Google has announced the launch of MedGemma, a new suite of open AI models aimed at helping developers build applications for medical imaging. The models are designed to support functions such as image analysis, report generation, patient triage, and answering medical image-related questions. MedGemma is available in two versions: a 4B multimodal model for image-based applications and a 27B text-only model for language-based tasks. The multimodal model can assist in fields like radiology, digital pathology, skin image analysis, and fundus imaging, while the text-only model is intended for use cases including patient interviewing, clinical triage, summarization, and decision support.
Google emphasized the importance of adapting the models to specific use cases. The company stated, "MedGemma is a developer model that requires validation on the developer's intended use case. MedGemma can be fine-tuned for improved performance on the existing tasks it's been trained on, or to add additional tasks to its repertoire." The company also noted that the larger, text-only model will generally yield the best performance across most applications. According to research published, AI in medical imaging has improved diagnostic accuracy and personalized care. Other firms like Viz.ai, Aidoc, Lunit, and iCAD are also incorporating AI in healthcare imaging tools.
