It was certainly a good choice to extend this year’s EuSoMII Annual Meeting to a two-day event. Imaging Informatics, with its many facets, is getting more and more important and there is only so much time in one day.
Day one in Valencia marked a new era in AI in radiology by focusing on integration [1]. Where in lasts year’s meeting the focus was mainly technical developments, this meeting’s presentations revolved more around practical aspects. It is clear that the time has come to acknowledge and address the ‘crux of the issue’: responsible implementation into clinical practice. Again, the first half of the day was composed of educational lectures focusing on cutting edge technologies, but mostly on their ethical, legal and societal implications.
The first speaker of the day, Bram van Ginneken of Radboud UMC Nijmegen, opened the discussion on if and when algorithms should be employed autonomously. He advocated directing research efforts towards cost-benefit analysis; for example, to address the questions of “How much should an algorithm cost?” and “How much benefit does it add?” – a relevant statement in times where ‘value-based healthcare’ is being discussed. This laid the perfect foundation for the second speaker, Angel Alberrich-Bayarri (Valencia, Spain), who presented some historical context on imaging biomarkers research in medical imaging and provided suggestions on how the performance of algorithms could potentially be surveyed and monitored after implementation into clinical routine, as well as what legal frameworks already apply (e.g. GDPR). He specifically recommended treating algorithms as drugs, in the sense that the same requirements for rigorous evaluation in trials should apply, and, additionally, post-marketing surveillance to guarantee the continuous performance of AI tools.
Following this introduction of imaging biomarkers, Bettina Baeßler (Zurich, Switzerland) provided a more practical and in-depth look at radiomics research echoing the issues raised during the first day’s keynote lecture [2] by outlining a stepwise approach to ensure standardized, accurate, robust and repeatable research in this field; welcome advice given the current ‘reproducibility crisis’ in medical research in general. The educational part of the event closed with a very practical talk on the value of dose management and optimization by Federica Zanca (Leuven, Belgium), and a summarizing overview by EuSoMII’s president, Erik Ranschaert, on how IT solutions, in general, can help to improve efficiency and quality of patient care in radiology.
Another major difference from previous EuSoMII meetings was that, for the first time, participants were offered the possibility to take part in a written exam. The exam covered all topics from the educational lectures and should be a first step towards the eMIIP certification that was mentioned on the first day. Although the certification procedures have not yet been finalized with the ESR, this year’s exam offered a first impression of how members would respond to such offers. Not surprisingly, almost all attendees opted to take part underlining the interest for further education on imaging informatics topics and the need for a corresponding subspecialization.
After this already packed half-day, attendees were again provided with a complimentary lunch and the parallel poster exhibition. Following this short break, the afternoon started off with a special talk by Raym Geis from the Society for Imaging Informatics in Medicine (SIIM), summarizing the recently published multi-society statement on ethics in AI [3]. From the audience’s engagement in the discussion of this topic, it was very clear that after the initial hype about AI in radiology, many more differentiated views and opinions are now being voiced.
The afternoon then continued with the second joint session featuring The European Society of Musculoskeletal Radiology (ESSR), moderated by Mario Maas from Amsterdam UMC, who acts as ‘Chair Artificial Intelligence Initiative’ at the ESSR. Topics included applications of algorithms in spinal cross-sectional imaging and sports medicine, especially automatic diagnosis in knee MR. This part of the meeting also featured Oge Marques as a special guest, presenting on the ACR’s AI-Lab, and a fantastic keynote lecture by Tim Leiner from UMC Utrecht on how to bring AI to the clinic (see upcoming post). After that, the day closed with a wide range of scientific presentations.
All in all, the EuSoMII Annual Meeting fits well with other events dedicated to imaging informatics and AI, such as the ESR/ESOR premium AI event [4]. There is clearly a need for more in-depth education and discussion of such topics. Certainly, next year’s EuSoMII Annual Meeting will be able to continue this year’s success, where a record number of submitted abstracts and attendees was achieved.
To learn more about EuSoMII, go to https://www.eusomii.org. Stay tuned for an upcoming post summarizing the keynote lecture from day two.
[1] https://ai.myesr.org/education/a-look-back-at-the-eusomii-annual-meeting-2019/
[2] https://ai.myesr.org/education/keynote-lecture-at-eusomii-annual-meeting-2019-imaging-biomarkers-and-radiomics-source-of-big-data-for-ai/
[3] https://ai.myesr.org/publications/the-ethics-of-ai-in-radiology-a-multisociety-statement/
[4] https://www.maiesr.org
