EMory BrEast imaging Dataset (EMBED)
收藏资源简介:
EMory BrEast imaging Dataset (EMBED)是由埃默里大学创建的一个大型、细粒度且种族多样的乳腺影像数据集,包含365万次筛查和诊断性乳腺X光片,涉及11.6万名女性,其中白人和非洲裔美国患者各占一半。数据集还包含4万个注释病变,与结构化影像描述符和61个基于严重程度分类的病理结果相关联。创建过程涉及使用自动化和半自动化技术,确保数据的高质量和一致性。该数据集旨在支持开发和验证公平服务于所有患者的乳腺AI模型,减少医疗AI中的偏见,特别关注改善非洲裔美国女性的乳腺影像研究代表性。
The EMory BrEast imaging Dataset (EMBED) is a large-scale, fine-grained, and racially diverse breast imaging dataset developed by Emory University. It comprises 3.65 million screening and diagnostic mammography examinations across 116,000 female patients, with half of the cohort identifying as White and the other half as African American. Additionally, the dataset contains 40,000 annotated lesions, which are associated with structured imaging descriptors and 61 severity-stratified pathological findings. Its development utilized automated and semi-automated techniques to ensure high data quality and consistency. This dataset aims to support the development and validation of breast AI models that equitably serve all patients, mitigate bias in medical AI, with a particular focus on enhancing the representation of African American women in breast imaging research.

- 1The EMory BrEast imaging Dataset (EMBED): A Racially Diverse, Granular Dataset of 3.5M Screening and Diagnostic Mammograms埃默里大学 · 2022年



