PET2Rep
收藏资源简介:
PET2Rep是首个用于评估视觉语言模型在PET放射学报告生成中的基准数据集。它包含全身PET/CT图像-报告对,覆盖数十个器官的代谢信息,这对于反映PET成像在肿瘤诊断中的真实临床全面性至关重要。
PET2Rep is the first benchmark dataset for evaluating vision-language models in the task of PET radiology report generation. It contains whole-body PET/CT image-report pairs, covering metabolic information of dozens of organs, which is critical for reflecting the real-world clinical comprehensiveness of PET imaging in tumor diagnosis.
PET2Rep 数据集概述
数据集简介
PET2Rep是首个用于评估视觉语言模型(VLMs)在正电子发射断层扫描(PET)放射学报告生成中的基准数据集。该数据集包含全身PET/CT图像-报告对,覆盖数十个器官的代谢信息,对于反映PET成像在肿瘤诊断中的真实临床全面性至关重要。
数据内容
- 数据类型:全身PET/CT图像-报告对
- 覆盖范围:数十个器官的代谢信息
- 应用领域:肿瘤诊断中的PET成像
数据集特点
- 首个专注于PET放射学报告生成的基准数据集
- 包含反映真实临床全面性的图像-报告对
- 专注于视觉语言模型在医疗报告生成中的评估
相关资源
- 论文地址:https://arxiv.org/abs/2508.04062
- 代码仓库:https://github.com/YichiZhang98/PET2Rep
引用信息
bibtex @article{zhang2025pet2rep, title={PET2Rep: Towards Vision-Language Model-Drived Automated Radiology Report Generation for Positron Emission Tomography}, author={Zhang, Yichi and Zhang, Wenbo and Ling, Zehui and Feng, Gang and Peng, Sisi and Chen, Deshu and Liu, Yuchen and Zhang, Hongwei and Wang, Shuqi and Li, Lanlan and others}, journal={arXiv preprint arXiv:2508.04062}, year={2025} }




