Electromagnetic sources of biomarkers in magnetoencephalography (MEG) enables recognition of excessive discharges generators in epilepsy. Neurologists rely on MEG biomarker source localization for diagnostic purposes (presurgical epilepsy inquiry). Several ways are proposed to overcome the forward and inverse source localization problems. The goal of this study is to evaluate four distributed inverse problems: minimum norm estimation MNE, standardized lowresolution brain electromagnetic tomography sLORETA, wavelet maximum entropy on the mean wMEM, and dynamic statistical parametric maps dSPM used to define network connectivity of oscillatory epileptic events. We employed Jmail et al.'s 2016 pre-processing chain to assess epileptic oscillations among MEG sources in pharmaco-resistant patients. Then, we explored granger causality and spatial granger causality metrics among extended active sources for each inverse approach. Granger causality determines causal influence of one time series on another, providing insights into directional flow of information between brain regions. Spatial Granger causality extends this concept to incorporate spatial information, evaluating how brain activity in one region can predict future activity in another region, accounting of brain spatial structure. As a result, for granger causality, wMEM method identifies strong connections between active sources, suggesting robust interaction and influence among these regions. In contrast, dSPM method reveals weaker connections, indicating a less pronounced directional influence between active sources. For spatial granger causality wMEM again demonstrates strong connections between active sources, highlighting its sensitivity and effectiveness in detecting robust functional interactions. On the other hand, MNE and sLORETA methods exhibit weaker connections, suggesting that these techniques may be less effective in capturing directional influences between active sources. These findings underscore the importance of using multiple localization techniques to to accurately determine interictal MEG oscillation locations and epileptic zones.

Electroencephalogram (EEG) is a widely used supplementary test in epilepsy diagnosis. While traditional methods and deep learning networks have shown considerable success, they often rely on extensive human effort, particularly in feature extraction. This study proposes a hybrid Convolutional Neural Network and Recurrent Neural Network (CNN-RNN) algorithm that can autonomously learn features from EEG data. The RNN component calculates dependency and continuity features from the CNN's intermediate layer output, which are then connected to the final fully connected network for classification prediction. The algorithm was trained and tested on four datasets from the University of Bonn. For binary classification tasks, the proposed CNN-RNN model achieved accuracies of 98.33% on dataset 1 and 100% on dataset 2. When tested on datasets 3 and 4, the model attained 98.00% and 100% accuracy, respectively. Notably, the algorithm demonstrated superior performance on datasets 2 and 4 compared to state-of-the-art methods. These results suggest that the proposed CNN-RNN model offers a promising approach for automated EEG analysis in epilepsy diagnosis, potentially reducing the need for manual feature extraction and improving classification accuracy.

High frequency oscillations (HFO) from, MEG (magnetoencephalography) and intracerebral EEG are considered as effective tools to identify cognitive status and several cortical disorders especially in epilepsy diagnosis. The aim of our study is to evaluate stationary wavelet transform (SWT) technique performance in efficient reconstruction of pure epileptic high frequency oscillations, reputed as biomarkers of epileptogenic zones: generators of inter ictal epileptic discharges, and offhand seizures. We applied SWT on simulated and real database to detect non-contaminated HFO by spiky element. For simulated data, we computed the GOF of reconstruction that reaches for all studied constraint (relative amplitude, frequency, SNR and overlap) a promising results. For real data we used time frequency domain to evaluate SWT robustness of HFO reconstruction. We proved that SWT is an efficient filtering technique for separation HFO from spiky events. Our results would have an important impact on the definition of epileptogenic zones.

Locating electromagnetic sources among magnetoencephalography (MEG) allows the definition of responsible generators of excessive discharges in epilepsy. Source localization of MEG biomarkers is considered as a diagnostic aid for neurologists (pre surgical investigation of epilepsy ). Several techniques are proposed to resolve forward and inverse problem of source localization. Our goal in this study is to compare three distributed methods of inverse problem: MNE, sLORETA and dSPM in defining networks connectivity of spiky epileptic events. We used a pre processing chain to evaluate the rate of epileptic spikes connectivity among MEG for five pharmaco resistant patients and two groups of spiky events. For each inverse technique, we calculated the cross correlation between active sources, in fact dsPM shows the highest level of connectivity, MNE gives also a connectivity between the entire active sources but with lowest rate then sLORETA finally dsPM depicts a low number of connection between active regions. These results promote the combination of several localization methods during the investigation of epileptogenics zones.

The emergency of embedded systems puts new challenges for the design of different system in many fields. One of the embedded application's fields is the medical one. The major difficulty is the embedded system's reduced energy and computational resources that must be carefully used to execute complex application often in unpredictable environments. In this paper, the used application is the detection of evoked potential with variable latency and multiple trials using consensus matching pursuit. Fitting to the noisy Evoked Potential (EP) signal persistent in all response, we use the Consensus version of the matching pursuit algorithm (CMP). EP is a resulted wave from a stimulus. The EP can be explained with a good quality of energy ratio factor (QR). If we use a noisy EP, we cannot reconstruct the original data because of the random atoms of CMP dictionary. We select the significant atoms to rebuild and EP signals. This application is embedded on a Xilinx ML 507. We used an adaptive architecture based on dynamically partial reconfiguration.

The current trend in the diagnosis of epilepsy is oriented towards high frequency activity that is used to characterize the epileptogenic zone. Still, interictal spikes, i.e. transient activity, remain a classical marker. A difficult question is to separate these two classes of activity that overlap in the time-frequency plane. We propose to use a temporal basis set in order to describe and subtract spiking activity from the traces, as a preprocessing step for further space-time-frequency analysis. The set is based on singular value decomposition. We select the first three temporal components, construct a model of the spikes by projecting the data on the new basis and obtain a spike-free signal by subtracting the model from the original signal. We prove that despikifying results in a better characterization of oscillatory activity, promoting time and frequency separation. The resulting spatio-temporal maps could help for a better characterization of the oscillatory activity in electrophysiology of epilepsy.