Researchers have created CT-RATE, the first dataset pairing 3D chest CT scans with radiology reports, enabling powerful new AI foundation models for medical imaging analysis.

• Developed CT-CLIP for encoding 3D medical images with corresponding text reports
• Created CT-CHAT, a chat-based interface for direct textual interaction with 3D medical scans
• Built from 25,692 chest CT scans from over 21,000 patients
• Bridges the gap between advanced computer vision and medical imaging AI

This research significantly advances medical imaging interpretation by bringing the power of large language models to radiology. These foundation models could enhance diagnostic accuracy, improve workflow efficiency, and enable more intuitive interaction with complex 3D medical data.

Developing Generalist Foundation Models from a Multimodal Dataset for 3D Computed Tomography