Next generation medical image interpretation with MedGemma 1.5 and medical speech to text with MedASR

The adoption of artificial intelligence in healthcare is accelerating dramatically, with the healthcare industry adopting AI at twice the rate of the broader economy. In support of this transformation, last year Google published the MedGemma collection of open medical generative AI models through our Health AI Developer Foundations (HAI-DEF) program. HAI-DEF models like MedGemma are intended as starting points for developers to evaluate and adapt to their medical use cases, and they can be easily scaled on Google Cloud through Vertex AI. The response to the MedGemma release has been incredible, with millions of downloads and hundreds of community-built variants published on Hugging Face.

Today, we’re building on that momentum by releasing MedGemma 1.5 4B and launching the MedGemma Impact Challenge hackathon on Kaggle. Guided by direct feedback from the community, this model update enables developers to more effectively adapt MedGemma for applications that involve several medical imaging modalities:

• High-dimensional medical imaging: Computed tomography (CT), magnetic resonance imaging (MRI), and histopathology
• Longitudinal medical imaging: Chest X-ray time series review
• Anatomical localization: Localization of anatomical features in chest X-rays
• Medical document understanding: Extracting structured data from medical lab reports

MedGemma 1.5 4B also improves accuracy on core capabilities for text, medical records and 2D images over MedGemma 1 4B. We are publishing the updated 4B model size today to provide an ideal compute-efficient starting point for developers that is small enough to run offline, and developers can continue to use our MedGemma 1 27B parameter model for more complex text-based applications. Full details of the MedGemma 1.5 4B model and performance benchmarks appear in the MedGemma 1.5 model card.