RT @AkshayGoelMD: 🚀 Excited to share two open-source releases today from our team at @GoogleResearch . • 𝐋𝐚𝐧𝐠𝐄𝐱𝐭𝐫𝐚𝐜𝐭 – Gemini-powered infor….

Radiology reports were designed to be a means of physician-to-physician communication. To be useful to patients, the reports need to be translated into clear explanations free of medical jargon. Enter RadGPT. Developed by Stanford Radiology clinicians and researchers, RadGPT.

Excited to share that our paper "Beyond the Prompt: Deploying Medical Foundation Models on Diverse Chest X-ray Populations" is accepted at #MIDL2025 and will be presented at the conference next week!. In this work, we explore how to reliably deploy foundation models in real-world.

RT @_akhaliq: SMMILE. An Expert-Driven Benchmark for Multimodal Medical In-Context Learning

RT @Michael_D_Moor: 🚨New preprint! 🚨In-context learning (ICL) is the intriguing ability of LLMs to learn to solve tasks purely from context….

You can now transform a report into structured plain language, i.e.:. Mild pulmonary edema with probable small bilateral pleural effusions. More focal opacities at lung bases may reflect atelectasis but infection cannot be completely excluded. To:.Observation: mild edema.

Documentation of F1-RadGraph has been improved as well. Typically, the RG_ER reward has become widely used (CheXagent, Maira-2, GREEN,.LLaVA-Rad,. ).

RadGraph now works with modernBert, just call:.model_type = "modern-radgraph-xl".radgraph = RadGraph(model_type=model_type).

💥 RadGraph repo just leveled up!.The latest update now runs on ModernBERT, crushing long medical reports with ease. Even slightly beats the official RadGraph-XL model from the original paper. Also, less dependencies and updated doc 👇.

This work builds on our recent study on Automated Structured Radiology Report Generation ( which introduces the dataset and evaluation framework.

📄 Paper: 🌐 Project Page: 🤗 Models & Data: All models and datasets are fully open-source — we hope this contributes to the broader medical AI community! 🤝. Huge thanks to the amazing team at Stanford.

💥 Excited to share our latest work: Structuring Radiology Reports: Challenging LLMs with Lightweight Models. In this study, we explore how small, task-specific encoder-decoder models can rival (and sometimes outperform) much larger LLMs in structuring radiology reports; all.

Paper, soon to appear at #ACL2025 main: Project page, with all resources (datasets, models, ontology) and usage notes: All models and datasets are publicly available as open-source:.

4) We conduct a reader study to create a radiologist-validated test set for both the automated structured radiology report task, as well as utterances disease labels from our new ontology. Additionally, we introduce a new evaluation framework, which includes results from our new.

3) We fine-tune popular RRG system on this restructured findings and impression, namely:.- Chexagent @StanfordAIMI.- MAIRA-2 @MSFTResearch.- RaDialog @TU_Muenchen.- Chexpert-plus @StanfordAIMI. As well as a BERT architecture for the disease classification system on our new.

2) Since each reported observation, whether in the findings or impression sections, is expressed as a single utterance (1.5M unique utterances in total), we use a large language model to label each one according to a new ontology comprising 72 critical chest X-ray (CXR).

1) We leverage LLM to restructure MIMIC-CXR and Chexpert-plus (180K Findings sections and 400K Impression sections) into reports categorized by organ system, under strict rules. Example findings:. Lungs and Airways:. - No evidence of pneumothorax. Pleura:. - Bilateral.

💥 We unveil our paper accepted at the #ACL2025 Main Conference:.Automated Structured Report Generation. Let's revisit automated radiology report generation for CXR. Free-form reports make it hard for AI systems to learn accurate generation, and even harder to evaluate. 🧵👇.