Zurich iEEG HFO Dataset

This dataset was obtained from the publication [1].

There are 20 subjects with HFO events. We converted the dataset into BIDS format. The channels that were included in the resected region and channels excluded from analysis are included in the clinical Excel file under the sourcedata/ directory. The channels were extracted from the Supplementary table at: https://static-content.springer.com/esm/art%3A10.1038%2Fs41598-017-13064-1/MediaObjects/41598_2017_13064_MOESM1_ESM.pdf.

The original uploader: adam2392 obtained explicit permission from the authors of the dataset to upload this to openneuro. Adam worked on an open-source Python implementation of HFO detection algorithms, and uses this dataset in validation. Even though the publication involves a Morphology HFO detector, we have implemented our interpretation of the RMS, LineLength and Hilbert detectors in the mne-hfo repository [2].For more information, visit: https://github.com/mne-tools/mne-hfo.

Note from the paper

"We excluded all electrode contacts where electrical stimulation evoked motor or language responses (Table S1). In TLE patients, we included only the 3 most mesial bipolar channels".

BIDS Conversion

MNE-BIDS was used to convert the dataset into BIDS format. The code inside code/ was used to generate the data.

HFO Events From Original Paper

The HFO events from the original paper that were validated and detected are stored in the *events.tsv file per dataset run. The format is similar to mne-hfo and can be easily read in using mne-bids and/or mne-python.

Each row in the events.tsv file corresponds to a HFO detected in the original source dataset. The trial_type column stores the information pertaining type of HFO (e.g. ripple, fr for fast ripple, or frandr for fast ripple and ripple). The channel name (possibly in bipolar reference) is "-" character delimited and appended to the type of HFO with a "_" separating. For example: <hfo_type>_<channel_name> is the form.

Reference Dataset

The following website was where the original data was downloaded.

http://crcns.org/data-sets/methods/ieeg-1

References

[1] Fedele T, Burnos S, Boran E, Krayenbühl N, Hilfiker P, Grunwald T, Sarnthein J. Resection of high frequency oscillations predicts seizure outcome in the individual patient. Scientific Reports. 2017;7(1):13836. https://www.nature.com/articles/s41598-017-13064-1 doi:10.1038/s41598-017-13064-1

[2] Dataset meta analysis with mne-hfo. 10.5281/zenodo.4485036

[3] Appelhoff, S., Sanderson, M., Brooks, T., Vliet, M., Quentin, R., Holdgraf, C., Chaumon, M., Mikulan, E., Tavabi, K., Höchenberger, R., Welke, D., Brunner, C., Rockhill, A., Larson, E., Gramfort, A. and Jas, M. (2019). MNE-BIDS: Organizing electrophysiological data into the BIDS format and facilitating their analysis. Journal of Open Source Software 4: (1896). https://doi.org/10.21105/joss.01896

[4] Holdgraf, C., Appelhoff, S., Bickel, S., Bouchard, K., D'Ambrosio, S., David, O., … Hermes, D. (2019). iEEG-BIDS, extending the Brain Imaging Data Structure specification to human intracranial electrophysiology. Scientific Data, 6, 102. https://doi.org/10.1038/s41597-019-0105-7

SWEC iEEG Dataset

  Dataset summary

The SWEC iEEG Dataset contains fully anonymised multi-channel iEEG recordings collected from a total of 68 subjects suffering from pharmacoresistent epilepsy undergoing pre-surgical evaluation for epilepsy. The data was recorded at the Sleep Wake Epilepsy Center (SWEC) of the Department of Neurology at the Inselspital in Bern, Switzerland. The dataset includes a total of 9328 hours of signal and 704 ictal events, annotated by… See the full description on the dataset page: https://huggingface.co/datasets/NeuroTec/SWEC_iEEG_Dataset.

Dataset description This dataset is part of a bigger dataset of intracranial EEG (iEEG) called RESPect (Registry for Epilepsy Surgery Patients), a dataset recorded at the University Medical Center of Utrecht, the Netherlands. It consists of 12 patients: six patients recorded intraoperatively using electrocorticography (acute ECoG), six patients with long-term recordings (3 patients recorded with ECoG and 3 patients recorded with stereo-encephalography SEEG). For a detailed description see (Demuru M, van Blooijs D, Zweiphenning W, Hermes D, Leijten F, Zijlmans M, on behalf of the RESPect group. “A Practical Workflow for Organizing Clinical Intraoperative and Long-Term iEEG data in BIDS”.).

This data is organized according to the Brain Imaging Data Structure specification. A community- driven specification for organizing neurophysiology data along with its metadata. For more information on this data specification, see https://bids-specification.readthedocs.io/en/stable/

Each patient has their own folder (e.g., sub-RESP0280) which contains the iEEG recordings data for that patient, as well as the metadata needed to understand the raw data and event timing.

Two different implementation of the BIDS structure were done according to the different type of recordings (i.e. intraoperative or long-term) Intraoperative ECoG Surgery with intraoperative ECoG is composed of three main situations that can be logically grouped into BIDS sessions:

• Pre-resection sessions, consisting of all recordings (with different configurations of the grid and strips/depth) carried out before the surgeon has started the planned resection.

• Intermediate sessions, consisting of all subsequent recordings performed before any iterative extension of the resection area.

• Post-resection sessions, consisting of all the recordings performed after the last resection.

Each situation is labelled with an increasing number starting from 1, indicative of the period in time respective to the surgical resection and a consecutive letter (starting from A) indicative of the position of the grid and strip/depth for a given session. As an example see patient RESP0280 who had 4 sessions recorded: two pre-resection sessions, one intermediate sessions and one post-resection session. The first session is SITUATION1A consisting of the first recording, then the grid was moved to another position, resulting in SITUATION1B. After that, the surgeon resected part of the brain and then there was another recording(SITUATION2A). Finally the surgeon applied a resection for the last time and the recording after that was defined as SITUATION3A.

Long-term iEEG In long-term recordings, data that are recorded within one monitoring period are logically grouped in the same BIDS session and stored across runs indicating the day and time point of recording in the monitoring period. If extra electrodes were added/removed during this period, the session was divided into different sessions (e.g. ses-1A and ses-1b). We use the optional run key-value pair to specify the day and the start time of the recording (e.g. run-021315, day 2 after implantation, which is day 1 of the monitoring period, at 13:15). The task key-value pair in long-term iEEG recordings describes the patient’s state during the recording of this file. Different tasks have been defined, such as “rest” when a patient is awake but not doing a specific task, “sleep” when a patient is sleeping the majority of the file, or “SPESclin” when the clinical SPES protocol has been performed in this file. Other task definitions can be found in the annotation syntax (https://github.com/UMCU-EpiLAB/umcuEpi_longterm_ieeg_respect_bids/master/manuals/IFU_annotatingtrc_ECoG).

License This dataset is made available under the Public Domain Dedication and License CC v1.0, whose full text can be found at https://creativecommons.org/publicdomain/zero/1.0/. We hope that all users will follow the ODC Attribution/Share-Alike Community Norms (http://www.opendatacommons.org/norms/odc-by-sa/); in particular, while not legally required, we hope that all users of the data will acknowledge by citing Demuru M, van Blooijs D, Zweiphenning W, Hermes D, Leijten F, Zijlmans M, on behalf of the RESPect group. “A Practical Workflow for Organizing Clinical Intraoperative and Long-Term iEEG data in BIDS”. Submitted to Neuroinformatics. in any publications.

Code available at: https://github.com/UMCU-EpiLAB.

Acknowledgements We would like to thank the patients for providing their data for this dataset, the RESPect team of University Medical Center of Utrecht, for the acquisition of the dataset. Please cite Demuru M, van Blooijs D, Zweiphenning W, Hermes D, Leijten F, Zijlmans M, on behalf of the RESPect group. “A Practical Workflow for Organizing Clinical Intraoperative and Long-Term iEEG data in BIDS”. Submitted to Neuroinformatics. in any publications.

Hundreds of 90-s iEEG records are typically captured from each NeuroPace RNS System patient between clinic visits. While these records provide invaluable information about the patient’s electrographic seizure and interictal activity patterns, manually classifying them into electrographic seizure/non-seizure activity, and manually identifying the seizure onset channels and times is an extremely time-consuming process. A convolutional neural network based Electrographic Seizure Classifier (ESC) model was developed in an earlier study. In this study, the classification model is tested against iEEG annotations provided by three expert reviewers board certified in epilepsy. The three experts individually annotated 3,874 iEEG channels from 36, 29, and 35 patients with leads in the mesiotemporal (MTL), neocortical (NEO), and MTL + NEO regions, respectively. The ESC model’s seizure/non-seizure classification scores agreed with the three reviewers at 88.7%, 89.6%, and 84.3% which was similar to how reviewers agreed with each other (92.9%–86.4%). On iEEG channels with all 3 experts in agreement (83.2%), the ESC model had an agreement score of 93.2%. Additionally, the ESC model’s certainty scores reflected combined reviewer certainty scores. When 0, 1, 2 and 3 (out of 3) reviewers annotated iEEG channels as electrographic seizures, the ESC model’s seizure certainty scores were in the range: [0.12–0.19], [0.32–0.42], [0.61–0.70], and [0.92–0.95] respectively. The ESC model was used as a starting-point model for training a second Seizure Onset Detection (SOD) model. For this task, seizure onset times were manually annotated on a relatively small number of iEEG channels (4,859 from 50 patients). Experiments showed that fine-tuning the ESC models with augmented data (30,768 iEEG channels) resulted in a better validation performance (on 20% of the manually annotated data) compared to training with only the original data (3.1s vs 4.4s median absolute error). Similarly, using the ESC model weights as the starting point for fine-tuning instead of other model weight initialization methods provided significant advantage in SOD model validation performance (3.1s vs 4.7s and 3.5s median absolute error). Finally, on iEEG channels where three expert annotations of seizure onset times were within 1.5 s, the SOD model’s seizure onset time prediction was within 1.7 s of expert annotation.

2 females

This dataset was collected for a study to replicate previous work on hippocampal physiology predictive of successful encoding using an associative memory paradigm in a surgical intracranial electroencephalography (iEEG) group and extend those investigations to a chronic ambulatory iEEG population using RNS System devices (NeuroPace, Inc.). The primary objective was to compare the hippocampal gamma response during associative learning in the RNS System to the gold standard of conventional iEEG recordings. Subjects in this study were asked to complete a face-profession association task. To measure the hippocampal physiology of associative learning, the researcher performed this face-professional paradigm with five patients with convention iEEG and with three patients with chronic iEEG monitored by RNS System, adapting the devices to permit task synchronization to the RNS System iEEG without modifying the clinical system.

Eligible subjects for the study were epilepsy patients between the ages of 18 and 70 years old with previously implanted RNS Systems for treatment of refractory focal epilepsy, had the RNS System with at least one hippocampal depth lead, had a Full-Scale Intelligence Quotient (FSIQ) greater than 70, were able to provide informed consent, and were native English speakers. Patients undergoing surgical evaluation with iEEG monitoring for epilepsy surgery at NYU Langone Health provided comparison data.

We present an electrophysiological dataset recorded from nine subjects during a verbal working memory task. Subjects were epilepsy patients undergoing intracranial monitoring for localization of epileptic seizures. Subjects performed a modified Sternberg task in which the encoding of memory items, maintenance, and recall were temporally separated. The dataset includes simultaneously recorded scalp EEG with the 10-20 system, intracranial EEG (iEEG) recorded with depth electrodes, waveforms and spike times of 1526 units recorded in the medial temporal lobe, and the MNI coordinates and anatomical labels of all intracranial electrodes. Subject characteristics and information on sessions (set size, match/mismatch, correct/incorrect, response, response time for each trial) are also provided. This dataset enables the investigation of working memory by providing simultaneous scalp EEG and iEEG recordings, which can be used for connectivity analysis, alongside hard to obtain unit recordings from humans.

Dataset of intracranial EEG from human epilepsy patients performing a visuospatial working memory task

Description:

We present an electrophysiological dataset recorded from ten subjects during a visuospatial working memory task. Subjects were epilepsy patients undergoing intracranial monitoring for localization of epileptic seizures. Subjects completed 60 trials (five sessions) of Memory Matrix - a visuospatial working memory game on the Lumosity platform (https://www.lumosity.com/; Lumos Labs, Inc, San Francisco, CA) - during interictal iEEG recording.

Repository structure:

Main directory (iEEG from children during gameplay) Contains iEEG files of each participant in the study. Folders are explained below.

Subfolders:

1. sub-/: Contains folders for each subject, named sub- and session information.
2. sub-/ses-: Contains folders for base and task.
3. sub-/ses-/ieeg/: Contains the raw iEEG data in .edf format for each subject. Each subject performed 60 working memory trials (ses-task). Each *ieeg.edf file contains continuous iEEG data during the working memory task. Details about the channels are given in the corresponding .tsv file. We also provide the information on the timing of the stimulus onset and finger tapping on ieeg/edf file by specifying the start and end sample of each trial. (101 is for task display, 401 is for finger tapping to the successful grid, and 501 is for finger tapping to the failed grid). Each subject also had baseline periods (ses-base). To establish baseline, we selected 60 non-overlapping 2,000-ms time windows during periods of spontaneous, resting, eye-open wakefulness immediately preceding the game sessions.

The exquisite spatiotemporal precision of human intracranial EEG recordings (iEEG) permits characterizing neural processing with a level of detail that is inaccessible to scalp-EEG, MEG, or fMRI. However, the same qualities that make iEEG an exceptionally powerful tool also present unique challenges. Until now, the fusion of anatomical data (MRI and CT images) with the electrophysiological data and its subsequent analysis has relied on technologically and conceptually challenging combinations of software. Here, we describe a comprehensive protocol that addresses the complexities associated with human iEEG, providing complete transparency and flexibility in the evolution of raw data into illustrative representations. The protocol is directly integrated with an open source toolbox for electrophysiological data analysis (FieldTrip). This allows iEEG researchers to build on a continuously growing body of scriptable and reproducible analysis methods that, over the past decade, have been developed and employed by a large research community. We demonstrate the protocol for an example complex iEEG data set to provide an intuitive and rapid approach to dealing with both neuroanatomical information and large electrophysiological data sets. We explain how the protocol can be largely automated and readily adjusted to iEEG data sets with other characteristics. The protocol can be implemented by a graduate student or post-doctoral fellow with minimal MATLAB experience and takes approximately an hour, excluding the automated cortical surface extraction.

This collection contains the data described in the protocol and that can be used to replicate all results.

Data for the manuscript titled "Utility of iEEG networks in modeling brain activity depends on re-referencing and connectivity choice"

HUP iEEG dataset

This dataset was prepared for release as part of a manuscript by Bernabei & Li et al. (in preparation). A subset of the data has been featured in Kini & Bernabei et al., Brain (2019) [1], and Bernabei & Sinha et al., Brain (2022) [2].

Dataset description

These files contain de-identified patient data collected as part of surgical treatment for drug resistant epilepsy at the Hospital of the University of Pennsylvania. Each of the 58 subjects underwent intracranial EEG with subdural grid, strip, and depth electrodes (ECoG) or purely stereotactically-placed depth electrodes (SEEG). Each patient also underwent subsequent treatment with surgical resection or laser ablation. Electrophysiologic data for both interictal and ictal periods is available, as are electrode localizations in ICBM152 MNI space. Furthermore, clinically-determined seizure onset channels are provided, as are channels which overlap with the resection/ablation zone, which was rigorously determined by segmenting the resection cavity.

BIDS Conversion

MNE-BIDS was used to convert the dataset into BIDS format.

References

[1] Kini L.*, Bernabei J.M.*, Mikhail F., Hadar P., Shah P., Khambhati A., Oechsel K., Archer R., Boccanfuso J.A., Conrad E., Stein J., Das S., Kheder A., Lucas T.H., Davis K.A., Bassett D.S., Litt B., Virtual resection predicts surgical outcome for drug resistant epilepsy. Brain, 2019.

[2] Bernabei J.M.*, Sinha N.*, Arnold T.C., Conrad E., Ong I., Pattnaik A.R., Stein J.M., Shinohara R.T., Lucas T.H., Bassett D.S., Davis K.A., Litt B., Normative intracranial EEG maps epileptogenic tissues in focal epilepsy. Brain, 2022

[3] Appelhoff, S., Sanderson, M., Brooks, T., Vliet, M., Quentin, R., Holdgraf, C., Chaumon, M., Mikulan, E., Tavabi, K., Höchenberger, R., Welke, D., Brunner, C., Rockhill, A., Larson, E., Gramfort, A. and Jas, M. (2019). MNE-BIDS: Organizing electrophysiological data into the BIDS format and facilitating their analysis. Journal of Open Source Software 4: (1896). https://doi.org/10.21105/joss.01896

[4] Holdgraf, C., Appelhoff, S., Bickel, S., Bouchard, K., D'Ambrosio, S., David, O., … Hermes, D. (2019). iEEG-BIDS, extending the Brain Imaging Data Structure specification to human intracranial electrophysiology. Scientific Data, 6, 102. https://doi.org/10.1038/s41597-019-0105-7

This dataset was utilized for the publication of the manuscript by Zhang et al. (in preparation). A subset of the data has been employed in [1] and [2].

Summary: This data set comprises the de-identified subjects with interictal iEEG recordings with sleep from University of California Los Angels Mattel Children’s Hospital, and Children’s Hospital of Michigan, Detroit. Subject-wise information is contained in each folder, including iEEGs collected from 185 subjects during sleep. The channel name and valuables, such as the anatomical label and the resection status, are attached to each folder. The outcome and background information of all the subjects are summarized in ‘paticipant.tsv’ located in the parental directory.

Derivatives The processed data for HFO detection and classification are shown in the derivatives/folder. The HFO analysis contains detection from two methods: RMS and MNI detectors.

Ref

[1] Zhang Y, Lu Q, Monsoor T, et al. Refining epileptogenic high-frequency oscillations using deep learning: a reverse engineering approach. Brain Commun. 2022;4(1):fcab267. doi:10.1093/braincomms/fcab267

[2] Kuroda N, Sonoda M, Miyakoshi M, et al. Objective interictal electrophysiology biomarkers optimize prediction of epilepsy surgery outcome. Brain Commun. 2021;3(2):fcab042. doi:10.1093/braincomms/fcab042

This dataset was curated for publication as part of the manuscript in Sakakura et al. (in preparation). It contains iEEGs collected from 114 individuals during slow wave sleep. The available Matlab code can be found at https://github.com/kaz1126/MI_HFO. The iEEG coordinate system employed in this dataset is MNI305.

Fragility Multi-Center Retrospective Study

iEEG and EEG data from 5 centers is organized in our study with a total of 100 subjects. We publish 4 centers' dataset here due to data sharing issues.

Acquisitions include ECoG and SEEG. Each run specifies a different snapshot of EEG data from that specific subject's session. For seizure sessions, this means that each run is a EEG snapshot around a different seizure event.

For additional clinical metadata about each subject, refer to the clinical Excel table in the publication.

Data Availability

NIH, JHH, UMMC, and UMF agreed to share. Cleveland Clinic did not, so requires an additional DUA.

All data, except for Cleveland Clinic was approved by their centers to be de-identified and shared. All data in this dataset have no PHI, or other identifiers associated with patient. In order to access Cleveland Clinic data, please forward all requests to Amber Sours, SOURSA@ccf.org:

Amber Sours, MPH Research Supervisor | Epilepsy Center Cleveland Clinic | 9500 Euclid Ave. S3-399 | Cleveland, OH 44195 (216) 444-8638

You will need to sign a data use agreement (DUA).

Sourcedata

For each subject, there was a raw EDF file, which was converted into the BrainVision format with mne_bids. Each subject with SEEG implantation, also has an Excel table, called electrode_layout.xlsx, which outlines where the clinicians marked each electrode anatomically. Note that there is no rigorous atlas applied, so the main points of interest are: WM, GM, VENTRICLE, CSF, and OUT, which represent white-matter, gray-matter, ventricle, cerebrospinal fluid and outside the brain. WM, Ventricle, CSF and OUT were removed channels from further analysis. These were labeled in the corresponding BIDS channels.tsv sidecar file as status=bad. The dataset uploaded to openneuro.org does not contain the sourcedata since there was an extra anonymization step that occurred when fully converting to BIDS.

Derivatives

Derivatives include: * fragility analysis * frequency analysis * graph metrics analysis * figures

These can be computed by following the following paper: Neural Fragility as an EEG Marker for the Seizure Onset Zone

Events and Descriptions

Within each EDF file, there contain event markers that are annotated by clinicians, which may inform you of specific clinical events that are occuring in time, or of when they saw seizures onset and offset (clinical and electrographic).

During a seizure event, specifically event markers may follow this time course:

* eeg onset, or clinical onset - the onset of a seizure that is either marked electrographically, or by clinical behavior. Note that the clinical onset may not always be present, since some seizures manifest without clinical behavioral changes.
* Marker/Mark On - these are usually annotations within some cases, where a health practitioner injects a chemical marker for use in ICTAL SPECT imaging after a seizure occurs. This is commonly done to see which portions of the brain are active metabolically.
* Marker/Mark Off - This is when the ICTAL SPECT stops imaging.
* eeg offset, or clinical offset - this is the offset of the seizure, as determined either electrographically, or by clinical symptoms.

Other events included may be beneficial for you to understand the time-course of each seizure. Note that ICTAL SPECT occurs in all Cleveland Clinic data. Note that seizure markers are not consistent in their description naming, so one might encode some specific regular-expression rules to consistently capture seizure onset/offset markers across all dataset. In the case of UMMC data, all onset and offset markers were provided by the clinicians on an Excel sheet instead of via the EDF file. So we went in and added the annotations manually to each EDF file.

Seizure Electrographic and Clinical Onset Annotations

For various datasets, there are seizures present within the dataset. Generally there is only one seizure per EDF file. When seizures are present, they are marked electrographically (and clinically if present) via standard approaches in the epilepsy clinical workflow.

Clinical onset are just manifestation of the seizures with clinical syndromes. Sometimes the maker may not be present.

Seizure Onset Zone Annotations

What is actually important in the evaluation of datasets is the clinical annotations of their localization hypotheses of the seizure onset zone.

These generally include:

* early onset: the earliest onset electrodes participating in the seizure that clinicians saw
* early/late spread (optional): the electrodes that showed epileptic spread activity after seizure onset. Not all seizures has spread contacts annotated.

Surgical Zone (Resection or Ablation) Annotations

For patients with the post-surgical MRI available, then the segmentation process outlined above tells us which electrodes were within the surgical removed brain region.

Otherwise, clinicians give us their best estimate, of which electrodes were resected/ablated based on their surgical notes.

For surgical patients whose postoperative medical records did not explicitly indicate specific resected or ablated contacts, manual visual inspection was performed to determine the approximate contacts that were located in later resected/ablated tissue. Postoperative T1 MRI scans were compared against post-SEEG implantation CT scans or CURRY coregistrations of preoperative MRI/post SEEG CT scans. Contacts of interest in and around the area of the reported resection were selected individually and the corresponding slice was navigated to on the CT scan or CURRY coregistration. After identifying landmarks of that slice (e.g. skull shape, skull features, shape of prominent brain structures like the ventricles, central sulcus, superior temporal gyrus, etc.), the location of a given contact in relation to these landmarks, and the location of the slice along the axial plane, the corresponding slice in the postoperative MRI scan was navigated to. The resected tissue within the slice was then visually inspected and compared against the distinct landmarks identified in the CT scans, if brain tissue was not present in the corresponding location of the contact, then the contact was marked as resected/ablated. This process was repeated for each contact of interest.

References

• Adam Li, Chester Huynh, Zachary Fitzgerald, Iahn Cajigas, Damian Brusko, Jonathan Jagid, Angel Claudio, Andres Kanner, Jennifer Hopp, Stephanie Chen, Jennifer Haagensen, Emily Johnson, William Anderson, Nathan Crone, Sara Inati, Kareem Zaghloul, Juan Bulacio, Jorge Gonzalez-Martinez, Sridevi V. Sarma. Neural Fragility as an EEG Marker of the Seizure Onset Zone. bioRxiv 862797; doi: https://doi.org/10.1101/862797

• Appelhoff, S., Sanderson, M., Brooks, T., Vliet, M., Quentin, R., Holdgraf, C., Chaumon, M., Mikulan, E., Tavabi, K., Höchenberger, R., Welke, D., Brunner, C., Rockhill, A., Larson, E., Gramfort, A. and Jas, M. (2019). MNE-BIDS: Organizing electrophysiological data into the BIDS format and facilitating their analysis. Journal of Open Source Software 4: (1896). https://doi.org/10.21105/joss.01896

• Holdgraf, C., Appelhoff, S., Bickel, S., Bouchard, K., D'Ambrosio, S., David, O., … Hermes, D. (2019). iEEG-BIDS, extending the Brain Imaging Data Structure specification to human intracranial electrophysiology. Scientific Data, 6, 102. https://doi.org/10.1038/s41597-019-0105-7

• Pernet, C. R., Appelhoff, S., Gorgolewski, K. J., Flandin, G., Phillips, C., Delorme, A., Oostenveld, R. (2019). EEG-BIDS, an extension to the brain imaging data structure for electroencephalography. Scientific Data, 6, 103. https://doi.org/10.1038/s41597-019-0104-8

This dataset contains iEEG neural recordings and mood measurements from adult human subjects undergoing inpatient monitoring for seizure localization. The experimental design, including explanation of metrics of interest, is detailed in Rao et al (2018) Current Biology. Each subject received an identifier: EC##. Electrode coverage varied across subjects and was solely dictated by clinical need. Note, each probe (with multiple channels) was named by the anatomical target of the probe. The spacing of electrodes and size of targets means that not all electrodes within a given probe will contact tissue in the anatomical target. MRI and CT reconstructions allowed us to assign the activity recorded on specific electrodes to different brain regions of interest.

The repository contains analysis scripts and data used in Wang Z, Magnotti J, Beauchamp MS, Li M. Functional Group Bridge for Simultaneous Regression and Support Estimation, 2020. The data contains high-gamma brain responses across 8 subjects from "congruency" audio-visual experiment.

Eight mixed-breed canines with naturally occurring epilepsy were implanted with a mobile iEEG recording device and monitored continuously for multiple months. Four dogs had an inadequate number of seizures for algorithm training and testing. Lead seizures are defined as seizures separated by a minimum of 4 hours. Dogs with fewer than 5 lead seizures (italicized) were excluded from analysis. Dog 1 (Buck) died after approximately a year of iEEG monitoring.Testing data.

Processed data and code for reproducing the main results and figures of the paper "Fluctuations in EEG band power at subject-specific timescales over minutes to days explain changes in seizure evolutions".

We analysed publicly available data from subjects with drug-resistant focal epilepsy. A total of 2656 hours of long-term intracranial electroencephalography (iEEG) from 18 subjects was obtained using the "The SWEC-ETHZ iEEG Database and Algorithms" (available at http://ieeg-swez.ethz.ch) (Burrello et al., 2019).

Reference A. Burrello, L. Cavigelli, K. Schindler, L. Benini, A. Rahimi, ‘‘Laelaps: An Energy-Efficient Seizure Detection Algorithm from Long-term Human iEEG Recordings without False Alarms’’ in proceedings of the ACM/IEEE Design, Automation, and Test in Europe Conference (DATE), Florence, Italy, March 25-29, 2019.

The dataset used in this study consists of 8 neurosurgical patients who underwent intracranial electroencephalography (iEEG) recordings while listening to speech.

Dataset for event encoded analog EEG signals for detection of Epileptic seizures

This dataset contains events that are encoded from the analog signals recorded during pre-surgical evaluations of patients at the Sleep-Wake-Epilepsy-Center (SWEC) of the University Department of Neurology at the Inselspital Bern. The analog signals are sourced from the SWEC-ETHZ iEEG Database

This database contains event streams for 10 seizures recorded from 5 patients and generated by the DYnamic Neuromorphic Asynchronous Processor (DYNAP-SE2) to demonstrate a proof-of-concept of encoding seizures with network synchronization. The pipeline consists of two parts (I) an Analog Front End (AFE) and (II) an SNN termed as"Non-Local Non-Global" (NLNG) network.

In the first part of the pipeline, the digitally recorded signals from SWEC-ETHZ iEEG Database are converted to analog signals via an 18-bit Digital-to-Analog converter (DAC) and then amplified and encoded into events by an Asynchronous Delta Modulator (ADM). Then in the second part, the encoded event streams are fed into the SNN that extracts the features of the epileptic seizure by extracting the partial synchronous patterns intrinsic to the seizure dynamics.

Details about the neuromorphic processing pipeline and the encoding process are included in a manuscript under review. The preprint is available in bioRxiv

InstallationThe installation requires Python>=3.x and conda (or py-venv) package. Users can then install the requirements inside a conda environment using

conda env create -f requirements.txt -n sez

Once created the conda environment can be activated with conda activate sez

The main files in the database are described in the hierarchy below.

EventSezDataset/

├─ data/

│ ├─ P x S x

│ │ ├─ Pat x Sz x _CH x .csv

├─ LSVM_Params/

│ ├─ opt_svm_params/

│ ├─ pat_x_features_SYNCH/

├─ fig_gen.py

├─ sync_mat_gen.py

├─ SeizDetection_FR.py

├─ SeizDetection_SYNCH.py

├─ support.py

├─ run.sh

├─ requirements.txt

where x represents the Patient ID and the Seizure ID respectively.

requirements.txt: This file lists the requirements for the execution of the Python code.

fig_gen.py: This file plots the analog signals and the associated AFE and NLNG event streams. The execution of the code happens with `python fig_gen.py 1 1 13', where patient 2, seizure 1, and channel 13 of the recording are plotted.

sync_mat_gen.py: This file describes the function for plotting the synchronization matrices emerging from the ADM and the NLNG spikes with either linear or log colorbar. The execution of the code happens with python sync_mat_gen.py 1 1' orpython sync_mat_gen.py 1 1 log'. This execution generated four figures for pre-seizure, First Half of seizure, Second Half of seizure, and post-seizure time periods, where patient 1 and seizure 1. The third option can either be left blank or input as lin or log, for respective color bar scales. The time is the signal-time as mentioned in the table below.

run.sh: A simple Linux script to run the above code for all patients and seizures.

SeizDetection_FR.py: This file runs the LSVM on the ADM and NLNG spikes, using the firing rate (FR) as a feature. The code is currently set up with plotting with pre-computed features (in the LSVM_Params/opt_svm_params/ folder). Users can use the code for training the LSVM with different parameters as well.

SeizDetection_SYNCH.py: This file runs the LSVM on the kernelized ADM and NLNG spikes, using the flattened SYNC matrices as a feature. The code is currently set up with plotting with pre-computed features (in the LSVM_Params/pat_x_features_SYNCH/ folder). Users can use the code for training the LSVM with different parameters as well.

LSVM_Params: Folder containing LSVM features with different parameter combinations.

support.py: This file contains the necessary functions.

data/P1S1/: This folder, for example, contains the event streams for all channels for seizure 1 of patient 1.

Pat1_Sz_1_CH1.csv: This file contains the spikes of the AFE and the NLNG layers with the following tabular format (which can be extracted by the fig_gen.py)

Comments

SStart: 180 //Start of the Seizure in signal time# SEnd: 276.0 //Start of the Seizure in signal time# Pid: 2 // The patient ID as per the SWEC-ETHZ iEEG Database # Sid: 1 // The Seizure ID as per the SWEC-ETHZ iEEG Database # Channel_No: 1 // The channel number

SYS_time signal_time dac_value ADMspikes NLNGspikes

The time from the interface FPGA The time of the signal as per the SWEC ETHZ Database The value of the analog signals as recorded in the SWEC ETHZ Database The event-steam is the output of the AFE in boolean format. True represents a spike The spike-steam is the output of the SNN in boolean format. True represents a spike