{"success":true,"database":"eegdash","data":{"_id":"69d16e04897a7725c66f4c54","dataset_id":"ds007591","associated_paper_doi":"10.1101/2024.05.09.591996","authors":["Motoshige Sato","Yasuo Kabe","Sensho Nobe","Akito Yoshida","Masakazu Inoue","Mayumi Shimizu","Kenichi Tomeoka","Shuntaro Sasai"],"bids_version":"1.9.0","contact_info":null,"contributing_labs":null,"data_processed":false,"dataset_doi":"doi:10.18112/openneuro.ds007591.v1.0.1","datatypes":["eeg"],"demographics":{"subjects_count":3,"ages":[],"age_min":null,"age_max":null,"age_mean":null,"species":null,"sex_distribution":null,"handedness_distribution":null},"experimental_modalities":null,"external_links":{"paper_url":"https://doi.org/10.1101/2024.05.09.591996"},"funding":["JST, Moonshot R&D Grant Number JPMJMS2012"],"ingestion_fingerprint":"ca0969d2737094040a4c293089706039909ada71ef6c0156fd62bcdfb8d637c2","license":"CC0","n_contributing_labs":null,"name":"Delineating neural contributions to EEG-based speech decoding","readme":"# Delineating neural contributions to EEG-based speech decoding\n## Overview\n128-channel EEG recordings during speech production tasks.\nParticipants produced one of 5 color words (green, magenta, orange, violet, yellow)\nunder three speech conditions: overt, minimally overt, and covert.\nEach trial consists of 5 repetitions of the same word (1.25 sec per repetition).\n## Channel layout (139 channels total)\n- Channels 1-128: EEG\n- Channels 129-130: DISPLAY (bipolar pair, misc)\n- Channels 131-132: MIC (bipolar pair, misc)\n- Channels 133-134: EOG (bipolar pair)\n- Channels 135-136: EMG upper orbicularis oris (bipolar pair)\n- Channels 137-138: EMG lower orbicularis oris (bipolar pair)\n- Channel 139: TRIGGER (marks trial onsets)\n## Session types\n- calibration: Offline data collection for decoder training\n- online: Real-time decoding with trained decoder\n## Preprocessing note\nThe EEG channels were recorded with a 10x preamp gain.\nRaw values have been converted to Volts (×1e-6).\n## Code\nCode for data loading, preprocessing, and decoding models is available at:\nhttps://github.com/arayabrain/uhd-gmail-public","recording_modality":["eeg"],"senior_author":null,"sessions":["20230511","20230512","20230516","20230523","20230524","20230529"],"size_bytes":1737952647,"source":"openneuro","storage":{"backend":"s3","base":"s3://openneuro.org/ds007591","raw_key":"dataset_description.json","dep_keys":["CHANGES","README","events.json","participants.json","participants.tsv"]},"study_design":null,"study_domain":null,"tasks":["covert","minimallyovert","overt"],"timestamps":{"digested_at":"2026-05-31T16:31:32.448670+00:00","dataset_created_at":null,"dataset_modified_at":null},"total_files":21,"computed_title":"Delineating neural contributions to EEG-based speech decoding","nchans_counts":[{"val":139,"count":21}],"sfreq_counts":[{"val":256.0,"count":21}],"stats_computed_at":"2026-05-31T19:34:32.603722+00:00","total_duration_s":24393.0,"tagger_meta":{"model":"openai/gpt-4o","tagged_at":"2026-06-10T08:19:41Z","source":"eegdash-llm-tagger"},"tags":{"pathology":["Other"],"modality":["Auditory"],"type":["Other"],"confidence":{"pathology":0.7,"modality":0.9,"type":0.8},"reasoning":{"few_shot_analysis":"There are no directly matching few-shot examples in terms of task paradigm or population. However, the 'EEG: Reinforcement Learning in Parkinson's' and 'EEG: Three-Stim Auditory Oddball and Rest in Acute and Chronic TBI' examples provide insight into how datasets with specific tasks like decision-making or involving clinical conditions like TBI or Parkinson's are labeled. These examples indicate that when a dataset involves a clear cognitive task or medical condition, specific labels such as 'Decision-making' and a condition-specific pathology are used, respectively.","metadata_analysis":"The metadata reveals that the study involves EEG recordings during speech production tasks ('128-channel EEG recordings during speech production tasks'). The tasks involve producing color words under different speech conditions, with an emphasis on the overt, minimally overt, and covert speech ('Participants produced one of 5 color words under three speech conditions: overt, minimally overt, and covert'). It also mentions that the study employs Explainable AI techniques to a convolutional neural network predicting spoken words based on EEG signals. This indicates a focus on the neural processing of speech rather than purely on perception or motor aspects.","paper_abstract_analysis":"The abstract supports the focus on EEG-based speech decoding. It discusses the use of ultra-high-density EEG for facilitating communication in individuals with speech impairments through BCIs ('Speech Brain-computer interfaces (BCIs) have emerged as a pivotal technology in facilitating communication for individuals with speech impairments'). It highlights the role of neural activities and distinguishes them from muscular artifacts during speech production.","evidence_alignment_check":"1. Pathology: Metadata suggests the use of speech decoding to aid individuals with speech impairments, but there's no explicit mention of a medical condition being studied, which aligns with the few-shot examples under 'Healthy'.\n2. Modality: Metadata explicitly states the task involves speech, suggesting 'Auditory', aligning with the focus on speech-decoding.\n3. Type: The use of AI techniques to decode speech and differentiate neural contributions suggests 'Other', as there's no clear focus on known cognitive types like decision-making or perception.","decision_summary":"For 'Pathology', the most accurate label is 'Other' due to the focus on speech decoding for aiding communication and an indirect reference to speech impairment. 'Healthy' was considered but was discarded as the main focus isn't a healthy cohort study. For 'Modality', 'Auditory' is the most supported label since the task involves speech production, aligning with auditory processing. For 'Type', 'Other' is more fitting due to the specific aim of exploring EEG-based speech recognition, which doesn't align neatly with known categories like perception or motor. The evidence from the metadata aligns strongly with these selections, providing high confidence in the chosen labels."}},"canonical_name":null,"name_confidence":0.84,"name_meta":{"suggested_at":"2026-04-14T10:18:35.343Z","model":"openai/gpt-5.2 + openai/gpt-5.4-mini + deterministic_fallback"},"name_source":"author_year","author_year":"Sato2026_Delineating","bad_channels_info":null,"acknowledgements":"We thank Ryota Kanai, for helpful discussions, and Rousslan Fernand Julien Dossa for data collection. Special thanks to Anna Maria Hadjiev for her meticulous proofreading, significantly enhancing our manuscript's quality.","ethics_approvals":["Shiba Palace Clinic Ethics Review Committee","Declaration of Helsinki"],"how_to_acknowledge":"Please cite the following paper (doi: https://doi.org/10.1101/2024.05.09.591996) and this dataset using its OpenNeuro DOI.","references_and_links":["https://doi.org/10.1101/2024.05.09.591996"],"associated_paper_meta":{"channel":"text/how_to_acknowledge","confidence":"high","author_overlap":8,"is_oa":true,"oa_status":"preprint","source":"paper_resolver"}}}