Federated attention consistent learning models for prostate cancer diagnosis and Gleason grading

作者全名：Kong, Fei; Wang, Xiyue; Xiang, Jinxi; Yang, Sen; Wang, Xinran; Yue, Meng; Zhang, Jun; Zhao, Junhan; Han, Xiao; Dong, Yuhan; Zhu, Biyue; Wang, Fang; Liu, Yueping

作者地址：[Kong, Fei; Dong, Yuhan] Tsinghua Univ, Shenzhen Int Grad Sch, Shenzhen 518055, Peoples R China; [Wang, Xiyue] Sichuan Univ, Coll Biomed Engn, Chengdu 610065, Peoples R China; [Xiang, Jinxi; Yang, Sen; Zhang, Jun; Han, Xiao] Tencent AI Lab, Shenzhen 518057, Peoples R China; [Wang, Xinran; Yue, Meng; Liu, Yueping] Hebei Med Univ, Hosp 4, Dept Pathol, Shijiazhuang, Peoples R China; [Zhao, Junhan] Massachusetts Gen Hosp, Boston, MA 02114 USA; [Zhao, Junhan] Harvard TH Chan Sch Publ Hlth, Boston, MA 02115 USA; [Zhao, Junhan] Harvard Med Sch, Dept Biomed Informat, Boston, MA 02115 USA; [Zhu, Biyue] Chongqing Med Univ, Dept Pharm, Childrens Hosp, Chongqing 400014, Peoples R China; [Wang, Fang] Qingdao Univ, Affiliated Yantai Yuhuangding Hosp, Dept Pathol, Yantai 264000, Peoples R China

通信作者：Zhang, J (通讯作者)，Tencent AI Lab, Shenzhen 518057, Peoples R China.; Liu, YP (通讯作者)，Hebei Med Univ, Hosp 4, Dept Pathol, Shijiazhuang, Peoples R China.; Wang, F (通讯作者)，Qingdao Univ, Affiliated Yantai Yuhuangding Hosp, Dept Pathol, Yantai 264000, Peoples R China.

来源：COMPUTATIONAL AND STRUCTURAL BIOTECHNOLOGY JOURNAL

ESI学科分类：BIOLOGY & BIOCHEMISTRY

WOS号：WOS:001227271500001

JCR分区：Q2

影响因子：4.4

年份：2024

卷号：23

期号：　

开始页：1439

结束页：1449

文献类型：Article

关键词：Histopathology; Cancer detection; Federated learning; Consistent learning; Attention mechanism

摘要：Artificial intelligence (AI) holds significant promise in transforming medical imaging, enhancing diagnostics, and refining treatment strategies. However, the reliance on extensive multicenter datasets for training AI models poses challenges due to privacy concerns. Federated learning provides a solution by facilitating collaborative model training across multiple centers without sharing raw data. This study introduces a federated attentionconsistent learning (FACL) framework to address challenges associated with large-scale pathological images and data heterogeneity. FACL enhances model generalization by maximizing attention consistency between local clients and the server model. To ensure privacy and validate robustness, we incorporated differential privacy by introducing noise during parameter transfer. We assessed the effectiveness of FACL in cancer diagnosis and Gleason grading tasks using 19,461 whole -slide images of prostate cancer from multiple centers. In the diagnosis task, FACL achieved an area under the curve (AUC) of 0.9718, outperforming seven centers with an average AUC of 0.9499 when categories are relatively balanced. For the Gleason grading task, FACL attained a Kappa score of 0.8463, surpassing the average Kappa score of 0.7379 from six centers. In conclusion, FACL offers a robust, accurate, and cost-effective AI training model for prostate cancer pathology while maintaining effective data safeguards.

基金机构：Beijing Jingjian Foundation for The Advancement of Pathology [JJlXA2022-010]; Science, Technology and Innovation Commission of Shenzhen Municipality [WDZC20200818121348001]

基金资助正文：This research is supported in part by Beijing Jingjian Foundation for The Advancement of Pathology (No. JJlXA2022-010) Science, Technology and Innovation Commission of Shenzhen Municipality (No. WDZC20200818121348001).