Google announced the launch of MedGemma, which the company says contains the most capable open models that help developers build healthcare-based AI applications to examine medical images, generate medical image reports, triage patients and answer questions about medical images.
The open models, which users can fine-tune, come in two variants: a 4B multimodal version and a 27B text-only version.
The tech giant says the multimodal version can be used as a starting point for apps classifying medical images in radiology, digital pathology, skin images and fundus.
The models can also be used for apps aimed at generating medical reports and answering questions about images.
The text-only version can be a launching point for applications that require medical knowledge, such as those aimed at interviewing patients, triaging and summarization, as well as for clinical decision support.
MedGemma is a developer model that requires validation on the developer's intended use case.
Based on those validation results, the user will likely need to further adapt the model to improve performance.
【MORE】