The BETTER platform combines the best-in-class open-source Personal Health Train infrastructure with modern federated analytics capabilities and GDPR-aligned governance.
BETTER is designed as a true federated infrastructure, not a data warehouse. The platform never centralizes patient data — instead, it orchestrates privacy-preserving computations across distributed sites.
Stations
Each clinical partner operates a "Station" — an on-premise analytical node that hosts local FAIR-prepared patient data and executes federated queries in a secure environment.
Trains
A "Train" is a containerized analytical task — a statistical query, machine learning algorithm, or data harmonization script — that travels to each Station to execute locally.
Central Aggregator
Orchestrates Train dispatching, collects privacy-preserving results from Stations and aggregates insights — while never observing raw clinical records.
Step 01
A federated query is created on researcher workstations.
Step 02
The analytical task travels to each participating institution.
Step 03
The query processes data locally; raw data never leaves the site.
Step 04
Only privacy-preserving aggregated results are shared centrally.
Centralized Approach
Data location
Centralised in a single repository
Privacy risk
High patient re-identification risk
GDPR compliance
Difficult across borders
Cross-border transfer
Requires extensive legal agreements
Research scalability
Limited by transfer agreements
BETTER Federated Approach
Data location
Remains at local medical centres
Privacy risk
Low; only aggregated results leave sites
GDPR compliance
Compliant by design
Cross-border transfer
No raw data ever leaves local site
Research scalability
Enables pan-European collaboration
PADME data principles (Findable, Accessible, Interoperable, Reusable) are applied at every BETTER Station through semantic standardization and ontology mapping.
Each data source is harmonized using HL7 FHIR resources, SNOMED CT terminologies, and domain-specific ontologies (e.g., HPO for phenotypes). This enables cross-institutional querying without prior manual curation — the federated network only works when local data speaks the same semantic language.
Vantage6 is an open-source federated learning platform developed for the Personal Health Train paradigm. It orchestrates the secure execution of containerized algorithms across distributed healthcare data sources.
The BETTER platform extends Vantage6 with additional features: multi-modal data integration, FAIR metadata services, advanced cryptographic protocols and user interfaces tailored for clinical research workflows.
Patient data never leaves the Station. All data operations are executed locally behind institutional firewalls. No patient record ever transmits to the Aggregator — only statistical summaries and model parameters pass through the network.
Each Station exposes metadata through standardized FAIR catalogues (DCAT, schema.org) that describe available datasets without exposing content. Researchers can discover what exists across Europe before requesting federated access.
Every Train execution is logged, auditable, and reversible. Stations retain full sovereignty: they can inspect queries before execution, reject non-compliant analyses, and withdraw participation at any time without affecting other institutions.
Stations maintain complete governance authority over their data. No central authority can force data access. Patient consent is validated locally. All operations comply with national data protection laws specific to each participating country.
