Neural Noise

Human iEEG/LFP dataset with multilingual cognitive and eye-tracking tasks

3 min read

  • Human brain local field potential (LFP) recordings were collected during a battery of multilingual cognitive and eye-tracking tasks.1
  • The dataset directly supports decoding and BCI research; LFP is core electrophysiology for invasive interfaces.1
  • Nature-published; strong method and data utility for human neuroscience.1 1

Weekly enrichment (2026-07-20)

  • The dataset was published as a Data Descriptor in Nature’s Scientific Data (2025, DOI 10.1038/s41597-025-05222-2); recordings come from stereo-EEG (sEEG) depth electrodes implanted for seizure localization in drug-resistant epilepsy patients.2 3
  • It currently comprises recordings from 23 patients with 143 tasks performed; 15 patients completed all tasks and 11 repeated at least one task across two runs (separated by roughly 24–72 hours), enabling within-subject reproducibility analyses.2 3
  • Data were collected at two clinical sites — St. Anne’s University Hospital, Masaryk University (Brno, Czech Republic) and Uniwersytecki Szpital Kliniczny, Medical University of Wroclaw (Poland) — with patients tested in their native Polish, Czech, or Slovak (site language counts: Czech 11, Slovak 7, Polish 5).2 3
  • Each case yielded over 100 macro- and micro-contact LFP channels; electrophysiology was sampled at 5 kHz (Brno, Easys2/M&I) and 4 kHz for macro plus 32 kHz for micro contacts (Wroclaw, Neuralynx Digital Lynx SX), using AdTech and DIXI hybrid depth electrodes with 5–10 mm contact spacing.2
  • The battery contained five tasks in the order Smooth Pursuit (SP), Free Recall (FR, 12 words), Anti-saccade/pro-saccade (AP), Paired-Associate Learning (PAL, 6 word pairs with cued recall), and Word Screening (WS); WS presented 180 words over 5 trials so each noun appeared 5 times.2
  • A shared pool of 180 common nouns translated from an English source list into Polish, Czech, and Slovak was reused across paradigms and days, allowing analysis of selective neural responses to specific word stimuli across languages.2
  • Gaze and pupil data were captured with the i4tracking system (Medicton Group) via a camera sampling up to 150 Hz at ~0.1 mm/pixel resolution, vocal responses recorded at 44,100 Hz, all synchronized via TTL pulses; raw signals are stored in MEF3 format and organized in BIDS.2
  • The de-identified dataset is released under a Data Use Agreement through the EBRAINS Knowledge Graph (DOI 10.25493/4fzh-zcg), with open-source MEF readers for Python (pymef) and MATLAB (matmef), directly supporting invasive decoding and BCI research.2 4
  • Context: it complements other 2025 open iEEG releases such as the Cogitate Consortium’s multi-center dataset from 38 pharmaco-resistant epilepsy patients (4,771 electrodes; 1,238 surface and 3,533 depth) probing conscious visual perception with synchronized eye-tracking and a reproducible Jupyter analysis pipeline.5

Footnotes

  1. https://news.google.com/rss/articles/CBMiX0FVX3lxTE16Q3lDeEstZ0FjS3FCem1iYmhaQk5JR0d5RDUtWFRtbndLZDBiYzFxWXdqeWxiWFNoSUdNblZwLXM4M1RaSVdPLU5Td0ZhRVN4SnpEQjJSd0dUZ0JuUlQw?oc=5 ↩ ↩2 ↩3 ↩4

  2. https://www.nature.com/articles/s41597-025-05222-2 ↩ ↩2 ↩3 ↩4 ↩5 ↩6 ↩7 ↩8

  3. https://doi.org/10.1038/s41597-025-05222-2 ↩ ↩2 ↩3

  4. https://doi.org/10.25493/4fzh-zcg ↩

  5. https://www.nature.com/articles/s41597-025-04833-z ↩

Graph View