From the Department of Diagnostic Medicine (A.D., J.V.), Livestrong Cancer Institutes (J.V.), and Department of Oncology (J.V.), Dell Medical School, University of Texas at Austin, 210 E 24th St, Austin, TX 78712; and Oden Institute for Computational Engineering and Sciences, University of Texas at Austin, Austin, Tex (J.V.).
Radiol Imaging Cancer. 2024 Jan;6(1):e230100. doi: 10.1148/rycan.230100.
Purpose To characterize the demographic distribution of The Cancer Imaging Archive (TCIA) studies and compare them with those of the U.S. cancer population. Materials and Methods In this retrospective study, data from TCIA studies were examined for the inclusion of demographic information. Of 189 studies in TCIA up until April 2023, a total of 83 human cancer studies were found to contain supporting demographic data. The median patient age and the sex, race, and ethnicity proportions of each study were calculated and compared with those of the U.S. cancer population, provided by the Surveillance, Epidemiology, and End Results Program and the Centers for Disease Control and Prevention U.S. Cancer Statistics Data Visualizations Tool. Results The median age of TCIA patients was found to be 6.84 years lower than that of the U.S. cancer population ( = .047) and contained more female than male patients (53% vs 47%). American Indian and Alaska Native, Black or African American, and Hispanic patients were underrepresented in TCIA studies by 47.7%, 35.8%, and 14.7%, respectively, compared with the U.S. cancer population. Conclusion The results demonstrate that the patient demographics of TCIA data sets do not reflect those of the U.S. cancer population, which may decrease the generalizability of artificial intelligence radiology tools developed using these imaging data sets. Ethics, Meta-Analysis, Health Disparities, Cancer Health Disparities, Machine Learning, Artificial Intelligence, Race, Ethnicity, Sex, Age, Bias Published under a CC BY 4.0 license.
描述癌症影像学档案（TCIA）研究的人口统计学分布，并将其与美国癌症人群进行比较。
在这项回顾性研究中，对 TCIA 研究中的数据进行了检查，以纳入人口统计学信息。在截至 2023 年 4 月的 TCIA 中，共有 189 项研究，其中有 83 项人类癌症研究包含支持性人口统计学数据。计算并比较了每项研究的中位患者年龄以及性别、种族和民族比例，与美国癌症人群（由美国癌症研究所和疾病控制与预防中心的监测、流行病学和最终结果计划以及美国癌症统计数据可视化工具提供）进行了比较。
TCIA 患者的中位年龄比美国癌症人群低 6.84 岁（ =.047），女性患者多于男性患者（53%比 47%）。与美国癌症人群相比，TCIA 研究中美洲印第安人和阿拉斯加原住民、黑人和非裔美国人以及西班牙裔患者的比例分别低 47.7%、35.8%和 14.7%。
结果表明，TCIA 数据集的患者人口统计学特征与美国癌症人群不相符，这可能会降低使用这些影像学数据集开发的人工智能放射学工具的普遍性。
这项研究是回顾性的，因此不需要伦理审查。
作者没有报告任何利益冲突。
这项研究发表在 CC BY 4.0 许可下。
