E2EFP-MIL: End-to-end and high-generalizability weakly supervised deep convolutional network for lung cancer classification from whole slide image

被引：13

作者：

Cao, Lei ^{[1
]}

Wang, Jie ^{[2
]}

Zhang, Yuanyuan ^{[4
]}

Rong, Zhiwei ^{[3
]}

Wang, Meng ^{[1
]}

Wang, Liuying ^{[1
]}

Ji, Jianxin ^{[1
]}

Qian, Youhui ^{[5
]}

Zhang, Liuchao ^{[1
]}

Wu, Hao ^{[5
]}

Song, Jiali ^{[3
]}

Liu, Zheng ^{[6
]}

Wang, Wenjie ^{[1
]}

Li, Shuang ^{[1
]}

Wang, Peiyu ^{[6
]}

Xu, Zhenyi ^{[1
]}

Zhang, Jingyuan ^{[7
]}

Zhao, Liang ^{[1
]}

Wang, Hang ^{[2
]}

Sun, Mengting ^{[2
]}

Huang, Xing ^{[7
]}

Yin, Rong ^{[8
]}

Lu, Yuhong ^{[3
]}

Liu, Ziqian ^{[9
]}

Deng, Kui ^{[10
]}

Wang, Gongwei ^{[4
]}

Qiu, Mantang ^{[6
]}

Li, Kang ^{[1
]}

Wang, Jun ^{[6
]}

Hou, Yan ^{[3
]}

机构：

[1] Harbin Med Univ, Sch Publ Hlth, Dept Biostat, Harbin 150081, Peoples R China

[2] Jiangsu Canc Hosp, Jiangsu Inst Canc Res, Dept Tumor Biobank, Nanjing 210009, Peoples R China

[3] Peking Univ, Sch Publ Hlth, Dept Biostat, Beijing 100191, Peoples R China

[4] Peking Univ Peoples Hosp, Dept Pathol, Beijing 100044, Peoples R China

[5] Shenzhen Univ, Dept Thorac Surg, Affiliated Hosp 1, Shenzhen 518000, Peoples R China

[6] Peking Univ Peoples Hosp, Dept Thorac Surg, Beijing 100044, Peoples R China

[7] Nanjing Med Univ, Jiangsu Canc Hosp, Dept Pathol, Affiliated Canc Hosp, Nanjing 210009, Peoples R China

[8] Jiangsu Canc Hosp, Dept Thorac Surg, Jiangsu Key Lab Mol & Translat Canc Res, Nanjing 210009, Peoples R China

[9] Johnson & Johnson Vis Care Inc, Biostat & SAS Programming, Clin Sci, Jacksonville, FL 32256 USA

[10] Vanderbilt Univ, Vanderbilt Epidemiol Ctr, Dept Med, Div Epidemiol,Med Ctr, Nashville, TN 37232 USA

来源：

MEDICAL IMAGE ANALYSIS | 2023年 / 88卷

基金：

中国国家自然科学基金;

关键词：

Classification; Lung cancer; Weakly supervised learning; Deep learning; Convolutional neural network; Computational pathology; Subtype diagnosis; COMPUTATIONAL PATHOLOGY; SURVIVAL;

D O I：

10.1016/j.media.2023.102837

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

Efficient and accurate distinction of histopathological subtype of lung cancer is quite critical for the indi-vidualized treatment. So far, artificial intelligence techniques have been developed, whose performance yet remained debatable on more heterogenous data, hindering their clinical deployment. Here, we propose an end-to-end, well-generalized and data-efficient weakly supervised deep learning-based method. The method, end-to-end feature pyramid deep multi-instance learning model (E2EFP-MIL), contains an iterative sampling module, a trainable feature pyramid module and a robust feature aggregation module. E2EFP-MIL uses end-to-end learning to extract generalized morphological features automatically and identify discriminative histomorphological patterns. This method is trained with 1007 whole slide images (WSIs) of lung cancer from TCGA, with AUCs of 0.95-0.97 in test sets. We validated E2EFP-MIL in 5 real-world external heterogenous cohorts including nearly 1600 WSIs from both United States and China with AUCs of 0.94-0.97, and found that 100-200 training images are enough to achieve an AUC of >0.9. E2EFP-MIL overperforms multiple state-of-the-art MIL-based methods with high accuracy and low hardware requirements. Excellent and robust results prove generalizability and effectiveness of E2EFP-MIL in clinical practice. Our code is available at https://github.com/raycaohmu/E2EFP-MIL.

引用

页数：11