A Stacked Sparse Autoencoder-Based Detector for Automatic Identification of Neuromagnetic High Frequency Oscillations in Epilepsy

DeepReducer: A linear transformer-based model for MEG denoising

Measuring event-related magnetic fields (ERFs) in magnetoencephalography (MEG) is crucial for investigating perceptual and cognitive information processing in both neuroscience research and clinical practice. However, the magnitude of the ERF in cortical sources is comparable to the noise in a single trial. Consequently, numerous repetitive recordings are needed to distinguish these sources from background noise, requiring lengthy time for data acquisition. Herein, we introduce DeepReducer, a linear transformer-based deep learning model designed to reliably and efficiently denoise ERFs, thereby reducing the number of required trials. DeepReducer was trained on a mix of limited-trial and multi-trial averaged ERFs, employing mean squared error as the loss function to effectively capture and model the complex signal fluctuations inherent in MEG recordings. Validation on both semi-synthetic and experimental task-related MEG data showed that DeepReducer outperforms conventional trial-averaging techniques, significantly improving the signal-to-noise ratio of ERFs and reducing source localization errors. The practical significance of DeepReducer encompasses optimizing MEG data acquisition by reducing participant stress (particularly for patients) and minimizing associated artifacts.

Magnetoencephalography-based approaches to epilepsy classification

Epilepsy is a chronic central nervous system disorder characterized by recurrent seizures. Not only does epilepsy severely affect the daily life of the patient, but the risk of premature death in patients with epilepsy is three times higher than that of the normal population. Magnetoencephalography (MEG) is a non-invasive, high temporal and spatial resolution electrophysiological data that provides a valid basis for epilepsy diagnosis, and used in clinical practice to locate epileptic foci in patients with epilepsy. It has been shown that MEG helps to identify MRI-negative epilepsy, contributes to clinical decision-making in recurrent seizures after previous epilepsy surgery, that interictal MEG can provide additional localization information than scalp EEG, and complete excision of the stimulation area defined by the MEG has prognostic significance for postoperative seizure control. However, due to the complexity of the MEG signal, it is often difficult to identify subtle but critical changes in MEG through visual inspection, opening up an important area of research for biomedical engineers to investigate and implement intelligent algorithms for epilepsy recognition. At the same time, the use of manual markers requires significant time and labor costs, necessitating the development and use of computer-aided diagnosis (CAD) systems that use classifiers to automatically identify abnormal activity. In this review, we discuss in detail the results of applying various different feature extraction methods on MEG signals with different classifiers for epilepsy detection, subtype determination, and laterality classification. Finally, we also briefly look at the prospects of using MEG for epilepsy-assisted localization (spike detection, high-frequency oscillation detection) due to the unique advantages of MEG for functional area localization in epilepsy, and discuss the limitation of current research status and suggestions for future research. Overall, it is hoped that our review will facilitate the reader to quickly gain a general understanding of the problem of MEG-based epilepsy classification and provide ideas and directions for subsequent research.

A multi-head self-attention deep learning approach for detection and recommendation of neuromagnetic high frequency oscillations in epilepsy

Magnetoencephalography is a noninvasive neuromagnetic technology to record epileptic activities for the pre-operative localization of epileptogenic zones, which has received increasing attention in the diagnosis and surgery of epilepsy. As reported by recent studies, pathological high frequency oscillations (HFOs), when utilized as a biomarker to localize the epileptogenic zones, result in a significant reduction in seizure frequency, even seizure elimination in around 80% of cases. Thus, objective, rapid, and automatic detection and recommendation of HFOs are highly desirable for clinicians to alleviate the burden of reviewing a large amount of MEG data from a given patient. Despite the advantage, the performance of existing HFOs rarely satisfies the clinical requirement. Consequently, no HFOs have been successfully applied to real clinical applications so far. In this work, we propose a multi-head self-attention-based detector for recommendation, termed MSADR, to detect and recommend HFO signals. Taking advantage of the state-of-the-art multi-head self-attention mechanism in deep learning, the proposed MSADR achieves a more superior accuracy of 88.6% than peer machine learning models in both detection and recommendation tasks. In addition, the robustness of MSADR is also extensively assessed with various ablation tests, results of which further demonstrate the effectiveness and generalizability of the proposed approach.

Transformer-Based High-Frequency Oscillation Signal Detection on Magnetoencephalography From Epileptic Patients

High-frequency oscillations (HFOs), observed within 80–500 Hz of magnetoencephalography (MEG) data, are putative biomarkers to localize epileptogenic zones that are critical for the success of surgical epilepsy treatment. It is crucial to accurately detect HFOs for improving the surgical outcome of patients with epilepsy. However, in clinical practices, detecting HFOs in MEG signals mainly depends on visual inspection by clinicians, which is very time-consuming, labor-intensive, subjective, and error-prone. To accurately and automatically detect HFOs, machine learning approaches have been developed and have demonstrated the promising results of automated HFO detection. More recently, the transformer-based model has attracted wide attention and achieved state-of-the-art performance on many machine learning tasks. In this paper, we are investigating the suitability of transformer-based models on the detection of HFOs. Specifically, we propose a transformer-based HFO detection framework for biomedical MEG one-dimensional signal data. For signal classification, we develop a transformer-based HFO (TransHFO) classification model. Then, we investigate the relationship between depth of deep learning models and classification performance. The experimental results show that the proposed framework outperforms the state-of-the-art HFO classifiers, increasing classification accuracy by 7%. Furthermore, we find that shallow TransHFO (< 10 layers) outperforms deep TransHFO models (≥10 layers) on most data augmented factors.

HFO Detection in Epilepsy: A Stacked Denoising Autoencoder and Sample Weight Adjusting Factors-Based Method

High-frequency oscillations (HFOs) recorded by the intracranial electroencephalography (iEEG) are the promising biomarkers of epileptogenic zones. Accurate detection of HFOs is the key to pre-operative assessment for epilepsy. Due to the subjective bias caused by manual features and the class imbalance between HFOs and false HFOs, it is difficult to obtain satisfactory detection performance by the existing methods. To solve these problems, we put forward a novel method to accurately detect HFOs based on the stacked denoising autoencoder (SDAE) and the ensemble classifier with sample weight adjusting factors. First, the adjustable threshold of Hilbert envelopes is proposed to isolate the events of interest (EoIs) from background activities. Then, the SDAE network is utilized to automatically extract features of EoIs in the time-frequency domain. Finally, the AdaBoost-based support vector machine ensemble classifier with sample weight adjusting factors is devised to separate HFOs from EoIs by using the extracted features. These adjusting factors are used to solve the class imbalance problem by adjusting sample weights when learning the base classifiers. Our HFO detection method is evaluated by using clinical iEEG data recorded from 20 patients with medically refractory epilepsy. The experimental results show that our detection method outperforms some existing methods in terms of sensitivity and false discovery rate. In addition, the HFOs detected by our method are effective for localizing seizure onset zones.

Diagnosis of Infantile Hip Dysplasia with B-mode Ultrasound via Two-stage Meta-learning Based Deep Exclusivity Regularized Machine

The B-mode ultrasound (BUS) based computer-aided diagnosis (CAD) has shown its effectiveness for developmental dysplasia of the hip (DDH) in infants. In this work, a two-stage meta-learning based deep exclusivity regularized machine (TML-DERM) is proposed for the BUS-based CAD of DDH. TML-DERM integrates deep neural network (DNN) and exclusivity regularized machine into a unified framework to simultaneously improve the feature representation and classification performance. Moreover, the first-stage meta-learning is mainly conducted on the DNN module to alleviate the overfitting issue caused by the significantly increased parameters in DNN, and a random sampling strategy is adopted to self-generate the meta-tasks; while the second-stage meta-learning mainly learns the combination of multiple weak classifiers by a weight vector to improve the classification performance, and also optimizes the unified framework again. The experimental results on a DDH ultrasound dataset show the proposed TML-DERM achieves the superior classification performance with the mean accuracy of 85.89%, sensitivity of 86.54%, and specificity of 85.23%.

Neuromagnetic high frequency spikes are a new and noninvasive biomarker for localization of epileptogenic zones

OBJECTIVE

One barrier hindering high frequency brain signals (HFBS, >80 Hz) from wide clinical applications is that the brain generates both pathological and physiological HFBS. This study was to find specific biomarkers for localizing epileptogenic zones (EZs).

METHODS

Twenty three children with drug-resistant epilepsy and age/sex matched healthy controls were studied with magnetoencephalography (MEG). High frequency oscillations (HFOs, > 4 oscillatory waveforms) and high frequency spikes (HFSs, > 1 spiky or sharp waveforms) in 80-250 Hz and 250-600 Hz bands were blindly detected with an artificial intelligence method and validated with visual inspection. The magnitude of HFOs and HFSs were quantified with spectral analyses. Sources of HFSs and HFOs were localized and compared with clinical EZs determined by invasive recordings and surgical outcomes.

RESULTS

HFOs in 80-250 Hz and 250-600 Hz were identified in both epilepsy patients (18/23, 12/23, respectively) and healthy controls (6/23, 4/23, respectively). HFSs in 80-250 Hz and 250-600 Hz were detected in patients (16/23, 11/23, respectively) but not in healthy controls. A combination of HFOs and HFSs localized EZs for 22 (22/23, 96%) patients.

CONCLUSIONS

The results indicate, for the first time, that HFSs are a newer and more specific biomarker than HFOs for localizing EZs because HFOs appeared in both epilepsy patients and healthy controls while HFSs appeared only in epilepsy patients.

Detecting High Frequency Oscillations for Stereoelectroencephalography in Epilepsy via Hypergraph Learning

Successful epilepsy surgeries depend highly on pre-operative localization of epileptogenic zones. Stereoelectroencephalography (SEEG) records interictal and ictal activities of the epilepsy in order to precisely find and localize epileptogenic zones in clinical practice. While it is difficult to find distinct ictal onset patterns generated the seizure onset zone from SEEG recordings in a confined region, high frequency oscillations are commonly considered as putative biomarkers for the identification of epileptogenic zones. Therefore, automatic and accurate detection of high frequency oscillations in SEEG signals is crucial for timely clinical evaluation. This work formulates the detection of high frequency oscillations as a signal segment classification problem and develops a hypergraph-based detector to automatically detect high frequency oscillations such that human experts can visually review SEEG signals. We evaluated our method on 4,000 signal segments from clinical SEEG recordings that contain both ictal and interictal data obtained from 19 patients who suffer from refractory focal epilepsy. The experimental results demonstrate the effectiveness of the proposed detector that can successfully localize interictal high frequency oscillations and outperforms multiple peer machine learning methods. In particular, the proposed detector achieved 90.7% in accuracy, 80.9% in sensitivity, and 96.9% in specificity.

Detection of high-frequency oscillations in electroencephalography: A scoping review and an adaptable open-source framework

PURPOSE

High frequency oscillations (HFOs) are putative biomarkers of epileptogenicity. These electrophysiological phenomena can be effectively detected in electroencephalography using automated methods. Nonetheless, the implementation of these methods into clinical practice remains challenging as significant variability exists between algorithms and their characterizations of HFOs. Here, we perform a scoping review of the literature pertaining to automated HFO detection methods. In addition, we propose a framework for defining and detecting HFOs based on a simplified single-stage time-frequency based detection algorithm with clinically-familiar parameters.

METHODS

Several databases (OVID Medline, Web of Science, PubMed) were searched for articles presenting novel, automated HFO detection methods. Details related to the algorithm and various stages of data acquisition, pre-processing, and analysis were abstracted from included studies.

RESULTS

From the 261 records screened, 57 articles presented novel, automated HFO detection methods and were included in the scoping review. These algorithms were categorized into 3 groups based on their most salient features: energy thresholding, time-frequency analysis, and data mining/machine learning. Algorithms were optimized for specific datasets and suffered from low specificity. A framework for user-constrained inputs is proposed to circumvent some of the weaknesses of highly performant detectors.

CONCLUSIONS

Further efforts are required to optimize and validate existing automated HFO detection methods for clinical utility. The proposed framework may be applied to understand and standardize the variations in HFO definitions across institutions.

Recent advances in the noninvasive detection of high-frequency oscillations in the human brain

In recent decades, a significant body of evidence based on invasive clinical research has showed that high-frequency oscillations (HFOs) are a promising biomarker for localization of the seizure onset zone (SOZ), and therefore, have the potential to improve postsurgical outcomes in patients with epilepsy. Emerging clinical literature has demonstrated that HFOs can be recorded noninvasively using methods such as scalp electroencephalography (EEG) and magnetoencephalography (MEG). Not only are HFOs considered to be a useful biomarker of the SOZ, they also have the potential to gauge disease severity, monitor treatment, and evaluate prognostic outcomes. In this article, we review recent clinical research on noninvasively detected HFOs in the human brain, with a focus on epilepsy. Noninvasively detected scalp HFOs have been investigated in various types of epilepsy. HFOs have also been studied noninvasively in other pathologic brain disorders, such as migraine and autism. Herein, we discuss the challenges reported in noninvasive HFO studies, including the scarcity of MEG and high-density EEG equipment in clinical settings, low signal-to-noise ratio, lack of clinically approved automated detection methods, and the difficulty in differentiating between physiologic and pathologic HFOs. Additional studies on noninvasive recording methods for HFOs are needed, especially prospective multicenter studies. Further research is fundamental, and extensive work is needed before HFOs can routinely be assessed in clinical settings; however, the future appears promising.

Smartphone Motion Sensor-Based Complex Human Activity Identification Using Deep Stacked Autoencoder Algorithm for Enhanced Smart Healthcare System

Human motion analysis using a smartphone-embedded accelerometer sensor provided important context for the identification of static, dynamic, and complex sequence of activities. Research in smartphone-based motion analysis are implemented for tasks, such as health status monitoring, fall detection and prevention, energy expenditure estimation, and emotion detection. However, current methods, in this regard, assume that the device is tightly attached to a pre-determined position and orientation, which might cause performance degradation in accelerometer data due to changing orientation. Therefore, it is challenging to accurately and automatically identify activity details as a result of the complexity and orientation inconsistencies of the smartphone. Furthermore, the current activity identification methods utilize conventional machine learning algorithms that are application dependent. Moreover, it is difficult to model the hierarchical and temporal dynamic nature of the current, complex, activity identification process. This paper aims to propose a deep stacked autoencoder algorithm, and orientation invariant features, for complex human activity identification. The proposed approach is made up of various stages. First, we computed the magnitude norm vector and rotation feature (pitch and roll angles) to augment the three-axis dimensions (3-D) of the accelerometer sensor. Second, we propose a deep stacked autoencoder based deep learning algorithm to automatically extract compact feature representation from the motion sensor data. The results show that the proposed integration of the deep learning algorithm, and orientation invariant features, can accurately recognize complex activity details using only smartphone accelerometer data. The proposed deep stacked autoencoder method achieved 97.13% identification accuracy compared to the conventional machine learning methods and the deep belief network algorithm. The results suggest the impact of the proposed method to improve a smartphone-based complex human activity identification framework.

MEG Sensor Selection for Neural Speech Decoding

Direct decoding of speech from the brain is a faster alternative to current electroencephalography (EEG) speller-based brain-computer interfaces (BCI) in providing communication assistance to locked-in patients. Magnetoencephalography (MEG) has recently shown great potential as a non-invasive neuroimaging modality for neural speech decoding, owing in part to its spatial selectivity over other high-temporal resolution devices. Standard MEG systems have a large number of cryogenically cooled channels/sensors (200 - 300) encapsulated within a fixed liquid helium dewar, precluding their use as wearable BCI devices. Fortunately, recently developed optically pumped magnetometers (OPM) do not require cryogens, and have the potential to be wearable and movable making them more suitable for BCI applications. This design is also modular allowing for customized montages to include only the sensors necessary for a particular task. As the number of sensors bears a heavy influence on the cost, size, and weight of MEG systems, minimizing the number of sensors is critical for designing practical MEG-based BCIs in the future. In this study, we sought to identify an optimal set of MEG channels to decode imagined and spoken phrases from the MEG signals. Using a forward selection algorithm with a support vector machine classifier we found that nine optimally located MEG gradiometers provided higher decoding accuracy compared to using all channels. Additionally, the forward selection algorithm achieved similar performance to dimensionality reduction using a stacked-sparse-autoencoder. Analysis of spatial dynamics of speech decoding suggested that both left and right hemisphere sensors contribute to speech decoding. Sensors approximately located near Broca's area were found to be commonly contributing among the higher-ranked sensors across all subjects.

Variability analysis of epileptic EEG using the maximal overlap discrete wavelet transform

Purpose

To determine if there is a difference in the wavelet variances of seizure and non-seizure channels in the EEG of an epileptic subject.

Methods

A six-level decomposition was applied using the Maximal Overlap Discrete Wavelet Transform (MODWT). The wavelet variance and 95% CIs were calculated for each level of the decomposition. The number of changes in variance for each level were found using a change-point detection method of Whitcher. The Kruskal-Wallis test was used to determine if there were differences in the median number of change points within channels and across frequency bands (levels).

Results

No distinctive pattern was found for the wavelet variances to differentiate the seizure and non-seizure channels. The seizure channels tended to have lower variances for each level and overall, but this pattern only held for one of the three seizure channels (RAST4). The median number of change points did not differ between the seizure and non-seizure channels either within each channel or across the frequency bands.

Conclusion

The use of the MODWT in examining the variances and changes in variance did not show specific patterns which differentiate between seizure and non-seizure channels.

Automatic and Accurate Epilepsy Ripple and Fast Ripple Detection via Virtual Sample Generation and Attention Neural Networks

About 1% of the population around the world suffers from epilepsy. The success of epilepsy surgery depends critically on pre-operative localization of epileptogenic zones. High frequency oscillations including ripples (80-250 Hz) and fast ripples (250-500 Hz) are commonly used as biomarkers to localize epileptogenic zones. Recent literature demonstrated that fast ripples indicate epileptogenic zones better than ripples. Thus, it is crucial to accurately detect fast ripples from ripples signals of magnetoencephalography for improving outcome of epilepsy surgery. This paper proposes an automatic and accurate ripple and fast ripple detection method that employs virtual sample generation and neural networks with an attention mechanism. We evaluate our proposed detector on patient data with 50 ripples and 50 fast ripples labeled by two experts. The experimental results show that our new detector outperforms multiple traditional machine learning models. In particular, our method can achieve a mean accuracy of 89.3% and an average area under the receiver operating characteristic curve of 0.88 in 50 repeats of random subsampling validation. In addition, we experimentally demonstrate the effectiveness of virtual sample generation, attention mechanism, and architecture of neural network models.

Expert consensus on clinical applications of high-frequency oscillations in epilepsy

Studies in animal models of epilepsy and pre-surgical patients have unanimously found a strong correlation between high-frequency oscillations (HFOs, > 80 Hz) and the epileptogenic zone, suggesting that HFOs can be a potential biomarker of epileptogenicity and epileptogenesis. This consensus includes the definition and standard detection techniques of HFOs, the localizing value of pathological HFOs for epileptic foci, and different ways to distinguish physiological from epileptic HFOs. The latest clinical applications of HFOs in epilepsy and the related findings are also discussed. HFOs will advance our understanding of the pathophysiology of epilepsy.

Modeling Volatility Characteristics of Epileptic EEGs using GARCH Models

Objective: To determine if there was a difference in the volatility characteristics of seizure and non-seizure onset channels in the intracranial electroencephalogram (EEG) in a patient with temporal lobe epilepsy. Methods: The half-life of volatility for the different EEG channels was determined using Autoregressive Moving Average–Generalized Autoregressive Conditional Heteroscedasticity (ARMA–GARCH) models; confidence intervals were constructed using the delta method and an asymptotic method for comparing the half-lives. Results: Clinically determined seizure onsets occurred over strip electrodes named RAST (Right Anterior Subtemporal) and RMST (Right Mid Subtemporal), at locations 2, 3 and 4, on the strip electrodes. The half-lives of volatility for two of the three seizure channels, RAST3 and RAST4, were found to be significantly lower the rest of the channels for six one-minute EEG segments prior to seizure onset and nine one-minute EEG segments of an awake state. The half-lives of volatility for RAST3 and RAST4 were not significantly different to the non-seizure channels for ten one-minute segments of sleep and ten one-minute segments of sleep-to-awake states. The estimates for the half-lives were consistent for randomly selected one-minute EEG segments. Conclusions: The use of GARCH models may be a useful tool in determining hidden properties in epileptiform EEGs that may lead to better understanding of the seizure generating process.

EMS-Net: A Deep Learning Method for Autodetecting Epileptic Magnetoencephalography Spikes

Epilepsy is a neurological disorder characterized by sudden and unpredictable epileptic seizures, which incurs significant negative impacts on patients' physical, psychological and social health. A practical approach to assist with the clinical assessment and treatment planning for patients is to process magnetoencephalography (MEG) data to identify epileptogenic zones. As a widely accepted biomarker of epileptic foci, epileptic MEG spikes need to be precisely detected. Given that the visual inspection of spikes is time consuming, an automatic and efficient system with adequate accuracy for spike detection is valuable in clinical practice. However, current approaches for MEG spike autodetection are dependent on hand-engineered features. Here, we propose a novel multiview Epileptic MEG Spikes detection algorithm based on a deep learning Network (EMS-Net) to accurately and efficiently recognize the spike events from MEG raw data. The results of the leave-k-subject-out validation tests for multiple datasets (i.e., balanced and realistic datasets) showed that EMS-Net achieved state-of-the-art classification performance (i.e., accuracy: 91.82% - 99.89%; precision: 91.90% - 99.45%; sensitivity: 91.61% - 99.53%; specificity: 91.60% - 99.96%; f1 score: 91.70% - 99.48%; and area under the curve: 0.9688 - 0.9998).

Multiband entropy-based feature-extraction method for automatic identification of epileptic focus based on high-frequency components in interictal iEEG

Presurgical investigations for categorizing focal patterns are crucial, leading to localization and surgical removal of the epileptic focus. This paper presents a machine learning approach using information theoretic features extracted from high-frequency subbands to detect the epileptic focus from interictal intracranial electroencephalogram (iEEG). It is known that high-frequency subbands (>80 Hz) include important biomarkers such as high-frequency oscillations (HFOs) for identifying epileptic focus commonly referred to as the seizure onset zone (SOZ). In this analysis, the multi-channel interictal iEEG signals were splitted into segments and each segment was decomposed into multiple high-frequency subbands. The different types of entropy were calculated for each of the subbands and the sparse linear discriminant analysis (sLDA) was applied to select the prominent entropy features. Due to the imbalance of SOZ and non-SOZ channels in iEEG data, the use of machine learning techniques is always tricky. To deal with the imbalanced learning problem, an adaptive synthetic oversampling approach (ADASYN) with radial basis function kernel-based SVM was used to detect the focal segments. Finally, the epileptic focus was identified based on detection of focal segments on SOZ and non-SOZ channels. Eight patients were examined to observe the efficiency of the automatic detector. The experimental results and statistical tests indicate that the proposed automatic detector can identify the epileptic focus accurately and efficiently.

Kurtosis and skewness of high-frequency brain signals are altered in paediatric epilepsy

Intracranial studies provide solid evidence that high-frequency brain signals are a new biomarker for epilepsy. Unfortunately, epileptic (pathological) high-frequency signals can be intermingled with physiological high-frequency signals making these signals difficult to differentiate. Recent success in non-invasive detection of high-frequency brain signals opens a new avenue for distinguishing pathological from physiological high-frequency signals. The objective of the present study is to characterize pathological and physiological high-frequency signals at source levels by using kurtosis and skewness analyses. Twenty-three children with medically intractable epilepsy and age-/gender-matched healthy controls were studied using magnetoencephalography. Magnetoencephalographic data in three frequency bands, which included 2–80 Hz (the conventional low-frequency signals), 80–250 Hz (ripples) and 250–600 Hz (fast ripples), were analysed. The kurtosis and skewness of virtual electrode signals in eight brain regions, which included left/right frontal, temporal, parietal and occipital cortices, were calculated and analysed. Differences between epilepsy and controls were quantitatively compared for each cerebral lobe in each frequency band in terms of kurtosis and skewness measurements. Virtual electrode signals from clinical epileptogenic zones and brain areas outside of the epileptogenic zones were also compared with kurtosis and skewness analyses. Compared to controls, patients with epilepsy showed significant elevation in kurtosis and skewness of virtual electrode signals. The spatial and frequency patterns of the kurtosis and skewness of virtual electrode signals among the eight cerebral lobes in three frequency bands were also significantly different from that of the controls (2–80 Hz, P < 0.001; 80–250 Hz, P < 0.00001; 250–600 Hz, P < 0.0001). Compared to signals from non-epileptogenic zones, virtual electrode signals from epileptogenic zones showed significantly altered kurtosis and skewness (P < 0.001). Compared to normative data from the control group, aberrant virtual electrode signals were, for each patient, more pronounced in the epileptogenic lobes than in other lobes(kurtosis analysis of virtual electrode signals in 250–600 Hz; odds ratio = 27.9; P < 0.0001). The kurtosis values of virtual electrode signals in 80–250 and 250–600 Hz showed the highest sensitivity (88.23%) and specificity (89.09%) for revealing epileptogenic lobe, respectively. The combination of virtual electrode and kurtosis/skewness measurements provides a new quantitative approach to distinguishing pathological from physiological high-frequency signals for paediatric epilepsy. Non-invasive identification of pathological high-frequency signals may provide novel important information to guide clinical invasive recordings and direct surgical treatment of epilepsy.

A Long Short-Term Memory neural network for the detection of epileptiform spikes and high frequency oscillations

Over the last two decades, the evidence has been growing that in addition to epileptic spikes high frequency oscillations (HFOs) are important biomarkers of epileptogenic tissue. New methods of artificial intelligence such as deep learning neural networks can provide additional tools for automated analysis of EEG. Here we present a Long Short-Term Memory neural network for detection of spikes, ripples and ripples-on-spikes (RonS). We used intracranial EEG (iEEG) from two independent datasets. First dataset (7 patients) was used for network training and testing. The second dataset (5 patients) was used for cross-institutional validation. 1000 events of each class (spike, RonS, ripple and baseline) were selected from the candidates initially found using a novel threshold method. Network training was performed using random selections of 50-500 events (per class) from all patients from the 1st dataset. This 'global' network was then tested on other events for each patient from both datasets. The network was able to detect events with a good generalisability namely, with total accuracy and specificity for each class exceeding 90% in all cases, and sensitivity less than 86% in only two cases (82.5% for spikes in one patient and 81.9% for ripples in another patient). The deep learning networks can significantly accelerate the analysis of iEEG data and increase their diagnostic value which may improve surgical outcome in patients with localization-related intractable epilepsy.