Semiautomated breast ultrasound report generation using multimodal large language models and deep learning

The proposed framework enables accurate, efficient, and device-adaptable breast US report generation by combining multimodal DL and prompt-based LLM inference. It significantly reduces radiologist workload and demonstrates potential for scalable deployment in real-world clinical workflows.