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

Khadija Azhar¹Byoung-Dai Lee²Shi Sub Byon¹SeungJae Lee³Kyu Ran Cho⁴Sung Eun Song^4*

• ¹AI Laboratory, HealthHub Co., Ltd, Seoul, Republic of Korea
• ²Division of AI and Computer Engineering, Kyonggi University, Suwon-si, Republic of Korea
• ³Dnotitia Inc., Seoul, Republic of Korea
• ⁴Department of Radiology, Korea University Anam Hospital, Seoul, Republic of Korea

Introduction: Breast ultrasound (US) imaging is essential for early breast cancer detection, yet generating diagnostic reports is labor-intensive, particularly when incorporating multimodal elastography. Methods: This study presents a novel framework that combines multimodal large language models and deep learning to generate semiautomated breast US reports. This framework bridges the gap between manual and fully automated workflows by integrating radiologist annotations with advanced image classification and structured report compilation. A total of 2,119 elastography images and 60 annotated patient cases were retrospectively collected from two US machines. Results: The system demonstrated robust performance in elastography classification, achieving areas under the receiver operating characteristic curve of 0.92, 0.91, and 0.88 for shear-wave, strain, and Doppler images, respectively. In the evaluated dataset, the report generation module correctly identified all suspicious masses across both US machines, achieving 100% sensitivity in lesion detection, with an average report generation time of 31 seconds per patient using the GE Healthcare machine and 36 seconds using the Supersonic Image machine. Discussion: 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.