I am so proud to be a part of this big multi-disciplinary team, developing foundation models for cross-sectional imaging. Great work from Louis Blankemeier Akshay Chaudhari and so many others.

So excited and grateful to be able to contribute to such an innovative and impactful work! Congrats Louis Blankemeier for leading such a large effort and thank you Akshay Chaudhari for your constant feedback and expert direction! 🥳

Congrats to Louis Blankemeier and mentor Akshay Chaudhari for completing this important project! I’m inspired by your commitment and drive to move AI in Radiology forward. Excited to continue collaborating with you and team.

And here is a big one — VLM for abdominal CTs scans. 🎉 Really interesting work as scaling from 2D to 3D imaging involves its own set of technical challenges (see also: DBT, MRI).

⭐️ Check out Merlin led by Louis Blankemeier and mentor Akshay Chaudhari in this thread. Trained with diagnostic codes and rad reports and evaluated across various tasks comprehensively. Louis Blankemeier really laid a *solid foundation* for CT *foundation* models. :)

Congratulations Louis Blankemeier Akshay Chaudhari Curt Langlotz Amazing work. “Merlin,” a cutting-edge vision language model for 3D CT scans! 🩻✨ Trained with 1.8M+ EHR codes and 6M+ radiology report tokens, Merlin excels in classification, disease prediction, and report generation.

Not only is this Foundation Model work novel and impressive, but we only needed 1 GPU to run the training. Great work led by Louis Blankemeier.

🌟Nice study advancing #VLMs in #radiology with a 3D model for interpreting abdominal CT exams that leverages structured electronic health record (EHR) data and unstructured radiology reports for supervision.

Most doctors are often relying on summaries to navigate the overwhelming amount of electronic health record data. A new study shows AI's potential to assist physicians, reducing errors and improving patient care. stanford.io/3XoTh71

Inspired by RoPE for 1D sequences, Sophie Ostmeier & Brian Axelrod co-developed LieRE. This work leverages Lie groups for building positional embeddings for higher dimensional data. LieRE can flexibly improve training efficiency, label efficiency, and/or compute for transformers!

Open-source, to boot! 🚀 Any one can use this for internal QI projects or to build your own institutional AAA database/registry on Abd/Pelvic CT scans with and/or without contrast. Congrats Arash on winning second place! 🏆

In case you missed it, our team in collaboration with Akshay Chaudhari, Roxana Daneshjou MD/PhD, Oliver Aalami, and Vishnu Ravi, MD defined a framework for autonomous EHR systems, and implemented the first Level 1 EHR system with retrieval and extractive capabilities! Check it out!

📢 Our new 𝗿𝗲𝘃𝗶𝗲𝘄 𝗼𝗻 𝗱𝗲𝗲𝗽 𝗹𝗲𝗮𝗿𝗻𝗶𝗻𝗴 𝗳𝗼𝗿 𝗠𝗥𝗜 is now officially published! Here's the full link: link.springer.com/article/10.100…

Our RoentGen paper is now out in Nature Biomedical Engineering with more exps than before! There's lots of excitement/debate in best practices for synthetic data in training next-gen foundation models. We hope our work in medical imaging can add to this dialogue! Congrats Christian Bluethgen+Stanford Radiology team!

📅 Mark your calendar for the Sept. 4 #RadAIchat at 8 PM ET on "Checklist for #AI in Medical Imaging (CLAIM): 2024 Update – Part II" moderated by Mike Klontzas and panelists Tugba Akinci D'Antonoli Anthony A Gatti Linda Moy #radiomics #AI #DL #radres RSNA Charles Kahn, MD Hesham Elhalawani

The #AIMICenter celebrates #WomenInMedicineMonth! Here's to the incredible women of our center & beyond who are driving innovation & improving patient outcomes every day! You are an inspiration to us all. #WIMMonth #StanfordWIM #WomenInAI #WomenInStem #WIMM24

Featuring the AIMI Center's Akshay Chaudhari, Curt Langlotz, Tina Hernandez-Boussard, Olivier Gevaert & James Zou, Stanford Medicine Magazine explores how #GenAI, when used judiciously, can create data from scratch to enhance #MedicalAI. stan.md/3XyEkyQ

Have you ever used LLMs to aid in your own healthcare? We want to hear from you! stanforduniversity.qualtrics.com/jfe/form/SV_1H…

This Stanford Medicine Magazine article on the evolving use of #GenAI in medicine includes insights from #StanfordCDH Pilot Grant Awardee Akshay Chaudhari & Affiliate Faculty Members Tina Hernandez-Boussard, James Zou & Curt Langlotz. See it here: stan.md/3XyEkyQ