AMi-Br
收藏资源简介:
AMi-Br数据集是由维也纳兽医大学等多个研究机构合作创建的,专注于人类乳腺癌中的正常和异常有丝分裂图像。该数据集包含3720个有丝分裂图像,其中832个为异常有丝分裂,2888个为正常有丝分裂。数据来源于TUPAC和MIDOG 2021两个公开数据集,并通过三位专家的多数投票进行分类。数据集创建过程中,专家们对每个有丝分裂图像进行了独立分类,最终形成了具有高一致性的标签。该数据集旨在通过深度学习模型提高有丝分裂分类的准确性和可重复性,特别是在乳腺癌预后评估中的应用。
The AMi-Br dataset was collaboratively created by multiple research institutions including the University of Veterinary Medicine Vienna, focusing on normal and abnormal mitotic images in human breast cancer. It contains a total of 3720 mitotic images, among which 832 are abnormal mitoses and 2888 are normal mitoses. The dataset is sourced from two public datasets, TUPAC and MIDOG 2021, and its labels were generated via majority voting by three experts. During the dataset construction process, each mitotic image was independently classified by the experts, ultimately resulting in highly consistent labels. This dataset is intended to enhance the accuracy and reproducibility of mitotic classification using deep learning models, especially for applications in breast cancer prognosis evaluation.

- 1A Histologic Dataset of Normal and Atypical Mitotic Figures on Human Breast Cancer (AMi-Br)维也纳兽医大学, 施瓦茨曼动物医疗中心, 弗伦斯堡应用科学大学, 柏林自由大学, 维也纳医科大学, 因戈尔施塔特技术大学 · 2025年



