A workflow utilizing general-purpose large language models for efficient structuring and data mining of bone scintigraphy narratives.
Authors
Affiliations (13)
Affiliations (13)
- Shanghai Institute of Medical Imaging, Shanghai, 200032, China.
- Department of Nuclear Medicine, Zhongshan Hospital, Fudan University, No. 180 in Fenglin Road, Shanghai, 200032, P.R. China.
- Institute of Nuclear Medicine, Fudan University, Shanghai, 200032, China.
- Cancer Prevention and Treatment Center, Zhongshan Hospital, Fudan University, Shanghai, 200032, China.
- Department of Nuclear Medicine, Shanghai Xuhui Central Hospital, Fudan University, Shanghai, 200237, China.
- Department of Nuclear Medicine, Fujian Provincial Hospital, Shengli Clinical Medical College of Fujian Medical University, Fuzhou University Affiliated Provincial Hospital, Fuzhou, 350001, China.
- Fujian Research Institute of Nuclear Medcine, Fuzhou, 350001, China.
- Digital Medical Research Center, School of Basic Medical Sciences, Fudan University, No. 138 in Yixueyuan Road, Shanghai, 200032, P.R. China. [email protected].
- Shanghai Key Laboratory of MICCAI, Shanghai, China. [email protected].
- Shanghai Institute of Medical Imaging, Shanghai, 200032, China. [email protected].
- Department of Nuclear Medicine, Zhongshan Hospital, Fudan University, No. 180 in Fenglin Road, Shanghai, 200032, P.R. China. [email protected].
- Institute of Nuclear Medicine, Fudan University, Shanghai, 200032, China. [email protected].
- Cancer Prevention and Treatment Center, Zhongshan Hospital, Fudan University, Shanghai, 200032, China. [email protected].
Abstract
Whole-body bone scintigraphy is pivotal for skeletal evaluation in oncological monitoring, yet the unstructured nature of clinical reports impedes efficient data management and multicenter integration. This study aims to validate whether a clinical-logic-guided prompting framework can effectively constrain general-purpose large language models (LLMs) to achieve reliable structured information extraction from bone scintigraphy narratives without domain-specific fine-tuning, and to evaluate the performance of this workflow in real-world clinical scenarios. We first established a multicenter ground-truth dataset to benchmark four LLMs (DeepSeek-R1, DeepSeek-V3, GPT-o3, and Gemini 2.5 Pro), and DeepSeek-R1 demonstrated the highest accuracy and stability in structured extraction. Subsequently, using DeepSeek-R1, we executed two validation tasks. In a human-in-the-loop workflow, LLM assistance reduced manual processing time by 74.5%-82.6% while significantly enhancing accuracy. Furthermore, we constructed a bone metastasis atlas for eight common malignancies through the automated processing of data from initial 18,331 patients. Our study demonstrates that prompt engineering designed by clinical experts, integrating clinical logic and controlled vocabularies, can effectively guide general-purpose LLMs for bone scintigraphy narrative information extraction. This approach provides a validated low-code paradigm for physicians to transform medical narratives into analyzable structured data, thereby empowering large-scale clinical research.