For medical image segmentation, contrastive learning is the dominant practice to improve the quality of visual representations by contrasting semantically similar and dissimilar pairs of samples. This is enabled by the observation that without accessing ground truth label, negative examples with truly dissimilar anatomical features, if sampled, can significantly improve the performance. In reality, however, these samples may come from similar anatomical features and the models may struggle to distinguish the minority tail-class samples, making the tail classes more prone to misclassification, both of which typically lead to model collapse. In this paper, we propose ARCO, a semi-supervised contrastive learning (CL) framework with stratified group sampling theory in medical image segmentation. In particular, we first propose building ARCO through the concept of variance-reduced estimation, and show that certain variance-reduction techniques are particularly beneficial in medical image segmentation tasks with extremely limited labels. Furthermore, we theoretically prove these sampling techniques are universal in variance reduction. Finally, we experimentally validate our approaches on three benchmark datasets with different label settings, and our methods consistently outperform state-of-the-art semi- and fully-supervised methods. Additionally, we augment the CL frameworks with these sampling techniques and demonstrate significant gains over previous methods. We believe our work is an important step towards semi-supervised medical image segmentation by quantifying the limitation of current self-supervision objectives for accomplishing medical image analysis tasks.

翻译：对于医学图像分割，对比学习通过对比语义相似与不相似的样本对来提升视觉表征质量，已成为主流实践。这一方法的可行性源于以下观察：若能在不获取真实标签的情况下采样到真正具有不同解剖特征的负样本，则可显著提升性能。然而在实际应用中，这些样本可能源自相似的解剖结构，且模型难以区分小众尾类样本，导致尾类更易被错误分类——这两种情况通常都会引发模型坍塌。本文提出ARCO——一种基于分层组采样理论的半监督对比学习框架，用于医学图像分割。具体而言，我们首先通过方差缩减估计的概念构建ARCO，并证明某些方差缩减技术在标签极度有限的医学图像分割任务中具有特殊优势。其次，我们从理论上证明这些采样技术具有方差缩减的普适性。最后，我们在三个不同标签配置的基准数据集上进行了实验验证，所提方法始终优于最先进的半监督和全监督方法。此外，我们将这些采样技术融入对比学习框架，相较于先前方法取得了显著性能提升。我们相信，通过量化当前自监督目标函数在医学图像分析任务中的局限性，这项工作为半监督医学图像分割迈出了重要一步。