Follow us on X
Follow us on Linkedin
Real-world-data enabled assessment
for health regulatory decision-making

REALM 2nd Progress Meeting in Heraklion

Heraklion, May 22-24

The second progress meeting of the REALM project (Real-world-data Enabled Assessment for heaLth regulatory decision-Making) was successfully held in Heraklion from May 22nd to 24th. This milestone marks 18 months into the project’s implementation and was the first face-to-face progress meeting with around 30 consortium members, highlighting REALM's efforts to propose solutions for faster and more secure adoption of AI in medical device software.

Recent advancements in artificial intelligence have shown great potential in clinical tasks, driven by developments in big data, supercomputing, sensor networks, brain science, and other technologies. However, large-scale application in clinical practice remains elusive due to the lack of standardized processes, ethical and legal supervision, and other challenges. The REALM project aims to establish a process framework to ensure the safe and orderly development of AI in the medical industry. This framework will facilitate the design and implementation of AI products, enhance regulatory management, and ensure the selection of reliable and safe AI products for clinical use.

The meeting was opened by the host, Traqbeat, with the main messages delivered by the project coordinator, Michel Dumontier (Maastricht University) and co-coordinator, Gökhan Ertaylan (VITO).

The three-day agenda of the meeting was packed with key activities, including plenary meetings, working group sessions, and hands-on demonstrations. On the first day, plenary sessions were held with detailed reporting on each Work Package (WP), focusing on the progress and challenges encountered. The day concluded with hacking pitches to organize the work for the second and third days.

The second day featured intensive discussions in working groups on several critical topics:

  • COPD Data Transformation and ML Competition: Methods for transforming COPD data and preparations for the upcoming machine learning competition.
  • Evaluation of the RIANA Dashboard: Presentation and demo of the current version of the RIANA dashboard, followed by a QA session, direct feedback, and hands-on testing on the test server.
  • Implementation and Evaluation of Synthetic Data in REALM Infrastructure: Assessment of the use of synthetic data within the REALM infrastructure.
  • Evaluation of ML-Enabled Medical Devices: Raising awareness about the limited availability of independently verified evaluations of ML-enabled medical devices, especially in Europe. Highlighting the need for crucial information for clinicians to make informed decisions.
  • Blockchain for Healthcare: Ensuring transparency, traceability, and accountability. Implementation of model cards for clinical transparency, patient safety, and overall healthcare integrity. Focus on integrating blockchain technology into REALM applications.
  • OMOP Mappings for User Claims: Introduction to mapping user claims into OMOP terminology to create standard and reproducible queries for AI testing. Included a hands-on example with the TraqBeat AI model and a discussion on challenges and automation possibilities.
  • Ethics Diary: Ensuring compliance with the AI Act’s ethical and legal principles and exploring ways to connect ethical considerations with practical implementations.

The third day started with sessions from Anshu Ankolekar (Maastricht University) about Enhancing Healthcare through Living Labs: Real-World Testing and User-Centric Innovation and Ivett Jakab (YAGHMA) who reported on the outcomes of the Lorentz Centre Life Sciences with Industry Workshop 2024 where YAGHMA participated with their challenge.

Second part of the day included parallel sessions on different topics:

  • Stakeholder Questionnaire Overview & Q&A Session
  • The Realm Academy
  • Opportunity and risks to REALM from AI regulation

The REALM project's 2nd Progress Meeting was a significant milestone, fostering collaboration, innovation, and shared understanding among consortium members. The discussions and outcomes from these sessions are expected to significantly advance the project's goals and impact, ultimately leading to the safer and faster adoption of AI in medical device software.