BUS-CoT
收藏资源简介:
BUS-CoT数据集是一个包含10,897张乳腺超声图像的数据集,对应10,019个病变和4,838名患者,涵盖了所有99种组织病理学类型。该数据集旨在促进基于观察、特征、诊断和病理标签的链式思维(CoT)推理分析。数据集包含B模式超声、多普勒超声和弹性成像记录,数据标签包括病变特征、超声报告、BI-RADS评分和组织病理学类别。数据集由经验丰富的超声专家进行注释和验证。此外,为了提高鲁棒性,数据集还提供了18种不同设备类型的增强版本。该数据集的创建旨在解决乳腺超声图像分析的挑战,特别是AI系统在罕见情况下的泛化问题。
The BUS-CoT dataset is a collection of 10,897 breast ultrasound images, corresponding to 10,019 lesions and 4,838 patients, covering all 99 histopathological types. This dataset aims to facilitate chain-of-thought (CoT) reasoning analysis based on observations, features, diagnoses and pathological labels. It includes B-mode ultrasound, Doppler ultrasound and elastography records, with data labels encompassing lesion features, ultrasound reports, BI-RADS scores and histopathological categories. The dataset was annotated and validated by experienced ultrasound specialists. Furthermore, to enhance robustness, an augmented version of the dataset across 18 different device types is provided. This dataset was developed to address the challenges in breast ultrasound image analysis, especially the generalization issue of AI systems in rare clinical scenarios.

- 1通过北京大学, 北京协和医学院医院, 北京大学肿瘤医院, 中国医学科学院肿瘤医院/国家癌症中心/国家临床研究中心/肿瘤医院, 中国医科大学附属第一医院, 深圳市妇幼保健院, 西安交通大学, 一准医疗AI有限公司 · 2025年



