Artificial Intelligence Predicts Multiclass Molecular Signatures and Subtypes Directly From Breast Cancer Histology: a Multicenter Retrospective Study

Xiangyang Zhang; Yang Chen; Changjing Cai; Yifeng Wang; Jun Tan; Zijie Fang; Le Wei; Zhuchen Shao; Liwen Wang; Tiezheng Qi; Yihan Liu; Zhaohui Jiang; Yin Li; Ying Han; Tibera Kagemulo Rugambwa; Shan Zeng; Haoqian Wang; Hong Shen; Yongbing Zhang

doi:10.1097/JS9.0000000000002220

Artificial Intelligence Predicts Multiclass Molecular Signatures and Subtypes Directly From Breast Cancer Histology: a Multicenter Retrospective Study

Int J Surg. 2025 Jan 7. doi: 10.1097/JS9.0000000000002220. Online ahead of print.

Authors

Xiangyang Zhang^{1

2}, Yang Chen³, Changjing Cai¹, Yifeng Wang⁴, Jun Tan⁵, Zijie Fang³, Le Wei^{1

6}, Zhuchen Shao³, Liwen Wang⁶, Tiezheng Qi⁶, Yihan Liu^{1

6}, Zhaohui Jiang¹, Yin Li¹, Ying Han¹, Tibera Kagemulo Rugambwa¹, Shan Zeng¹, Haoqian Wang³, Hong Shen^{1

7}, Yongbing Zhang⁸

Affiliations

¹ Department of Oncology, Xiangya Hospital, Central South University, Changsha, Hunan, China.
² Department of Biomedical Sciences, City University of Hong Kong, Hong Kong, China.
³ Tsinghua Shenzhen International Graduate School,Tsinghua University, Shenzhen, China.
⁴ School of Science, Harbin Institute of Technology (Shenzhen), Shenzhen, China.
⁵ Department of Neurosurgery, Xiangya Hospital, Central South University, Changsha, China.
⁶ Xiangya School of Medicine,Central South University, Changsha, China.
⁷ National Clinical Research Center for Geriatric Disorders, Xiangya Hospital, Central South University, Changsha, China.
⁸ Computer Science and Technology, Harbin Institute of Technology (Shenzhen), Shenzhen, China.

PMID: 39764584
DOI: 10.1097/JS9.0000000000002220

Abstract

Detection of biomarkers of breast cancer incurs additional costs and tissue burden. We propose a deep learning-based algorithm (BBMIL) to predict classical biomarkers, immunotherapy-associated gene signatures, and prognosis-associated subtypes directly from hematoxylin and eosin stained histopathology images. BBMIL showed the best performance among comparative algorithms on the prediction of classical biomarkers, immunotherapy related gene signatures, and subtypes.