通过挖掘难负样本提升病理图像分类效果，降低训练成本。
Hard Negative Sample Mining for Whole Slide Image Classification
- 细粒度挖掘难负样本，优化特征表示。
- 在两个公开数据集上准确率提升，训练效率更高。
- 适合医学图像分析与弱监督学习研究者参考。
弱监督全切片图像（WSI）分类因缺乏局部标记和高计算成本而具有挑战性。当前先进方法利用自监督的局部特征表示进行多实例学习（MIL）。近期工作尝试通过伪标签对特征表示进行微调，但主要关注高质量正样本的选择。本文提出在微调阶段挖掘难负样本，从而获得更优的特征表示并降低训练成本。此外，我们设计了一种新型的局部排序损失函数，以更好利用这些难负样本。在两个公开数据集上的实验验证了所提方法的有效性。代码已开源：https://github.com/winston52/HNM-WSI。
原文摘要 · Abstract (English)
Weakly supervised whole slide image (WSI) classification is challenging due to the lack of patch-level labels and high computational costs. State-of-the-art methods use self-supervised patch-wise feature representations for multiple instance learning (MIL). Recently, methods have been proposed to fine-tune the feature representation on the downstream task using pseudo labeling, but mostly focusing on selecting high-quality positive patches. In this paper, we propose to mine hard negative samples during fine-tuning. This allows us to obtain better feature representations and reduce the training cost. Furthermore, we propose a novel patch-wise ranking loss in MIL to better exploit these hard negative samples. Experiments on two public datasets demonstrate the efficacy of these proposed ideas. Our codes are available at https://github.com/winston52/HNM-WSI
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。