iEEG Signals Research Articles

Chapter 19 - iEEG: Dura-lining electrodes

Intracranial electroencephalography (iEEG) is measured from electrodes placed in or on the brain. These measurements have an excellent signal-to-noise ratio and iEEG signals have often been used to decode brain activity or drive brain-computer interfaces (BCIs). iEEG recordings are typically done for seizure monitoring in epilepsy patients who have these electrodes placed for a clinical purpose: to localize both brain regions that are essential for function and others where seizures start. Brain regions not involved in epilepsy are thought to function normally and provide a unique opportunity to learn about human neurophysiology. Intracranial electrodes measure the aggregate activity of large neuronal populations and recorded signals contain many features. Different features are extracted by analyzing these signals in the time and frequency domain. The time domain may reveal an evoked potential at a particular time after the onset of an event. Decomposition into the frequency domain may show narrowband peaks in the spectrum at specific frequencies or broadband signal changes that span a wide range of frequencies. Broadband power increases are generally observed when a brain region is active while most other features are highly specific to brain regions, inputs, and tasks. Here we describe the spatiotemporal dynamics of several iEEG signals that have often been used to decode brain activity and drive BCIs.

Node-Centric Graph Learning From Data for Brain State Identification

Data-driven graph learning models a network by determining the strength of connections between its nodes. The data refers to a graph signal which associates a value with each graph node. Existing graph learning methods either use simplified models for the graph signal, or they are prohibitively expensive in terms of computational and memory requirements. This is particularly true when the number of nodes is high or there are temporal changes in the network. In order to consider richer models with a reasonable computational tractability, we introduce a graph learning method based on representation learning on graphs. Representation learning generates an embedding for each graph node, taking the information from neighbouring nodes into account. Our graph learning method further modifies the embeddings to compute the graph similarity matrix. In this work, graph learning is used to examine brain networks for brain state identification. We infer time-varying brain graphs from an extensive dataset of intracranial electroencephalographic (iEEG) signals from ten patients. We then apply the graphs as input to a classifier to distinguish seizure vs. non-seizure brain states. Using the binary classification metric of area under the receiver operating characteristic curve (AUC), this approach yields an average of 9.13 percent improvement when compared to two widely used brain network modeling methods.

Automatic detection of naturally occurring epilepsy in dogs using intracranial electroencephalogram signals

Abstract Automatic detection of epilepsy has become a crucial tool for the treatment of medical-refractory epilepsy and it has gained a great attention of researchers during recent years. Two major bottlenecks in the development of automated epilepsy detection systems are dearth of precise data and deficiency of general algorithms which detects epilepsy with minimum computational requirements. Thanks to advances in electroencephalogram (EEG) machines, scalp or intracranial EEG machines which provides high quality signals are now available. Robust algorithms which detect epilepsy with less computational power and memory is the prevalent requirement. In this work, an automatic detection method is presented to detect naturally occurring epilepsy in dogs by analyzing intracranial EEG (iEEG). A Modified log energy entropy feature is extracted from original iEEG signals and its first and second derivatives. One way analysis of variance (ANOVA) test is also carried out to study the efficiency of extracted features. The performance of the proposed method is evaluated using UPenn - Mayo Clinic Seizure Detection Challenge dataset. The result of k-fold cross validation in k-nearest neighbor (KNN) and support vector machine (SVM) classifiers are higher than that of the state-of-the-art methods.

Deep Learning Approach for Epileptic Focus Localization

The task of epileptic focus localization receives great attention due to its role in an effective epileptic surgery. The clinicians highly depend on the intracranial EEG data to make a surgical decision related to epileptic subjects suffering from uncontrollable seizures. This surgery usually aims to remove the epileptogenic region which requires precise characterization of that area using the EEG recordings. In this paper, we propose two methods based on deep learning targeting accurate automatic epileptic focus localization using the non-stationary EEG recordings. Our first proposed method is based on semi-supervised learning, in which a deep convolutional autoencoder is trained and then the pre-trained encoder is used with multi-layer perceptron as a classifier. The goal is to determine the location of the EEG signal that is responsible for the epileptic activity. In the second proposed method, unsupervised learning scheme is implemented by merging deep convolutional variational autoencoder and K-means algorithm for clustering the iEEG signals into two distinct clusters based on the seizure source. The proposed methods automate and integrate the features extraction and classification processes instead of manually extracting the features as done in the previous studies. Dimensionality reduction is achieved using the autoencoder, while the important spatio-temporal features are extracted from the EEG recordings using the convolutional layers. Moreover, we implemented the inference network of the semi-supervised model on FPGA. The results of our experiments demonstrate high classification accuracy and clustering performance in localizing the epileptic focus compared with the state of the art.

Unsupervised machine-learning classification of electrophysiologically active electrodes during human cognitive task performance

Identification of active electrodes that record task-relevant neurophysiological activity is needed for clinical and industrial applications as well as for investigating brain functions. We developed an unsupervised, fully automated approach to classify active electrodes showing event-related intracranial EEG (iEEG) responses from 115 patients performing a free recall verbal memory task. Our approach employed new interpretable metrics that quantify spectral characteristics of the normalized iEEG signal based on power-in-band and synchrony measures. Unsupervised clustering of the metrics identified distinct sets of active electrodes across different subjects. In the total population of 11,869 electrodes, our method achieved 97% sensitivity and 92.9% specificity with the most efficient metric. We validated our results with anatomical localization revealing significantly greater distribution of active electrodes in brain regions that support verbal memory processing. We propose our machine-learning framework for objective and efficient classification and interpretation of electrophysiological signals of brain activities supporting memory and cognition.

Epileptic Focus Localization via Brain Network Analysis on Riemannian Manifolds.

Brain network connectivity analysis plays an important role in computer-aided automatic localization of seizure onset zone (SOZ) from Intracranial Electroencephalography (iEEG). However, how to accurately compute brain network dynamics is still not well addressed. This work aims to develop an effective measure to find out the dynamics for SOZ localization. Given multiple-channel iEEG signals, the ictal process involves continuous changes of information propagation. In each time slot, the connectivity relationship between channels can be represented as a matrix. Since the matrices from different time slots do not lie on vector spaces, the similarity between them cannot be computed directly. In this paper, we regard the matrices as points on a Riemannian manifold, so that the similarity can be measured by the geodesic distance on the manifold. It addresses the information-losing problem in existing methods using a vector to approximate a matrix. With the Riemannian method, the brain network dynamics are figured out by clustering methods. A temporal segmentation process is applied to refine the segments for SOZ localization. Our method is evaluated on six epilepsy patients, and the SOZ localization performance is evaluated by the area under the curve (AUC) score. Overall, our method obtains an average AUC score of 0.875, which outperforms the existing approaches. Our method preserves more information in measuring the relationship between brain connectivity descriptors, thus is more robust for SOZ localization. Our method has great potentials for clinical epilepsy treatments.

Probabilistic prediction of Epileptic Seizures using SVM.

In this paper, an algorithm based on the linear Support Vector Machine (SVM) tool was proposed to classify intracranial electroencephalography (iEEG) signals as ictal or interictal to perform human seizure prediction, efficiently. Various univariate linear measures were extracted, and the developed classifier performed adequately well with numerous performance metrics, especially the dataset was suffering from a significant imbalanced class, with the majority of samples representing non-seizure events. The proposed tool was indeed able to forecast accurately such rare events, seizures, from a large set of EEG dataset. In fact, our model can predict some seizures with up to 0.4 probability and about 30-40 minutes in advance. The proposed work employed intracranial EEG recordings of 6 patients in the Freiburg EEG database, totalling trained and tested on 34 seizures of 140-hour-long. It exhibits a sensitivity of 78% and specificity of 100% employing a 2-second-long window with 10-fold cross-validation.

O-43 Data-driven selection of active iEEG channels during verbal memory task performance

Background Specific brain regions activated during a memory task can be accurately determined using spectral power and amplitude from intracranial electroencephalography (iEEG) signals sampled from the overlying electrodes implanted across the neocortex. We developed a data-driven method for active electrode determination and compared our results with a previously used approach based on a predefined arbitrary threshold. Material and methods We assessed neurophysiological activity of 57 epilepsy patients during a verbal short-term memory task, in which lists of 12 common nouns were presented for subsequent freely recall. Normalized amplitude of the high gamma frequency component of the signal was extracted from the word presentation epochs. A family of exponential functions was used to automatically determine patient-specific threshold values for selecting active channels. Results Proportion of active electrodes determined with our data-driven method was lower (2056 of 5857 channels – 35.28%) than for the method using a predefined threshold (3865 channels – 66.33%). There was 67.70% overlap between the channels selected with the two. Methods We observed significant differences between the two populations of selected active electrodes obtained with these methods, including the profile of average spectral power response during word presentation and memory effect differences between words that were subsequently recalled and forgotten. Conclusions The proposed method offers a fully automatic, patient-specific selection of active electrodes sampling from brain areas activated by verbal memory processing. Our results suggest improved sensitivity and specificity of active electrode selection that can potentially be applied to other cognitive, perceptual, and motor tasks.

High interictal connectivity within the resection zone is associated with favorable post-surgical outcomes in focal epilepsy patients.

Patients with drug-resistant focal epilepsy are often candidates for invasive surgical therapies. In these patients, it is necessary to accurately localize seizure generators to ensure seizure freedom following intervention. While intracranial electroencephalography (iEEG) is the gold standard for mapping networks for surgery, this approach requires inducing and recording seizures, which may cause patient morbidity. The goal of this study is to evaluate the utility of mapping interictal (non-seizure) iEEG networks to identify targets for surgical treatment. We analyze interictal iEEG recordings and neuroimaging from 27 focal epilepsy patients treated via surgical resection. We generate interictal functional networks by calculating pairwise correlation of iEEG signals across different frequency bands. Using image coregistration and segmentation, we identify electrodes falling within surgically resected tissue (i.e. the resection zone), and compute node-level and edge-level synchrony in relation to the resection zone. We further associate these metrics with post-surgical outcomes. Greater overlap between resected electrodes and highly synchronous electrodes is associated with favorable post-surgical outcomes. Additionally, good-outcome patients have significantly higher connectivity localized within the resection zone compared to those with poorer postoperative seizure control. This finding persists following normalization by a spatially-constrained null model. This study suggests that spatially-informed interictal network synchrony measures can distinguish between good and poor post-surgical outcomes. By capturing clinically-relevant information during interictal periods, our method may ultimately reduce the need for prolonged invasive implants and provide insights into the pathophysiology of an epileptic brain. We discuss next steps for translating these findings into a prospectively useful clinical tool.

A New Unsupervised Detector of High-Frequency Oscillations in Accurate Localization of Epileptic Seizure Onset Zones.

This paper presents a new unsupervised detector for automatically detecting high-frequency oscillations (HFOs) using intracranial electroencephalogram (iEEG) signals. This detector does not presuppose a specific number of clusters and has a good performance. First, the HFO candidates are detected by an initial detection method which distinguishes HFOs from background activities. Then, as significant features, fuzzy entropy, short-time energy, power ratio, and spectral centroid of the HFO candidates are investigated and constructed as a feature vector. Finally, the feature vector is used as the input of the fuzzy- -means-quantization-error-modeling-based expectation-maximization-Gaussian mixture model clustering algorithm. This algorithm has the advantages of detecting HFOs and avoiding false detection caused by artifacts. The concentrations of detected HFOs are used to localize epileptic seizure onset zones in epileptic iEEG signal analysis. A comparison shows that our detector provides better localization performance in terms of sensitivity and specificity than five existing detectors.

Intracerebral EEG Artifact Identification Using Convolutional Neural Networks

Manual and semi-automatic identification of artifacts and unwanted physiological signals in large intracerebral electroencephalographic (iEEG) recordings is time consuming and inaccurate. To date, unsupervised methods to accurately detect iEEG artifacts are not available. This study introduces a novel machine-learning approach for detection of artifacts in iEEG signals in clinically controlled conditions using convolutional neural networks (CNN) and benchmarks the method’s performance against expert annotations. The method was trained and tested on data obtained from St Anne’s University Hospital (Brno, Czech Republic) and validated on data from Mayo Clinic (Rochester, Minnesota, U.S.A). We show that the proposed technique can be used as a generalized model for iEEG artifact detection. Moreover, a transfer learning process might be used for retraining of the generalized version to form a data-specific model. The generalized model can be efficiently retrained for use with different EEG acquisition systems and noise environments. The generalized and specialized model F1 scores on the testing dataset were 0.81 and 0.96, respectively. The CNN model provides faster, more objective, and more reproducible iEEG artifact detection compared to manual approaches.

Continuous behavioral tracing-based online functional brain mapping with intracranial electroencephalography

Intracranial electroencephalography (iEEG) has proven to be a reliable tool in clinical functional brain mapping. While iEEG signals are believed to be highly informative, making online or real-time functional mapping possible, the utilization of corresponding behavioral information in traditional trial-based paradigms is typically low. This imbalance has limited the efficiency and effectiveness in current iEEG-based functional mapping approaches. Objective. In the current study, we were to investigate whether using more dedicate behavioral information can improve iEEG-based functional mapping. Approach. We continuously monitored behavioral outputs of patients during iEEG recording, transformed them into regressors, and assessed their correlations with iEEG signals. Main results. Functional cortical sites identified in preliminary test using the proposed method showed considerable consistency with the results from electrical cortical stimulation (ECS) as well as a trial-based method. Continuous behavioral tracing yielded more dependency to neural signals than sole event markers-derived regressor. Based on this, we successfully performed online functional brain mapping during five neurosurgery operations, which yielded satisfactory results. Significance. In the present study, we showed the utilization of continuous behavioral tracing can improve the efficiency of iEEG-based functional brain mapping. We then demonstrated this approach is adequate for multi-session intraoperative functional assessment within short time span, which is of considerable clinical value.

On generalized notions of the Epileptogenic Zone

Objectives Contemporary interpretations of the Epileptogenic Zone (EZ) are rooted in static attributions of pathological properties to brain tissue, resulting in complex interpretations of the dynamic discharge properties. We wish to critically consider alternative EZ interpretations based on traveling discharge patterns. Methods Most commonly observed patterns in intracranial electroencephalography (iEEG) recordings comprise fast oscillations (8–30 Hz) with gradually increasing amplitude. The activity of the folded sheet is observed through the local field potential recorded on the contacts of a depth electrode. We hypothesize that the gradual amplitude increase on the recorded signals is caused by the physical translation of the neural field and the subsequent spatial averaging of the source activity. We use The Virtual Brain platform to simulate the spread of the epileptic seizure on a piece of a cortical surface located in the vicinity of the intracranial electrode of interest. Results Simulated iEEG signals exhibit the typical feature of the seizure onset pattern – slowly increasing amplitude of the oscillations. The comparison of the recorded and simulated signals in terms of the timing of the apparent seizure onset and the duration of the period of amplitude increase show also quantitative consistency, demonstrating the plausibility of the mechanism on the realistic spatial and temporal scales. Conclusion We show that that field effects of traveling neural activity along the folded cortex may mimic onset patterns at distant iEEG electrodes, in particular those with increasing oscillatory amplitude. Notably, these effects can be observed in contacts distant from the true EZ, and under certain conditions even in absence of highly epileptogenic tissue. These results underscore the importance of the computational modeling in clinical neuroscience and suggest more dynamic interpretations of the concept of Epileptogenic Zone.

Epileptic Seizure Prediction Based on Permutation Entropy.

Epilepsy is a chronic non-communicable disorder of the brain that affects individuals of all ages. It is caused by a sudden abnormal discharge of brain neurons leading to temporary dysfunction. In this regard, if seizures could be predicted a reasonable period of time before their occurrence, epilepsy patients could take precautions against them and improve their safety and quality of life. However, the potential that permutation entropy(PE) can be applied in human epilepsy prediction from intracranial electroencephalogram (iEEG) recordings remains unclear. Here, we described the novel application of PE to track the dynamical changes of human brain activity from iEEG recordings for the epileptic seizure prediction. The iEEG signals of 19 patients were obtained from the Epilepsy Centre at the University Hospital of Freiburg. After preprocessing, PE was extracted in a sliding time window, and a support vector machine (SVM) was employed to discriminate cerebral state. Then a two-step post-processing method was applied for the purpose of prediction. The results showed that we obtained an average sensitivity (SS) of 94% and false prediction rates (FPR) with 0.111 h−1. The best results with SS of 100% and FPR of 0 h−1 were achieved for some patients. The average prediction horizon was 61.93 min, leaving sufficient treatment time before a seizure. These results indicated that applying PE as a feature to extract information and SVM for classification could predict seizures, and the presented method shows great potential in clinical seizure prediction for human.

EPINETLAB: A Software for Seizure-Onset Zone Identification From Intracranial EEG Signal in Epilepsy.

The pre-operative workup of patients with drug-resistant epilepsy requires in some candidates the identification from intracranial EEG (iEEG) of the seizure-onset zone (SOZ), defined as the area responsible of the generation of the seizure and therefore candidate for resection. High-frequency oscillations (HFOs) contained in the iEEG signal have been proposed as biomarker of the SOZ. Their visual identification is a very onerous process and an automated detection tool could be an extremely valuable aid for clinicians, reducing operator-dependent bias, and computational time. In this manuscript, we present the EPINETLAB software, developed as a collection of routines integrated in the EEGLAB framework that aim to provide clinicians with a structured analysis pipeline for HFOs detection and SOZ identification. The tool implements an analysis strategy developed by our group and underwent a preliminary clinical validation that identifies the HFOs area by extracting the statistical properties of HFOs signal and that provides useful information for a topographic characterization of the relationship between clinically defined SOZ and HFO area. Additional functionalities such as inspection of spectral properties of ictal iEEG data and import and analysis of source-space magnetoencephalographic (MEG) data were also included. EPINETLAB was developed with user-friendliness in mind to support clinicians in the identification and quantitative assessment of HFOs in iEEG and source space MEG data and aid the evaluation of the SOZ for pre-surgical assessment.

Promises and limitations of human intracranial electroencephalography.

Intracranial electroencephalography (iEEG), also known as electrocorticography when using subdural grid electrodes or stereotactic EEG when using depth electrodes, is blossoming in various fields of human neuroscience. In this article, we highlight the potentials of iEEG in exploring functions of the human brain while also considering its limitations. The iEEG signal provides anatomically precise information about the selective engagement of neuronal populations at the millimeter scale and the temporal dynamics of their engagement at the millisecond scale. If several nodes of a given network are monitored simultaneously with implanted electrodes, the iEEG signals can also reveal information about functional interactions within and across networks during different stages of neural computation. As such, human iEEG can complement other methods of neuroscience beyond simply replicating what is already known, or can be known, from noninvasive lines of research in humans or from invasive recordings in nonhuman mammalian brains.

Composition of Feature Extraction Methods Shows Interesting Performances in Discriminating Wakefulness and NREM Sleep

Intracranial electroencephalography (iEEG) is an invasive technique used to explore the cortical activity of the brain. In this letter, we focused on features of iEEG signals recorded during wakefulness and non-rapid eye movement (NREM) sleep in order to find differences between the two states, respectively. We preliminary screened the data using standard deviation analysis (STD). Then, we compared and combined STD values with coefficients from wavelet decomposition (Daubechies mother wavelet of order 4). Resulting parameters were classified using an artificial neural network. STD analysis underlined two brain areas [superior temporal sulcus (STS) and intraparietal-sulcus and parietal transverse (IPS)] with different electrical activity in the two states. STD values of STS and IPS channels were highly correlated in time; therefore, only STS was then used further in the features extraction analysis. Approximation and detail coefficients from Daubechies decomposition were used alone or in combination with the STD value. The overall accuracy of the pattern recognition was higher (98.57%), when features from different methods were used in combination. Our test was able to automatically recognize wake or NREM sleep status with very good discrimination performances using one single iEEG electrode.

Signal Complexity of Human Intracranial EEG Tracks Successful Associative-Memory Formation across Individuals.

Memory performance is highly variable among individuals. Most studies examining human memory, however, have largely focused on the neural correlates of successful memory formation within individuals, rather than the differences among them. As such, what gives rise to this variability is poorly understood. Here, we examined intracranial EEG (iEEG) recordings captured from 43 participants (23 male) implanted with subdural electrodes for seizure monitoring as they performed a paired-associates verbal memory task. We identified three separate but related signatures of neural activity that tracked differences in successful memory formation across individuals. High-performing individuals consistently exhibited less broadband power, flatter power spectral density slopes, and greater complexity in their iEEG signals. Furthermore, within individuals across three separate time scales ranging from seconds to days, successful recall was positively associated with these same metrics. Our data therefore suggest that memory ability across individuals can be indexed by increased neural signal complexity.SIGNIFICANCE STATEMENT We show that participants whose intracranial EEG exhibits less low-frequency power, flatter power spectrums, and greater sample entropy overall are better able to memorize associations, and that the same metrics track fluctuations in memory performance across time within individuals. These metrics together signify greater neural signal complexity, which may index the brain's ability to flexibly engage with information and generate separable memory representations. Critically, the current set of results provides a unique window into the neural markers of individual differences in memory performance, which have hitherto been underexplored.

Analysis of non-seizure and seizure activity using intracranial EEG signals and empirical mode decomposition based approximate entropy

Epilepsy is a series brain dynamical disorder, characterized by recurrent seizures. It is estimated that incidence rates (number of new cases) range from 24 to 53 per 100000 for every year. Approximately, 33% of epilepsy patients suffer from seizures that are not controlled by anti-convulsant medications. The patients with uncontrolled seizures experience several limitations in family, social and educational activities. Therefore, epilepsy research based on the diagnosis and treatment of seizure is considered to be greater importance. In this study, a method based on empirical mode decomposition and approximate entropy (ApEn) is proposed to analyze the intracranial electroencephalography (iEEG) recorded in non-seizure and seizure activity. For this purpose, a standard database based of University of Bonn, Germany is utilized. The iEEG signals of non-seizure and seizure class are subjected to empirical mode decomposition and Intrinsic Mode Functions (IMF) are obtained. Then the approximate entropy is computed from each IMF. The results show that IMF based ApEn is higher in non-seizure class. The ApEn extracted from the fourth IMF is found to perform better in terms of separating these signals. Highest percentage difference of 165% is obtained for this IMF. Further, the ApEn values extracted from all IMFs except IMF7 and IMF8 is found to have highly statistically significant (p<0.0001).

Automatic bad channel detection in intracranial electroencephalographic recordings using ensemble machine learning

ObjectiveIntracranial electroencephalographic (iEEG) recordings contain “bad channels”, which show non-neuronal signals. Here, we developed a new method that automatically detects iEEG bad channels using machine learning of seven signal features. MethodsThe features quantified signals’ variance, spatial–temporal correlation and nonlinear properties. Because the number of bad channels is usually much lower than the number of good channels, we implemented an ensemble bagging classifier known to be optimal in terms of stability and predictive accuracy for datasets with imbalanced class distributions. This method was applied on stereo-electroencephalographic (SEEG) signals recording during low frequency stimulations performed in 206 patients from 5 clinical centers. ResultsWe found that the classification accuracy was extremely good: It increased with the number of subjects used to train the classifier and reached a plateau at 99.77% for 110 subjects. The classification performance was thus not impacted by the multicentric nature of data. ConclusionsThe proposed method to automatically detect bad channels demonstrated convincing results and can be envisaged to be used on larger datasets for automatic quality control of iEEG data. SignificanceThis is the first method proposed to classify bad channels in iEEG and should allow to improve the data selection when reviewing iEEG signals.