Google Expands Healthcare AI with Two New Open Models
Google has launched two new artificial intelligence models focused on healthcare. The company named them MedGemma 1.5 and MedASR. This move is part of Google's expanding efforts in medical AI. Unlike some rivals that offer healthcare AI tools primarily as paid enterprise services, Google has opted for a more open approach. The company is releasing both models publicly for the wider research and developer community.
MedGemma 1.5 Targets Medical Images and Text
MedGemma 1.5 is the latest version of Google's medical vision-language model. It is built to analyse medical images alongside written information. The model can interpret scans and respond to questions related to visual medical data. It also assists with a range of research-oriented tasks.
According to Google Research, the updated version brings improved multimodal reasoning. It offers better performance when dealing with complex medical imagery. The model is also designed to be more flexible. This allows researchers to fine-tune it for specialised datasets and specific study requirements.
The model supports multiple forms of medical imaging. These include radiology scans and other clinically relevant visuals. Google said MedGemma 1.5 is intended for uses such as image-based question answering, report drafting, and structured data extraction. The company stressed that it is not meant to provide diagnoses or treatment advice. It should only be used as a support tool in research and development settings.
MedASR Focuses on Clinical Speech Recognition
Alongside MedGemma 1.5, Google introduced MedASR. This is an automatic speech recognition model designed specifically for healthcare environments. MedASR is built to transcribe spoken clinical conversations into text. It pays particular attention to medical terminology, diverse accents, and the challenges of real-world clinical audio.
Google said the model aims to reduce transcription errors. These errors often occur when general-purpose speech recognition systems are used in medical contexts. Potential use cases include transcribing doctor-patient discussions. It can also create clinical notes and convert dictated reports into text.
The company added that MedASR can be adapted for different healthcare settings. Researchers can fine-tune it to match specific clinical workflows or documentation standards.
Open Access for Developers and Researchers
Google said all versions of MedGemma and MedASR are available through Hugging Face and the Vertex AI platform. Developers can also access documentation and tutorials via the MedGemma GitHub repository.
This open access approach supports innovation and collaboration in healthcare AI development. It allows more people to experiment with these tools and potentially improve medical research.
The launch highlights Google's commitment to advancing AI in healthcare. By making these models publicly available, the company hopes to accelerate progress in medical technology. Researchers around the world can now leverage these tools for their studies.
