[论文解读] WSSS4LUAD: Grand Challenge on Weakly-supervised Tissue Semantic Segmentation for Lung Adenocarcinoma

Chu Han, Xipeng Pan+41 more|arXiv (Cornell University)|Apr 13, 2022

AI in cancer detection被引用 27

一句话总结

本文提出了一个关于弱监督组织语义分割在LUAD上的重大挑战和数据集，使用斑块级标签来推断像素级组织图，并获得来自顶尖团队的最先进结果。

ABSTRACT

Lung cancer is the leading cause of cancer death worldwide, and adenocarcinoma (LUAD) is the most common subtype. Exploiting the potential value of the histopathology images can promote precision medicine in oncology. Tissue segmentation is the basic upstream task of histopathology image analysis. Existing deep learning models have achieved superior segmentation performance but require sufficient pixel-level annotations, which is time-consuming and expensive. To enrich the label resources of LUAD and to alleviate the annotation efforts, we organize this challenge WSSS4LUAD to call for the outstanding weakly-supervised semantic segmentation (WSSS) techniques for histopathology images of LUAD. Participants have to design the algorithm to segment tumor epithelial, tumor-associated stroma and normal tissue with only patch-level labels. This challenge includes 10,091 patch-level annotations (the training set) and over 130 million labeled pixels (the validation and test sets), from 87 WSIs (67 from GDPH, 20 from TCGA). All the labels were generated by a pathologist-in-the-loop pipeline with the help of AI models and checked by the label review board. Among 532 registrations, 28 teams submitted the results in the test phase with over 1,000 submissions. Finally, the first place team achieved mIoU of 0.8413 (tumor: 0.8389, stroma: 0.7931, normal: 0.8919). According to the technical reports of the top-tier teams, CAM is still the most popular approach in WSSS. Cutmix data augmentation has been widely adopted to generate more reliable samples. With the success of this challenge, we believe that WSSS approaches with patch-level annotations can be a complement to the traditional pixel annotations while reducing the annotation efforts. The entire dataset has been released to encourage more researches on computational pathology in LUAD and more novel WSSS techniques.

研究动机与目标

• 丰富LUAD组织病理公开标注资源，以降低标注负担。
• 促进在WSIs上使用斑块级标签的弱监督语义分割技术。
• 开发并评估从LUAD粗略标注生成像素级组织图的策略。

提出的方法

• 组织一次使用斑块级标签对三个组织类别进行分割的WSSS竞赛：肿瘤上皮、肿瘤相关间质和正常组织。
• 采用病理学家参与的标注工作流来创建斑块级和像素级标注。
• 使用基于CAM的弱监督结合基于MLPS的弱监督分割方法来生成伪掩模。
• 结合数据增强和图像拼接以扩展分割的训练信号。
• 使用mIoU评估，排除白色背景，在GDPH和TCGA的WSIs上。

实验结果

研究问题

• RQ1 斑块级标签能否在LUAD中有效转换为准确的像素级组织分割图？
• RQ2 在LUAD组织分割中，最好的弱监督策略（CAM、伪掩模、数据增强）是什么？
• RQ3 病理学家参与的标注流程对标注效率和分割性能有何影响？
• RQ4 在弱监督下，LUAD的三种组织类别（肿瘤、间质、正常）之间的性能差距有多大？

主要发现

团队名称	mIoU	肿瘤	正常	基质
ChunhuiLin	0.8413	0.8389	0.8919	0.7931
baseline0412	0.8222	0.8402	0.8343	0.7921
Vison307	0.8058	0.8165	0.8554	0.7456
BinghongWu	0.8057	0.8045	0.8654	0.7471
adbertyoungdalu	0.8025	0.7967	0.8668	0.7440
DPPD	0.7815	0.7895	0.8397	0.7153
chenxl	0.7714	0.7897	0.8159	0.7088
sibet0222	0.7609	0.8121	0.7107	0.7599
guoxutao	0.7552	0.8179	0.6840	0.7636
shichuanyexi	0.7411	0.8192	0.6714	0.7325
zyw19990916	0.7382	0.8080	0.6868	0.7196
York	0.7239	0.8023	0.6710	0.6985
[e-mail]	0.7199	0.7557	0.7248	0.6791
Zlin3000	0.7064	0.7493	0.6863	0.6837

• 顶级团队的mIoU达到0.8413（肿瘤0.8389，间质0.7931，正常0.8919）。
• CAM基方法在该任务的弱监督语义分割中仍然流行，顶级团队采用逐步 dropout 和多层伪监督。
• 如CutMix和图像拼接等数据增强技术被广泛采用，以产生更可靠的像素级监督。
• 病理学家参与的工作流在通过选择性验证和重叠检查来加速标注的同时仍能保持质量。
• 数据集包含10,091个训练斑块和在验证+测试中的超过1.3亿个标注像素，覆盖GDPH和TCGA的87张WSI。