# === YAML Fixing Applied ===
id: "doi:10.57764/qb6h-em84"
name: Bridge2AI-Voice v1.0
title: "Bridge2AI-Voice: An ethically-sourced, diverse voice dataset linked to health information (version 1.0)"
description: >
  Bridge2AI-Voice is a comprehensive, ethically sourced collection of data
  derived from voice recordings linked to clinical and demographic information,
  intended to enable research in artificial intelligence for using voice as a
  biomarker of health. Version 1.0 contains 12,523 recordings from 306 adult
  participants collected across five North American sites. The initial release
  includes low-risk, de-identified derived data (e.g., spectrograms and acoustic
  features) and detailed phenotype/questionnaire data; original audio waveforms
  are not included in this release.
page: "https://docs.b2ai-voice.org"
doi: "doi:10.57764/qb6h-em84"
issued: 2024-11-27
version: "1.0"
keywords:
  - voice
  - bridge2ai
  - audio
created_by:
  - Alistair Johnson
  - Jean-Christophe Bélisle-Pipon
  - David Dorr
  - Satrajit Ghosh
  - Philip Payne
  - Maria Powell
  - Anaïs Rameau
  - Vardit Ravitsky
  - Alexandros Sigaras
  - Olivier Elemento
  - Yael Bensoussan
purposes:
  - name: Purpose
    response: >
      Create an ethically sourced flagship dataset to enable AI research and
      support insights into the use of voice as a biomarker of health by
      linking derived voice data with clinical and demographic information.
tasks:
  - name: Intended Tasks
    response: >
      AI/ML research and development using derived voice representations
      (e.g., spectrograms and acoustic/phonetic/prosodic features) for health-related
      applications across multiple disorder cohorts.
addressing_gaps:
  - name: Addressing Gap
    response: >
      Addresses the lack of large, high-quality, multi-institutional, demographically
      diverse voice datasets with standardized collection protocols and linked health
      information; overcomes prior limitations of small, heterogeneous datasets
      and inconsistent methods that impede external validity and clinical utility.
funders:
  - name: NIH Bridge2AI Voice Project
    grantor:
      name: National Institutes of Health
    grant:
      name: "Bridge2AI: Voice as a Biomarker of Health - Building an ethically sourced, bioacoustic database to understand disease like never before"
      grant_number: 3OT2OD032720-01S1
instances:
  - name: Instance description
    representation: Voice recordings linked with clinical/demographic and questionnaire data
    instance_type: Participants (n=306), sessions (one or more per participant), and recordings (12,523 total)
    data_type: >
      Derived data only in this release: spectrograms (513 × N), acoustic features (OpenSMILE),
      phonetic/prosodic measures (Parselmouth/Praat), and transcriptions; plus phenotype
      tabular data and feature tables.
    counts: 12523
    label: >
      Cohort categories and task_name per recording; detailed phenotype variables per participant.
    sampling_strategies:
      - name: Sampling and cohort selection
        is_sample:
          - yes
        is_random:
          - no
        source_data:
          - Patients presenting to specialty clinics and institutions across five North American sites
        is_representative:
          - not assessed; cohort-based clinical recruitment
        why_not_representative:
          - >
            Participants were recruited into predefined clinical cohorts (respiratory, voice,
            neurological, mood, pediatric) rather than via probabilistic sampling of a general population.
        strategies:
          - Clinical cohort-based recruitment at participating sites
    missing_information:
      - name: Missing information notes
        missing:
          - Raw audio waveforms not available in v1.0
          - Free speech transcripts removed
        why_missing:
          - Risk reduction and de-identification measures for this initial, low-risk release
relationships:
  - name: Instance relationships
    description:
      - Participant-to-session and session-to-recording relationships are explicit via participant_id and session_id
splits:
  - name: Recommended data splits
    description:
      - No recommended train/validation/test splits specified in v1.0
anomalies:
  - name: Data anomalies
    description:
      - None reported in the source; standardized preprocessing was applied
external_resources:
  - name: External resources
    external_resources:
      - Documentation website: "https://docs.b2ai-voice.org"
      - REDCap instrument release (Zenodo): "https://doi.org/10.5281/zenodo.14148755"
      - Preprocessing library (b2aiprep) source code: "https://github.com/sensein/b2aiprep"
    archival:
      - Versioned DOIs provided for dataset discovery
confidential_elements:
  - name: Confidentiality considerations
    description:
      - Clinical and questionnaire information linked to participants (de-identified); released data are low risk and derived
content_warnings:
  - name: Content warnings
    warnings:
      - None noted
subpopulations:
  - name: Identified subpopulations
    identification:
      - Adult cohort (v1.0 release includes adult participants only)
      - Disease cohorts: respiratory disorders, voice disorders, neurological/neurodegenerative disorders, mood/psychiatric disorders, pediatric (planned)
    distribution:
      - Distribution across cohorts not detailed in this source
sensitive_elements:
  - name: Sensitive elements
    description:
      - Health-related clinical and demographic variables; questionnaire responses
acquisition_methods:
  - name: Data acquisition
    description:
      - >
        Standardized clinical data collection protocol including demographics, health questionnaires,
        targeted confounder questionnaires, disease-specific information, and voice tasks (e.g., sustained vowels).
      was_directly_observed: "yes (audio tasks and recordings)"
      was_reported_by_subjects: "yes (questionnaires)"
      was_inferred_derived: "yes (derived features, spectrograms, and transcriptions from raw audio)"
      was_validated_verified: >
        Standardized protocol; REDCap-based capture with export/transform via open-source library; institutional ethics approvals
collection_mechanisms:
  - name: Collection mechanisms
    description:
      - >
        Custom tablet application; headset used when possible; REDCap used for data capture;
        export and conversion performed using an open-source library (b2aiprep).
data_collectors:
  - name: Data collectors
    description:
      - Project investigators at participating specialty clinics and institutions
collection_timeframes:
  - name: Collection timeframe
    description:
      - Not specified; multiple sites in North America; single or multiple sessions per participant as needed
ethical_reviews:
  - name: Ethical review and approvals
    description:
      - University of South Florida Institutional Review Board (approved)
      - University of Toronto Research Ethics Board (submitted)
data_protection_impacts:
  - name: Data protection impact
    description:
      - >
        Initial release limited to low-risk, de-identified derived data; raw audio withheld to reduce re-identification risk.
preprocessing_strategies:
  - name: Audio preprocessing and derivations
    description:
      - Convert raw audio to mono; resample to 16 kHz with Butterworth anti-aliasing filter
      - Compute spectrograms via STFT (25 ms window, 10 ms hop, 512-point FFT)
      - Extract acoustic features (OpenSMILE)
      - Compute phonetic/prosodic features (Parselmouth and Praat)
      - Generate transcriptions using OpenAI Whisper Large
    used_software:
      - name: openSMILE
      - name: Parselmouth
      - name: Praat
      - name: Torchaudio
      - name: OpenAI Whisper Large
      - name: b2aiprep
cleaning_strategies:
  - name: De-identification and cleaning
    description:
      - >
        HIPAA Safe Harbor identifiers removed; state/province removed (country retained); free speech transcripts removed;
        in v1.0, audio waveforms omitted and only derived features released.
labeling_strategies:
  - name: Labeling and annotations
    description:
      - Automated transcriptions generated with OpenAI Whisper Large; free speech transcripts removed in v1.0
raw_sources:
  - name: Raw data availability
    description:
      - Raw audio waveforms were collected but are not included in v1.0; future releases may include audio with additional safeguards
existing_uses:
  - name: Existing uses
    description:
      - Not specified
other_tasks:
  - name: Potential additional tasks
    description:
      - Not specified
future_use_impacts:
  - name: Considerations for future use
    description:
      - >
        Only low-risk derived data are included; omission of raw audio reduces privacy risks but may limit certain modeling approaches.
discouraged_uses:
  - name: Discouraged uses
    description:
      - Not specified
distribution_formats:
  - name: Distribution formats
    description:
      - spectrograms.parquet (Parquet; derived spectrograms with participant_id, session_id, task_name, 513×N arrays)
      - phenotype.tsv (tab-delimited; one row per participant with demographics/questionnaires/confounders)
      - phenotype.json (data dictionary for phenotype.tsv)
      - static_features.tsv (tab-delimited; one row per recording with derived features)
      - static_features.json (data dictionary for static_features.tsv)
distribution_dates:
  - name: Distribution dates
    description:
      - Initial public release: 2024-11-27 (v1.0)
license: Bridge2AI Voice Registered Access License
license_and_use_terms:
  name: License and access terms
  description:
    - Only credentialed users who sign the Data Use Agreement (DUA) can access files
    - Bridge2AI Voice Registered Access License applies
    - Data Use Agreement: Bridge2AI Voice Registered Access Agreement
    - Required training: "TCPS 2: CORE 2022"
    - Access via Health Data Nexus (credentialed access)
ip_restrictions:
  name: IP or third-party restrictions
  description:
    - None specified beyond the Registered Access License and DUA
regulatory_restrictions:
  name: Export control and regulatory restrictions
  description:
    - None specified
updates:
  name: Update plan
  description:
    - Future releases are planned to include voice audio with additional security precautions
version_access:
  name: Versioning and access
  description:
    - Version-specific DOI (v1.0): "https://doi.org/10.57764/qb6h-em84"
    - Latest-version DOI: "https://doi.org/10.57764/3sg0-7440"
is_deidentified:
  name: De-identification status
  description:
    - HIPAA Safe Harbor identifiers removed; state/province removed; free speech transcripts removed; audio waveforms omitted in v1.0
is_tabular: Partial; phenotype and feature data are tabular (TSV/JSON), spectrograms stored in Parquet
subsets:
  - id: bridge2ai-voice-v1.0-adult
    name: Adult cohort subset (v1.0)
    title: Adult cohort included in v1.0
    description: v1.0 release includes adult participants only; pediatric cohort planned for future releases
    is_subpopulation: adult participants only in this release
    is_data_split: no recommended ML splits provided
external_resources:
  - name: Documentation site
    external_resources:
      - https://docs.b2ai-voice.org
    restrictions:
      - None (documentation)
  - name: REDCap artifacts release
    external_resources:
      - https://doi.org/10.5281/zenodo.14148755
    restrictions:
      - None indicated for viewing the record
  - name: Preprocessing code (b2aiprep)
    external_resources:
      - https://github.com/sensein/b2aiprep
    restrictions:
      - Open-source code repository