A Domain Adaptive and Label-efficient Deep Learning Method for Ki-67 Scoring in Infiltrative Breast Ductal Carcinoma Cells
Yiqing Liu, Hufei Duan, Huijuan Shi, Yuxin Ni, Xin Wang, Anjia Han, Yonghong He · 2022
Purpose Ki-67 proliferation index is closely related to the diagnosis, grading, prognosis, and treatment of breast invasive ductal carcinoma (IDC). However, existing deep learning algorithms for automatic Ki-67 index scoring do not exclude normal cells, which leads to no clinically meaningful results. Considering that, we aim to develop a label-efficient and domain adaptive method to distinguish tumor cells from normal cells in Ki-67-stained images.