Getting ready to present data for a new exam you will LOVE! 💗 AAGL Mayo Clinic Radiology Mayo Clinic OB/GYN SRU Radiology Society of Abdominal Radiology | SAR Society of Gynecologic Surgeons ASRM Radiology #endometriosis #pelvicpain AAGLJMIG tatnai burnett Daniel Breitkopf Isabel Green MD, MEHP Sarah (Cohen) Rassier MD MPH

🚀 @TheoDapamede and I are thrilled to announce #RadPrompter - a Python package for simplified, reproducible, and composable #LLM prompting, providing a flexible framework for crafting complex prompts from reusable components.

1/7

Google struck back!

CT report generation (3D) leveraging the long context length (>1M) of Med-Gemini (🤯) among other 2D tasks.

Should be a very interesting read…

🔆 I will discuss some fun and important #AI topics today at room 306 #ARRS24 !

1. Fake it till you make it with GAN models!

2. Are AI models sure about what they predict? (Uncertainty quantification)

3. Can we explain what AI percepts?

ARRS

So far, the most impressive paper on the intersection of #AI , computer vision, and radiology in 2024; with an interesting approach, many robust evaluations, and offering a bright future for the application of #FoundationModels in radiology.

So proud of my amazing friend ManaMoassefi for delivering three insightful presentations at #ARRS24 on #uncertainty , #generativeai , and #interpretationai . Always a great learning experience!
MayoAILab Mayo Clinic Radiology

Shahriar Faghani MD @slowvak highlight the need for meticulous stress testing of #DeepLearning models for medical imaging doi.org/10.1148/ryai.2… MayoAILab #MSKRad #AI #MachineLearning

External #validation doesn't always guarantee generalizability!🤯
But why? And how can we ensure #AI model #trustworthiness and #generalizability ? Find out in this commentary. MayoAILab Mayo Clinic Radiology Penn Radiology #Radiology

Check out our collection of articles on 'Mitigating Bias' from MayoAILab Brad Erickson pubs.rsna.org/page/ai/mitiga… #ML #DeepLearning #Radiomics

A farewell lab meeting with pizza and conversations about the bright future of our fellows!🍕✨

Good luck in all your future endeavors, Kevin Zhang, Pouria Rouzrokh, MD , Shahriar Faghani MD and Bardia Khosravi, MD!🦾

And here is the link to our paper on Hugging Face Papers. Please Upvote if interested:

huggingface.co/papers/2404.13…

AK

In our recent publication, we explored how adding heterogeneity in the training set helps #glioblastoma tumor #segmentation generalization!

mcpdigitalhealth.org/article/S2949-…

Thanks to Shahriar Faghani MD and Gian Marco Conte and special thanks to my amazing mentor Brad Erickson!

Stress testing on an award-winning #DeepLearning model for bone age doi.org/10.1148/ryai.2… Shahriar Faghani MD MayoAILab Brad Erickson #PedsRad #BoneAge #AI

At the core of RadRotator is a conditional #diffusion model trained on >15 million pairs of Digitally Reconstructed Radiographs (DRRs) derived from 5,000 CT volumes. This extensive training enables the model to rotate an input XR according to a specified 3D transformation.
(2/3)

Excited to share the technical report for RadRotator, our latest #GenerativeAI tool, which enables the rotation of radiographs in 3D space💫

✅Technical Report:
🔗arxiv.org/abs/2404.13000

✅Website:
🔗pouriarouzrokh.github.io/RadRotator

✅Online Demo:
🔗huggingface.co/spaces/Pouriar…

(1/3)

With many thanks to all the valuable help from Bardia Khosravi, MD, who was my co-first author on this work, and the contributions of all colleagues and mentors from MayoOSAIL and MayoAILab.

Shahriar Faghani MD, Cody Wyles, Brad Erickson, Mayo Clinic Radiology.

Shahriar Faghani MD @slowvak highlight the need for meticulous stress testing of #DeepLearning models for medical imaging doi.org/10.1148/ryai.2… MayoAILab #robustness #ML #MachineLearning

Commentary on a study testing the robustness of #DeepLearning model to estimate bone age doi.org/10.1148/ryai.2… Shahriar Faghani MD MayoAILab Brad Erickson #MSKRad #PedsRad #MachineLearning

You can also follow this 🧵 by Shahriar Faghani MD to learn more about our work, as well as a brief introduction to #LLMs , #RAG , and uncertainty quantification with #conformalprediction .

x.com/ShahriarFaghan…

Interested in how to reliably use large language models ( #LLMs ) & mitigate their hallucinations? Then join Brad Erickson, Shahriar Faghani MD, & me as we introduce #CONFLARE , our recently proposed technique for measuring the uncertainty of content retrieval with LLMs.
#trustworthiness