ECR24 summary: Outlook on radiology AI

Synwisery’s outlook on radiology AI - where is the value? After numerous discussions with thought-leaders in the field at the European Congress of Radiology (ECR2024), we see a very colorful picture of the state of radiology AI in Europe - a spectrum of advances that lies between the bright “future of radiology” narrative and the pitch-black “the hype is already over” assertions.

A very common perception from the exhibition was that the interest from radiologists has shifted from general inquiries about what AI is and might do in the future to very practical questions about integration and adoption, performance and monitoring, and - perhaps most prominently, concrete value.

For the first two issues, vendors seem very well equipped by now. On the last point, however, we still see a lot of room for improvement and the biggest opportunity for players in the field: what is really in it for radiologists and managers?

Against this backdrop, here are our highlights of this year’s ECR, representing our current perception of current RadiologyAI value in Europe:

• AI for mammography screening shows clear value for double-reading scenarios with recent results in particular those from Screenpoint’s and Lunit’s clinical trials

• X-ray detection tools, such as fracture detection and chest X-ray applications, show impressive adoption records. Reasons for adoption vary, ranging from quality assurance during after-hours work to reducing radiologists’ workload. Gleamer and Annalise stand out here with impressive validation results.

• AI-augmented reporting is moving further into focus, with different approaches from players such as Annalise, Rad AI, Smart Reporting and RadReport. This also prompted questions around regulatory requirements, as voiced repeatedly by Hugh Harvey and others.

• Speaking of regulatory progress, we saw some fascinating advancements from b-rayz, who presented the first CE-marking of a dynamically adapting SaMD product and Oxipit, with the first autonomous imaging AI for high-probability normals in chest-x-ray in actual use at a Vilnius hospital. Glad to see these successes opening new avenues for the field!

• While some discussions at the AIX theater presented a somewhat disappointed perspective on current adoption, Hugues Brat offered an inspiring counterexample. He shared his experiences with the routine use of Incepto at 3R, which includes continuous monitoring, and highlighted the significant clinical impact of AI.

At Synwisery, we are dedicated to identifying the real-world value of healthtech applications and communicating their impact effectively. Reach out to us to discuss how to get the most from your healthtech applications.