=== YAML Fixing Applied ===
id: "doi:10.57764/qb6h-em84"
name: Bridge2AI-Voice
title: "Bridge2AI-Voice: An ethically-sourced, diverse voice dataset linked to health information"
description: >-
  Bridge2AI-Voice is a comprehensive, ethically sourced dataset derived from voice
  recordings linked to corresponding clinical information. v1.0 provides 12,523
  recordings for 306 participants collected across five North American sites.
  Participants were selected based on conditions known to manifest within the voice
  waveform, including voice disorders, neurological and neurodegenerative disorders,
  mood and psychiatric disorders, and respiratory disorders. The initial release
  contains low-risk, derived data (e.g., spectrograms and acoustic features) and
  clinical/demographic questionnaire data; original audio waveforms and free speech
  transcripts are not included in v1.0.
language: English
issued: 2024-11-27
page: "https://docs.b2ai-voice.org"
doi: "doi:10.57764/qb6h-em84"
version: "1.0"
keywords:
  - voice
  - bridge2ai
  - audio
  - spectrograms
  - clinical data
  - de-identified
  - biomarker
  - health
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
license: Bridge2AI Voice Registered Access License
status: "bibo:draft"
purposes:
  - id: purpose-1
    name: Dataset Purpose
    response: >-
      Enable research into the use of voice as a biomarker of health by providing
      ethically sourced, diverse, multi-site voice-derived data linked to clinical
      and demographic information.
tasks:
  - id: task-1
    name: Intended Tasks
    response: >-
      Artificial intelligence and machine learning research on health-related voice
      biomarkers, including exploratory analysis, feature development, and model
      building for conditions with known voice manifestations (voice, neurological,
      mood/psychiatric, and respiratory disorders).
addressing_gaps:
  - id: gap-1
    name: Addressed Gap
    response: >-
      Addresses the lack of large, high-quality, diverse, multi-institutional
      voice datasets with linked health information collected under standardized,
      ethically grounded protocols.
creators:
  - id: creator-1
    name: Bridge2AI-Voice Author
    principal_investigator:
      id: person-alistair-johnson
      name: Alistair Johnson
  - id: creator-2
    name: Bridge2AI-Voice Author
    principal_investigator:
      id: person-jean-christophe-belisle-pipon
      name: Jean-Christophe Bélisle-Pipon
  - id: creator-3
    name: Bridge2AI-Voice Author
    principal_investigator:
      id: person-david-dorr
      name: David Dorr
  - id: creator-4
    name: Bridge2AI-Voice Author
    principal_investigator:
      id: person-satrajit-ghosh
      name: Satrajit Ghosh
  - id: creator-5
    name: Bridge2AI-Voice Author
    principal_investigator:
      id: person-philip-payne
      name: Philip Payne
  - id: creator-6
    name: Bridge2AI-Voice Author
    principal_investigator:
      id: person-maria-powell
      name: Maria Powell
  - id: creator-7
    name: Bridge2AI-Voice Author
    principal_investigator:
      id: person-anais-rameau
      name: Anaïs Rameau
  - id: creator-8
    name: Bridge2AI-Voice Author
    principal_investigator:
      id: person-vardit-ravitsky
      name: Vardit Ravitsky
  - id: creator-9
    name: Bridge2AI-Voice Author
    principal_investigator:
      id: person-alexandros-sigaras
      name: Alexandros Sigaras
  - id: creator-10
    name: Bridge2AI-Voice Author
    principal_investigator:
      id: person-olivier-elemento
      name: Olivier Elemento
  - id: creator-11
    name: Bridge2AI-Voice Author
    principal_investigator:
      id: person-yael-bensoussan
      name: Yael Bensoussan
funders:
  - id: funding-nih-b2ai-voice
    name: NIH Bridge2AI Voice Grant
    grantor:
      id: org-nih
      name: National Institutes of Health (NIH)
    grant:
      id: grant-3OT2OD032720-01S1
      name: "Bridge2AI: Voice as a Biomarker of Health - Building an ethically sourced, bioaccoustic database to understand disease like never before"
      grant_number: 3OT2OD032720-01S1
instances:
  - id: instance-voice-derived
    name: Voice-derived data and linked clinical data
    representation: >-
      Voice recordings (sessions/tasks) linked to participant-level clinical and
      demographic data; v1.0 provides derived spectrograms and acoustic features
      with identifiers linking participants, sessions, and recordings.
    instance_type: Participants, sessions, and recordings (derived data per recording; clinical data per participant)
    data_type: >-
      Derived features from raw audio (spectrograms, acoustic/acoustic-phonetic
      features) and linked tabular clinical/demographic questionnaire data.
    counts: 12523
    label: >-
      No supervised task labels provided; includes task_name metadata for recording
      tasks and clinical questionnaire responses.
    sampling_strategies:
      - id: sampling-1
        name: Cohort design
        is_sample:
          - "yes"
        is_random:
          - "no"
        source_data:
          - Patients presenting at specialty clinics across five North American sites
        is_representative:
          - "no"
        why_not_representative:
          - Cohort selected for specific disease categories with known voice manifestations
        strategies:
          - Targeted recruitment into five disease cohort categories
    missing_information:
      - id: missing-1
        name: Free speech transcript removal
        missing:
          - Free speech transcripts
        why_missing:
          - Removed for de-identification and privacy protection
      - id: missing-2
        name: Raw audio omission
        missing:
          - Original audio waveforms
        why_missing:
          - Omitted in v1.0 to reduce re-identification risk; only derived data released
relationships:
  - id: relationships-1
    name: Participant-session-recording linkage
    description:
      - Spectrograms and features link participant_id, session_id, and task_name to derived data elements; phenotype links participant-level data
subpopulations:
  - id: subpop-1
    name: Adult cohort and disease-based subgroups
    identification:
      - Adult cohort (v1.0 only); disease categories: voice disorders, neurological/neurodegenerative disorders, mood/psychiatric disorders, respiratory disorders
    distribution:
      - Distribution across subgroups not provided in v1.0 overview
is_deidentified:
  id: deid-1
  name: De-identification
  description:
    - HIPAA Safe Harbor identifiers removed
    - State and province removed; country of data collection retained
    - Free speech transcripts removed
    - Original audio waveforms omitted in v1.0; only spectrograms and derived features released
sensitive_elements:
  - id: sensitive-1
    name: Health and biometric-related data
    description:
      - Contains health-related clinical and demographic questionnaire data
      - Voice-derived features and spectrograms (biometric characteristics), with de-identification applied
acquisition_methods:
  - id: acquisition-1
    name: Data acquisition
    description:
      - Standardized protocol across sites; demographic and health questionnaires; targeted questionnaires on acoustic confounders; disease-specific information; voice recording tasks (e.g., sustained vowel)
      - Data exported and converted from REDCap using an open-source library
    was_directly_observed: "yes (voice tasks, recordings)"
    was_reported_by_subjects: "yes (questionnaires)"
    was_inferred_derived: "yes (acoustic features, spectrograms, phonetic/prosodic metrics, ASR transcripts pre-removal)"
    was_validated_verified: "Standardized multi-site protocol; ethics approvals as noted"
collection_mechanisms:
  - id: collmech-1
    name: Collection mechanisms
    description:
      - Custom data collection application on a tablet; headset used when possible
      - Standardized data collection protocol across five North American sites
data_collectors:
  - id: collectors-1
    name: Site investigators
    description:
      - Participants screened and consented by project investigators at specialty clinics; standardized in-clinic data collection
ethical_reviews:
  - id: irb-1
    name: University of South Florida IRB
    description:
      - Data collection and sharing approved by the University of South Florida Institutional Review Board
  - id: irb-2
    name: University of Toronto Research Ethics Board
    description:
      - Submission for review to the University of Toronto Research Ethics Board
preprocessing_strategies:
  - id: prep-1
    name: Audio preprocessing and feature extraction
    description:
      - Raw audio converted to mono and resampled to 16 kHz with a Butterworth anti-aliasing filter
      - Spectrograms via STFT (25 ms window, 10 ms hop, 512-point FFT), yielding 513 x N matrices
      - Acoustic features with OpenSMILE
      - Phonetic and prosodic features via Parselmouth and Praat (F0, formants, voice quality)
      - Transcriptions generated using OpenAI Whisper Large (free speech transcripts removed prior to release)
    used_software:
      - id: sw-opensmile
        name: openSMILE
      - id: sw-praat
        name: Praat
      - id: sw-parselmouth
        name: Parselmouth
      - id: sw-torchaudio
        name: torchaudio
      - id: sw-whisper
        name: OpenAI Whisper (Large)
cleaning_strategies:
  - id: clean-1
    name: De-identification and redactions
    description:
      - Removal of HIPAA Safe Harbor identifiers (e.g., names, fine-grained dates, contact numbers, IDs, URLs, biometric identifiers)
      - Removal of state/province; retention of country
      - Removal of free speech transcripts
      - Omission of original audio waveforms in v1.0
labeling_strategies:
  - id: label-1
    name: Transcription and metadata labeling
    description:
      - ASR transcriptions generated via Whisper Large (free speech transcripts removed from release)
      - Task-level metadata (task_name) per recording; phenotype data dictionary with column descriptions
    used_software:
      - id: sw-whisper-2
        name: OpenAI Whisper (Large)
raw_sources:
  - id: raw-1
    name: Raw audio recordings
    description:
      - Original audio recordings were collected but are not included in v1.0; only derived spectrograms and features are released. Future releases aim to include audio with additional safeguards.
distribution_formats:
  - id: distfmt-1
    name: Released files and formats
    description:
      - spectrograms.parquet (derived spectrogram matrices; columns include participant_id, session_id, task_name, spectrogram)
      - static_features.tsv (one row per recording; acoustic feature set)
      - static_features.json (data dictionary for features)
      - phenotype.tsv (one row per participant; demographics, confounders, validated questionnaires)
      - phenotype.json (data dictionary for phenotype)
distribution_dates:
  - id: distdate-1
    name: Initial release date
    description:
      - 2024-11-27
license_and_use_terms:
  id: license-1
  name: Access, license, and use terms
  description:
    - Access is restricted to credentialed users who sign the Bridge2AI Voice Registered Access Agreement (DUA)
    - License: Bridge2AI Voice Registered Access License
    - Required training: TCPS 2: CORE 2022
    - Files are available under credentialed access via Health Data Nexus
updates:
  id: updates-1
  name: Update plan
  description:
    - v1.0 is the initial release; future releases aim to include voice waveforms with additional data security precautions
version_access:
  id: versioning-1
  name: Versioning and DOIs
  description:
    - Versioned DOI for v1.0: https://doi.org/10.57764/qb6h-em84
    - Latest-version DOI resolver: https://doi.org/10.57764/3sg0-7440
other_tasks:
  - id: othertask-1
    name: Additional potential uses
    description:
      - Research on disease screening and monitoring using voice-derived biomarkers across specified disease cohorts, subject to license and access restrictions
is_tabular: "Mixed: tabular phenotype/features (TSV/JSON) and matrix-form spectrograms (Parquet)"