R2GenCSR: Mining Contextual and Residual Information for LLMs-based Radiology Report Generation

Researchers have developed R2GenCSR, a new AI framework for generating radiology reports that uses Mamba architecture instead of Transformers to reduce computational complexity while maintaining performance. The system leverages context retrieval and large language models to produce high-quality medical reports from X-ray images.

• →R2GenCSR introduces Mamba as vision backbone with linear complexity, offering comparable performance to Transformer models with lower computational costs.
• →The framework performs context retrieval during training using both positive and negative samples to enhance feature representation.
• →The system combines vision tokens, context information, and prompts to generate medical reports using large language models.
• →Extensive testing on three X-ray datasets (IU X-Ray, MIMIC-CXR, CheXpert Plus) validated the model's effectiveness.
• →Source code has been made publicly available for research and development purposes.