XrayGPT
lastmod 2023-07-05created 2023-07-051 edits
Chest radiograph summarization using a medical vision-language model (VLM)
Features
- Research on automatically analyzing chest X-ray radiographs
- Fine-tunes the Vicuna LLM on medical data (100,000 real patient–doctor conversations) and 30,000 radiology conversations
- Improves LLM performance by generating clean, conversational summaries (~217k) from free-text radiology report datasets (MIMIC-CXR and Openl)
- Aligns a Frozen Medical Visual Encoder (MedClip) with the fine-tuned LLM (Vicuna) (using a Simple Linear Transformation)