TLE, n = 32), frontal lobe epilepsy (FLE, n = 14), and posterior quadrant epilepsy (PQE, n = 20) from our 3 Tesla EEG-fMRI database. We applied group analysis upon the blood oxygen-level dependent (BOLD) response associated with focal IEDs.Patients with TLE and FLE showed activations and deactivations, whereas in PQE only deactivations occurred. In TLE and FLE, the largest activation was in the mid-cingulate gyri bilaterally. In FLE, activations were also found in the ipsilateral frontal operculum, thalamus, and internal capsule, and in the contralateral cerebellum, whereas in TLE, we found additional activations in the ipsilateral mesial and neocortical temporal regions, insula, and cerebellar cortex. All three groups showed deactivations in default mode network regions, the most widespread being in the TLE group, and less in PQE and FLE.These results indicate that different epileptic syndromes result in unique and widespread networks related to focal IEDs. Default mode regions are deactivated in response to focal discharges in all three groups with syndrome specific pattern. We conclude that focal IEDs are associated with specific networks of widespread metabolic changes that may cause more substantial disturbance to brain function than might be appreciated from the focal nature of the scalp EEG discharges.
| S-EPMC4492710 | biostudies-literature
Project description:Most fMRI studies of the brain's intrinsic functional connectivity (FC) have assumed that this is static; however, it is now clear that it changes over time. This is particularly relevant in epilepsy, which is characterized by a continuous interchange between epileptic and normal brain states associated with the occurrence of epileptic activity. Interestingly, recurrent states of dynamic FC (dFC) have been found in fMRI data using unsupervised learning techniques, assuming either their sparse or non-sparse combination. Here, we propose an l1-norm regularized dictionary learning (l1-DL) approach for dFC state estimation, which allows an intermediate and flexible degree of sparsity in time, and demonstrate its application in the identification of epilepsy-related dFC states using simultaneous EEG-fMRI data. With this l1-DL approach, we aim to accommodate a potentially varying degree of sparsity upon the interchange between epileptic and non-epileptic dFC states. The simultaneous recording of the EEG is used to extract time courses representative of epileptic activity, which are incorporated into the fMRI dFC state analysis to inform the selection of epilepsy-related dFC states. We found that the proposed l1-DL method performed best at identifying epilepsy-related dFC states, when compared with two alternative methods of extreme sparsity (k-means clustering, maximum; and principal component analysis, minimum), as well as an l0-norm regularization framework (l0-DL), with a fixed amount of temporal sparsity. We further showed that epilepsy-related dFC states provide novel insights into the dynamics of epileptic networks, which go beyond the information provided by more conventional EEG-correlated fMRI analysis, and which were concordant with the clinical profile of each patient. In addition to its application in epilepsy, our study provides a new dFC state identification method of potential relevance for studying brain functional connectivity dynamics in general.
| S-EPMC6345787 | biostudies-literature
Project description:ObjectivesTo evaluate the clinical relevance of EEG/fMRI in patients with focal epilepsy, by assessing the information it adds to the scalp EEG in the definition of the epileptic focus.MethodsForty-three patients with focal epilepsy were studied with EEG/fMRI using a 3-T scanner. Blood oxygen level-dependent (BOLD) signal changes related to interictal epileptic discharges (IEDs) were classified as concordant or not concordant with the scalp EEG spike field and as contributory if the BOLD signal provided additional information to the scalp EEG about the epileptic focus or not contributory if it did not. We considered patients having intracerebral EEG or a focal lesion on MRI as having independent validation.ResultsThirty-three patients had at least 3 IEDs during the EEG/fMRI acquisition (active EEG), and all had a BOLD response. In 29 of 33 (88%) patients, the BOLD response was concordant, and in 21 of 33 (64%) patients, the BOLD response was contributory. Fourteen patients had an independent validation: in 12 of these 14, the BOLD responses were validated and in 2 they were invalidated.ConclusionsA BOLD response was present in all patients with active EEG, and more specific localization of the epileptic focus was gained from EEG/fMRI in half of the patients who were scanned, when compared with scalp EEG alone. This study demonstrates that EEG/fMRI, in the context of a clinical practice, may contribute to the localization of the interictal epileptic generator in patients with focal epilepsy.
| S-EPMC3345614 | biostudies-literature
Project description:The present study aims to investigate whether a newly developed fast fMRI called MREG (magnetic resonance encephalography) measures metabolic changes related to interictal epileptic discharges (IED). For this purpose BOLD changes are correlated with the IED distribution and variability.Patients with focal epilepsy underwent EEG-MREG using a 64 channel cap. IED voltage maps were generated using 32 and 64 channels and compared regarding their correspondence to the BOLD response. The extents of IEDs (defined as number of channels with >50% of maximum IED negativity) were correlated with the extents of positive and negative BOLD responses. Differences in inter-spike variability were investigated between interictal epileptic discharges (IED) sets with and without concordant positive or negative BOLD responses.17 patients showed 32 separate IED types. In 50% of IED types the BOLD changes could be confirmed by another independent imaging method. The IED extent significantly correlated with the positive BOLD extent (p = 0.04). In 6 patients the 64-channel EEG voltage maps better reflected the positive or negative BOLD response than the 32-channel EEG; in all others no difference was seen. Inter-spike variability was significantly lower in IED sets with than without concordant positive or negative BOLD responses (with p = 0.04).Higher density EEG and fast fMRI seem to improve the value of EEG-fMRI in epilepsy. The correlation of positive BOLD and IED extent could suggest that widespread BOLD responses reflect the IED network. Inter-spike variability influences the likelihood to find IED concordant positive or negative BOLD responses, which is why single IED analysis may be promising.
| S-EPMC4619722 | biostudies-literature
Project description:Epileptic networks, defined as brain regions involved in epileptic brain activity, have been mapped by functional connectivity in simultaneous electroencephalography and functional magnetic resonance imaging (EEG-fMRI) recordings. This technique allows to define brain hemodynamic changes, measured by the Blood Oxygen Level Dependent (BOLD) signal, associated to the interictal epileptic discharges (IED), which together with ictal events constitute a signature of epileptic disease. Given the highly time-varying nature of epileptic activity, a dynamic functional connectivity (dFC) analysis of EEG-fMRI data appears particularly suitable, having the potential to identify transitory features of specific connections in epileptic networks. In the present study, we propose a novel method, defined dFC-EEG, that integrates dFC assessed by fMRI with the information recorded by simultaneous scalp EEG, in order to identify the connections characterised by a dynamic profile correlated with the occurrence of IED, forming the dynamic epileptic subnetwork. Ten patients with drug-resistant focal epilepsy were included, with different aetiology and showing a widespread (or multilobar) BOLD activation, defined as involving at least two distinct clusters, located in two different lobes and/or extended to the hemisphere contralateral to the epileptic focus. The epileptic focus was defined from the IED-related BOLD map. Regions involved in the occurrence of interictal epileptic activity; i.e., forming the epileptic network, were identified by a general linear model considering the timecourse of the fMRI-defined focus as main regressor. dFC between these regions was assessed with a sliding-window approach. dFC timecourses were then correlated with the sliding-window variance of the IED signal (VarIED), to identify connections whose dynamics related to the epileptic activity; i.e., the dynamic epileptic subnetwork. As expected, given the very different clinical picture of each individual, the extent of this subnetwork was highly variable across patients, but was but was reduced of at least 30% with respect to the initially identified epileptic network in 9/10 patients. The connections of the dynamic subnetwork were most commonly close to the epileptic focus, as reflected by the laterality index of the subnetwork connections, reported higher than the one within the original epileptic network. Moreover, the correlation between dFC timecourses and VarIED was predominantly positive, suggesting a strengthening of the dynamic subnetwork associated to the occurrence of IED. The integration of dFC and scalp IED offers a more specific description of the epileptic network, identifying connections strongly influenced by IED. These findings could be relevant in the pre-surgical evaluation for the resection or disconnection of the epileptogenic zone and help in reaching a better post-surgical outcome. This would be particularly important for patients characterised by a widespread pathological brain activity which challenges the surgical intervention.
| S-EPMC7645285 | biostudies-literature
Project description:Emotion recognition from electroencephalogram (EEG) signals requires accurate and efficient signal processing and feature extraction. Deep learning technology has enabled the automatic extraction of raw EEG signal features that contribute to classifying emotions more accurately. Despite such advances, classification of emotions from EEG signals, especially recorded during recalling specific memories or imagining emotional situations has not yet been investigated. In addition, high-density EEG signal classification using deep neural networks faces challenges, such as high computational complexity, redundant channels, and low accuracy. To address these problems, we evaluate the effects of using a simple channel selection method for classifying self-induced emotions based on deep learning. The experiments demonstrate that selecting key channels based on signal statistics can reduce the computational complexity by 89% without decreasing the classification accuracy. The channel selection method with the highest accuracy was the kurtosis-based method, which achieved accuracies of 79.03% and 79.36% for the valence and arousal scales, respectively. The experimental results show that the proposed framework outperforms conventional methods, even though it uses fewer channels. Our proposed method can be beneficial for the effective use of EEG signals in practical applications.
| S-EPMC9523358 | biostudies-literature