Announcing CT Foundation, a new medical imaging embedding tool that accepts a computed tomography (CT) volume as input and returns a small, information-rich numerical embedding that can be used to rapidly train models. Learn more and try it out yourself →

Proud of @Google today

Great! Next up, MRI Foundation?

The researchers aim to expand the capabilities of Gemini, a powerful multimodal model, to handle medical tasks. They create a new family of models called Med-Gemini that can process diverse medical data like images, genomics, and text. Med-Gemini-2D sets a new standard for chest X-ray report generation, with 57% and 96% of AI reports on normal and abnormal cases evaluated as "equivalent or better" than original radiologist reports. Med-Gemini-2D also outperforms previous models on visual question answering and image classification tasks. Med-Gemini-Polygenic outperforms the standard linear approach for disease risk prediction and generalizes to new diseases. full paper:

This sounds like a game-changer for medical imaging! CT Foundation seems like a powerful tool that can really speed up model training.

This could drastically reduce the time and data needed to develop new diagnostic tools.

yes! more of this and less silly podcast generators

This is exactly the kind of progress we need!

What does the embedding itself point to, can it be used on its own?

@sundarpichai It's amazing

In the 70's the scientific community believed we would be able to transcribe dreams. Good work, by the way!