Now online in Cancer Discovery @aacrjournals.bsky.social - Machine Learning Predicts Hepatocellular Carcinoma Risk from Routine Clinical Data: A Large Population-Based Multicentric Study - by @janclusmann.bsky.social, @jnkt.bsky.social, Carolin Schneider, and colleagues doi.org/10.1158/2159...

Wrapped up #CCESummSchool25 with inspiring talks on early cancer + recognition session! Congratulations to Sena Alptekin @dkfz.bsky.social, Marck Opdam @nkinl.bsky.social, @janclusmann.bsky.social for Best Posters and Francesca Cogo @vhio.bsky.social for Best Participant Talk! 🎉

Thank you for sharing! It was a pleasure working on this, together with @twittlabs.bsky.social , @jnkt.bsky.social and many others...

Thank you very much for sharing! Was a pleasure working together with @twittlabs.bsky.social and @jnkt.bsky.social (and many others) on this project :)

@ekfzdigital.bsky.social

Great visit to EKFZ Digital Health Institute! Gave a talk at Jakob Kather’s lab, had great 1:1s with team members, and visited our PhD student Elisa Merelli, who's spending part of her PhD here💡🔬 #DigitalHealth #AI #Research
@jnkt.bsky.social @janclusmann.bsky.social

Finally out, in NEJM AI!
Everyday labels and watermarks can override medical knowledge of vision language models, just by their presence.

Was a lot of fun to work on this with @twittlabs.bsky.social , @jnkt.bsky.social and many others

The autoimmune hepatitis field is quite unique in the way clinicians and researchers on AIH are connected personally worldwide. Optimal starting point for generalizable prediction algorithms ...

Anger, despair, and defiance from a voice within the US federal research system People around the world are watching with disbelief as the new US government closes down its aid programme, withdraws from the World Health Organization (WHO) and the Paris agreement on climate change...

www.bmj.com/content/388/...

The disruptions of free research in the US is concerning, for the whole scientific community. Highly recommended opinion article to understand, why the voices of US science have been so quiet lately.

So happy to share this: secret instructions hidden in images can terribly alter the output of VLMs to create misdiagnosis. Models have different susceptibility, though, with Claude3.5 apparently being much better aligned to ethical outputs than GPT4o.

#aisafety
#LLMs
#VLMs