AI in Medical Imaging

Abstract

The research, described in "Advancing Multimodal Medical Capabilities of Gemini" (published on arXiv and the Google DeepMind research blog), adapted general-purpose large multimodal models for clinical use. The goal was to interpret medical data such as images from radiology, histopathology, ophthalmology, and dermatology. These images are not typically included in general LMMs.   Methods: Researchers developed several models within the Med-Gemini family. They optimized these models using a large and diverse set of biomedical data, including 2.2 million data samples in one suite.