Rafe McBeth

5.4: ….goblins….

@marslain Currently focused on updating tooling we already use for autoplanning. Early days but very excited about the direction.

@RafeMcbeth @bcherny Interesting. What are the use cases you tried?

Claude Code for Cancer? Claude can now control a cancer fighting radiation shooting robot.

@marslain @bcherny Yep. Lots of plumbing but its working.

@RafeMcbeth @bcherny Did you connect Claude to Eclipse and Truebeam?

Two figures. An unknown path ahead. Five years in, and the course has crystallized around one question: who writes the field guide? Training clinical scientists to answer that.

sf → {surf | kook | ai | medicine}. waves? coffee?

Cultivating next gen of health professionals and AI in Med scientists! @PennMedPhys @mcbethlab at Penn High School Summer Program explaining the role of AI in clinic and future pathways in #RadOnc and medicine. Excited to teach and mentor HS students through the summers!

The authors of a new Perspective report on the Development, Evaluation, and Assessment of Large Language Models (DEAL) checklist, which ensures comprehensive and consistent reporting of LLM studies. Learn more: nejm.ai/3RZB0tc

Starting #SIIM25 strong with talking about what’s Hip and what’s Hype in LLM research trends! Thanks to @RafeMcbeth, @DaniaDaye, and Hana Haver!

🚨 New in @NEJM_AI! We are excited to present the DEAL Checklist — a reproducibility & reporting framework for medical LLM research. Read the full article: tinyurl.com/zxvzp2b4 Let’s make #LLM science transparent & rigorous.

🚀 AI is transforming proton therapy! Join experts at #2025NPC to explore AI’s impact on treatment planning, efficiency & patient access. 🗓 April 27: 1:45– 2:30 PM 🎤 Dr. Rafe McBeth & Dr. William Green of @PennMedicine @PennCancer Register now! ⬇️ 🔗 nationalprotonconference.com

Don't miss our session at #SIIM25! Talking about what's "Hype" and what's actually "Hip" when it comes to #LLMs in #radiology With @DaniaDaye, Hana Haver, and @RafeMcbeth annualmeeting.siim.org/sessions/hip-v…

@JakobWasserthal Great idea. Recently built something like this for our in-house models. Will ask around and see if any lab members want to contribute

Introducing TotalSegmentator Annotate: a simple web-based tool where anyone can help improve segmentation data. Your contributions directly help us train a better, smarter TotalSegmentator. Start making an impact today. (Link in the comments)

@parmita Same. We talk about this often in Radiation Oncology. The physics error bars are much much tighter than the biology error bars and it makes us sad that clinical ground truth is uhhh.. fuzzy

I wish biology was like physics. I wish the objective truth could be assumed. I wish I knew what I knew is correct no matter what. I wish I could look at mRNA expression and PV=nRT it

Check out the latest edition of our newsletter - Radiation Communication - to learn about our student’s upcoming #AAPM2024 presentations, our global #medphys training program and more! tinyurl.com/y59mf823 @RafeMcbeth @KassaeeAK @CBaffic @michelemkim @Dr_SAvery

@RafeMcbeth talks about the last mile for AI adopting and integratation in the #Radonc clinic! #SIIM24 @mcbethlab @PennMedicine @PennCancer @penn

A great line up of talks and good wisdom in the morning session on #LLMs in the clinical workflow. Focusing both on DOs and DONT’s of LLM research. #SIIM24 Thanks to @RafeMcbeth, @asset25, and @DaniaDaye for making this is a success!

#SIIM24 @RafeMcbeth helping us understand the last mile to getting LLMs into the clinical workflow in #radiationoncology - one of the many specialities represented at @SIIM_Tweets

You don't want to miss this session! Exploring some crucial components of #radiology #LLM research and ways of integration in the clinical workflow. #siim24 @asset25 @DaniaDaye @RafeMcbeth @SIIM_Tweets

Thanks for posting and helping to organize @_DrOReilly_ I had a lot of fun with @YVerginadis