pipette
ENEnglish

An open vision-language model for diverse medical applications

Andrew Sellergren, Sahar Kazemzadeh, Fereshteh Mahvar, Atilla P. Kiraly, Madeleine Traverse, Timo Kohlberger, Chufan Gao, Ronnachai Jaroensri, Shawn Xu, Fayaz Jamil, Cían Owen Hughes, Charles T. Lau, Justin Anthony Chen, Liron Yatziv, Tiffany Chen, Bram Sterling, Sathaiah Baby, Susanna Maria Baby, Jeremy Lai, Samuel Schmidgall, Lu Yang, Lu Yang, Kejia Chen, Per Bjornsson, Shashir Reddy, Ryan Brush, Kenneth A. Philbrick, Bashir Sadjad, Mercy Asiedu, Ines Mezerreg, Howard Hu, Howard Yang, Richa Tiwari, Sunny Jansen, Preeti Singh, Yubin Kim, Daniel McDuff, Catherine Kozlowski, Rachelle Sico, Aishwarya Kamath, Johan Ferret, Shreya Pathak, Nino Vieillard, Ramona Merhej, Sarah Perrin, Tatiana Matejovicova, Alexandre Ramé, Morgane Rivière, Louis Rouillard, Thomas Mesnard y 10 más

Revista con revisión por paresUso en el mundo real

En palabras de los autores

Abstract Artificial intelligence has high potential for impact in healthcare applications, but its training and deployment are challenging due to diverse data, a complex spectrum of possible tasks and important privacy needs. High-performing foundation models that enable data-efficient fine-tuning for diverse downstream tasks can meaningfully accelerate development in this domain. We introduce MedGemma, a collection of medical vision-language foundation models based on Gemma 3. MedGemma demonstrates advanced medical understanding and reasoning across images and text and multiple medical imaging domains, exceeding the performance of similarly sized generative models while maintaining the general capabilities of the Gemma base models. For out-of-distribution tasks, MedGemma achieves improvements of 2.6–10% in medical image question answering, 15.5–18.1% in chest X-ray finding classification and 10.8% in agentic evaluations compared with the base models. Our results show that fine-tuning MedGemma can be more effective than fine-tuning the base Gemma 3 model for medical tasks, particularly in the setting of limited training data. We additionally introduce MedSigLIP, a medically tuned vision encoder derived from SigLIP. MedSigLIP powers the visual understanding capabilities of MedGemma and, as an encoder, achieves performance comparable to or better than that of many specialized medical image encoders. Taken together, the MedGemma collection provides a strong foundation of medical image and text capabilities, with the potential to accelerate medical research and the development of downstream applications.

Resultado principalEl resumen no menciona limitaciones.

Apareció: jueves, 8 de octubre. Nature Medicine. Revista con revisión por pares.

DOI: 10.1038/s41591-026-04626-w