Projects & Metadata Governance

Projects: Structured Collaboration Without Data Transfer

In CENTAURON, a Project defines the clinical question, cohort structure and data model for a collaborative AI initiative. Projects do not store medical data. Instead, they function as decentralized metadata registries that allow institutions to describe and reference their datasets while keeping all slides and annotations local.

Each Project includes:

• Clinical objective
  e.g., response prediction in colorectal cancer, biomarker surrogate tasks, rare tumor benchmarking
• Metadata schema & terminology
  Standardized taxonomies (e.g., ICD-O, UniProt, staining ontologies) ensure semantic consistency across institutions
• Participation and quality criteria
  Institutions contribute data that meets project-specific clinical and technical requirements
• Governance rules & cryptographic access controls
  Access and evaluation permissions are authenticated and recorded on the permissioned blockchain

By describing datasets rather than transmitting them, Projects allow cohorts to scale across institutions and jurisdictions without data leaving clinical environments.

Decentralized Metadata Layer

Project metadata forms a federated registry distributed across the network. Each institution publishes only the metadata it chooses to share, enabling: discovery of compatible datasets

• structured cohort assembly
• transparent contribution tracking
• auditability of data provenance

All contributions and access authorizations are immutably logged, allowing verifiable attribution and scientific transparency.

Outcome

Projects enable collaborative cohort building and discovery without centralized data aggregation. Pathology departments retain data sovereignty while benefiting from network-wide visibility, shared scientific standards and the ability to contribute securely to large-scale clinical AI initiatives.