Requirements Federated Learning and mUlti-party computation Techniques for prostatE cancer, published by HL7 Europe. This guide is not an authorized publication; it is the continuous build for version 0.1.0 built by the FHIR (HL7® FHIR® Standard) CI Build. This version is based on the current content of https://github.com/hl7-eu/flute-requirements/ and changes regularly. See the Directory of published versions

4.1 Methodology

Requirements collection: We reused the work performed by the TRUMPET project to define user interface and researcher requirements. Here, we cite TRUMPET’s deliverable D1.1 Section 3, summarize part of it to be self-contained, and present new FLUTE-specific requirements based on discussions with researchers / innovators who will use the platform, in particular statisticians and medical researchers from the three participating hospitals CHUL, IRST and VHIR, as well as data owners, in particular the representatives of the hospital partners CHUL, IRST and CHUL who are aware of ICT aspects, data protection and other requirements for the data owner nodes.

Dimensions considered to be out of scope: FLUTE is a focused effort, and to be efficient a number of dimensions are not considered to be a priority. These include the following:

• Preparation of federated learning: Typically, a data user (researcher, innovator) who wants to study a topic will need to search for data owners that have the appropriate data. There exist efforts, e.g., the AI4EU project and related initiatives, to build a platform that will allow data owners to advertise data, users to search for it, and to allow both to agree on collaboration, potentially for a reward or subject to conditions. While this is of interest in a large-scale exploitation phase, FLUTE’s main focus is to ensure privacy-preserving collaboration between multiple data owners, and then to pilot the technology in pre-defined applications. Therefore, at this stage FLUTE will not include such FL preparation features in its interface.

• Graphical user interfaces: All FLUTE data owners and data users are supposed to have the knowledge needed to interface their local software with the FLUTE platform. FLUTE will not develop graphical user interfaces (GUI), as doing so wouldn’t be the most effective use of resources. First, researchers/innovators are usually already using Python for their data analysis, which has much more flexibility than any GUI currently available. Second, hospitals use many different database systems, and while many GUIs exist to manage databases, they are often system-specific. The first priority of the data owners is therefore standardization, which is discussed below in this document. In a future exploitation phase, it will be possible to develop a GUI.