Storyline: Federated Analysis

Owned by Health-RI
Aug 23, 2024
3 min read

DATE: 23-08-2024 STATUS: ADOPTED

This article describes the storyline federated analysis.

This storyline describes how a federated analysis prepared by the data user in Storyline: Request data is performed by a Federated Analysis Coordinator and participating data providers and data processors. The requested data has been made available by the data providers for federated analysis in Storyline: Data delivery

Comments

  • This storyline describes a generic process for federated analysis. This also includes previously described concepts for federated analysis, such as the Personal Health Train (PHT).

  • This storyline describes federated analysis with a central aggregator on horizontally partitioned data.

  • In the case of vertically partitioned data, broader agreements are needed. These need to be further developed and the corresponding storyline (or adaptation/extension to this) will be described in a subsequent version of the architectural design.

  • This storyline shows a federated analysis with a single processor; in practice, N stations can be visited. The description shall indicate which actions will then be carried out by all participating stations. In the diagram, the bottom row is output per participating party to the processing.

  • In this storyline, the "Initiate Analysis" and "Process Results" functions are performed by the same Federated Analysis Coordinator. There are scenarios in which these functions are divided between two Federated Analysis Coordinators

Pre-condition

  • The data user has submitted a data request (including the federated analysis to be carried out), and this has been approved by the respective review committees.

  • The datasources are made FAIR based on a shared ontology that is know to the participating stations

  • The contracts have been signed

  • The data provider is connected to a network that supports federated analysis.

  • The requested data is made available on the federated data stations using Storyline: Data delivery

Process model

  1. The data user gives the Federated Analysis Coordinator the signal to start the federated analysis and indicates where the Federated Analysis Coordinator can retrieve or send the analysis approved in the application process.

  2. The Federated Analysis Coordinator retrieves the analysis approved in the request and sends it to all affected data processors.

  3. The data processor concerned performs the federated analysis determined by the data user on the data made available and sends the result to the Federated Analysis Coordinator in accordance with the agreed format and content.

  4. The Federated Analysis Coordinator will inquire with the data processor concerned whether the analysis has been completed. If this is the case, the Federated Analysis Coordinator retrieves the federated analysis result.

  5. The Federated Analysis Coordinator processes the results sent in accordance with the agreed format and content.

  6. Depending on the agreements made, the Federated Analysis Coordinator sends the final (partial) result of the federated analysis to the data user and indicates whether the requested federated analysis has been completed in full or in part.

  7. The data user receives the (partial) results of the federated analysis and processes them further. If a partial result has been sent by the Federated Analysis Coordinator, the data user determines whether the federated analysis can be stopped or whether federated analysis only stops after receiving a complete result.

  8. The data user generates research data by means of the process described in Storyline: Generating research data .

  9. The data user makes the newly generated research results available within the Health-RI ecosystem by means of the process described in Storyline: Making data available at the desired data holder.

  10. The data user has completed research and cleans up the used envrionment in accordance with established or applicable agreements, including

    • Datasets and algorithm/pipeline in data processors

    • Approved request and algorithm/pipeline at Federated Analysis Coordinator

Postcondition

  • The investigation has been carried out and the results have been secured.

  • The data management plan has been updated

  • The research project has ended

  • The research results are ready to be made available.

  • The (partial) results at the coordinator of federated analysis and the partial results and data at the data processors have been cleaned up.

  • The study is documented in such a way that it is repeatable.

 

image-20240501-132438.png
Process diagram “federated analysis”