←Back to feed
🧠 AI🟢 Bullish
Merlin: A Computed Tomography Vision-Language Foundation Model and Dataset
arXiv – CS AI|Louis Blankemeier, Ashwin Kumar, Joseph Paul Cohen, Jiaming Liu, Longchao Liu, Dave Van Veen, Syed Jamal Safdar Gardezi, Hongkun Yu, Magdalini Paschali, Zhihong Chen, Jean-Benoit Delbrouck, Eduardo Reis, Robbie Holland, Cesar Truyts, Christian Bluethgen, Yufu Wu, Long Lian, Malte Engmann Kjeldskov Jensen, Sophie Ostmeier, Maya Varma, Jeya Maria Jose Valanarasu, Zhongnan Fang, Zepeng Huo, Zaid Nabulsi, Diego Ardila, Wei-Hung Weng, Edson Amaro Junior, Neera Ahuja, Jason Fries, Nigam H. Shah, Greg Zaharchuk, Marc Willis, Adam Yala, Andrew Johnston, Robert D. Boutin, Andrew Wentland, Curtis P. Langlotz, Jason Hom, Sergios Gatidis, Akshay S. Chaudhari|
🤖AI Summary
Stanford researchers introduced Merlin, a 3D vision-language foundation model for analyzing abdominal CT scans that processes volumetric medical images alongside electronic health records and radiology reports. The model was trained on over 6 million images from 15,331 CT scans and demonstrated superior performance compared to existing 2D models across 752 individual medical tasks.
Key Takeaways
- →Merlin is the first 3D vision-language model specifically designed for abdominal CT scan interpretation, addressing radiologist shortage challenges.
- →The model was trained on a massive clinical dataset including over 6 million CT images, 1.8 million diagnosis codes, and 6 million tokens from radiology reports.
- →Merlin outperformed existing 2D vision-language models and CT foundation models across 752 individual diagnostic, prognostic, and quality-related medical tasks.
- →The researchers validated the model at scale using both internal testing on 5,137 CT scans and external testing on 44,098 CT scans from multiple independent sites.
- →Stanford has open-sourced the trained models, code, and dataset, potentially accelerating development of AI-powered medical imaging tools.
#medical-ai#computer-vision#healthcare#foundation-model#medical-imaging#stanford#open-source#ct-scans#radiology#3d-vision
Read Original →via arXiv – CS AI
Act on this with AI
Stay ahead of the market.
Connect your wallet to an AI agent. It reads balances, proposes swaps and bridges across 15 chains — you keep full control of your keys.
Related Articles