Scholarlink

This paper proposes deep-learning models to assist in classifying subtypes of luminal breast cancer from immunohistochemistry slides. The classification has traditionally been done by pathologists who evaluate Ki67 indexes by manually counting positive and negative cells that are dyed in different colors in the immunohistochemistry slides. This manual process is error-prone and time-consuming. Therefore, this paper explores the possibility of automatically classifying immunohistochemistry slides. The performances of 4 pre-trained, deep convolutional networks are compared. Based on the good performance of the models, we conclude that the task of Ki67 classification for determining subtypes of luminal breast cancer is well-suited for deep learning models and that for this reason, manual counting of Ki67-positive cells are neither needed nor desirable.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Deep Learning Classification of Subtypes of Luminal Breast Cancer from Immunohistochemistry Slides

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Vishnu Shaji,
  • Mridul Sharma,
  • Merin Mathew,
  • K. Malavika,
  • Dhanya Mary Louis,
  • Dehannathparambil Kottarathil Vijaykumar,
  • Lakshmi Malavika Nair,
  • Archana George,
  • Georg Gutjahr

摘要

This paper proposes deep-learning models to assist in classifying subtypes of luminal breast cancer from immunohistochemistry slides. The classification has traditionally been done by pathologists who evaluate Ki67 indexes by manually counting positive and negative cells that are dyed in different colors in the immunohistochemistry slides. This manual process is error-prone and time-consuming. Therefore, this paper explores the possibility of automatically classifying immunohistochemistry slides. The performances of 4 pre-trained, deep convolutional networks are compared. Based on the good performance of the models, we conclude that the task of Ki67 classification for determining subtypes of luminal breast cancer is well-suited for deep learning models and that for this reason, manual counting of Ki67-positive cells are neither needed nor desirable.