Dataset Metadata

Dataset metadata is divided into six tabs under the Metadata section of a dataset. Filling in thorough metadata improves discoverability and ensures your dataset meets FAIR principles.

For eye imaging datasets, metadata should be completed with AI-readiness in mind. At minimum, document modality, acquisition context, device information, access conditions, and provenance.

General Information

Found at /metadata/general-information.

Core descriptive fields:

• Titles - One or more titles for the dataset. You can add alternate titles with a type (e.g., subtitle, translated title).
• Descriptions - One or more descriptions. Types include abstract, methods, technical info, and others.
• Resource type - The type of resource (e.g., dataset, collection).
• Language - Primary language of the dataset content (default: English).
• Format - File format(s) of the dataset.
• Size - Approximate size of the dataset.
• Standards followed - Any data or metadata standards the dataset conforms to.

Recommended entries for Standards followed include CDS (dataset structure), DICOM (imaging format), and OMOP CDM (for associated clinical tables).

About

Found at /metadata/about.

Free-text fields for extended documentation of the dataset content, context, and purpose.

Identifiers

Found at /metadata/identifiers.

• DOI - Assigned automatically on publication.
• Alternate identifiers - Other identifiers for the dataset, such as accession numbers or internal IDs. Each alternate identifier has a type (e.g., accession-number, url).

Related Identifiers

Found at /metadata/related-identifiers.

Links to related resources such as publications, source datasets, or derived datasets. Each entry includes:

• Identifier - The related identifier (e.g., a DOI or URL)
• Type - The identifier type (e.g., doi, url, arxiv)
• Relation - The relationship (e.g., is-cited-by, is-derived-from, is-supplement-to)

Access Rights

Found at /metadata/access-rights.

Controls how the dataset can be accessed after publication:

• Access type - Public or controlled
• Access URL - Where users can request or download the data
• Consent - Data use restrictions and consent terms
• De-identification level - Method used (e.g., HIPAA Safe Harbor, k-anonymity, generalization)
• Rights - License and rights statement (e.g., CC BY 4.0)

Use this section to explicitly document participant consent constraints and any reuse restrictions that apply to downstream AI/ML usage.

Data Management

Found at /metadata/data-management.

• Managing organization - The organization responsible for the dataset
• Funders - Funding sources with award numbers and grant identifiers
• Dates - Key dates such as collection start/end, last modified, and publication date
• Subjects - Keywords and subject classifications

For ophthalmic data, include terms for imaging modality and disease focus (for example: OCT, OCTA, fundus, FLIO, AMD, glaucoma, diabetic retinopathy) to improve discoverability.

Team

Found at /metadata/team.

Lists all contributors to the dataset. Each contributor has:

• Name - Full name
• Affiliation - Institutional affiliation
• Role - Contribution type (e.g., creator, data collector, principal investigator)
• Identifier - Optional ORCID or other identifier

AI-Readiness Checklist

Before publishing, verify the metadata captures:

• Imaging modality and file format for each data type
• Device or acquisition context where relevant
• Participant-level de-identification approach
• Clear access model and licensing terms
• Sufficient provenance and contributor attribution
• Key related identifiers (publications, source datasets, derived resources)