Weakly Semi-supervised Whole Slide Image Classification by Two-level Cross Consistency Supervision

TL;DR

This work defines Weakly Semi-supervised WSI Classification (WSWC), addressing the costly labeling of patches within gigapixel WSIs by leveraging both labeled and unlabeled bags. The CroCo framework introduces two heterogeneous classifier branches that share an instance encoder and enforce cross-consistency at both bag- and instance-level, enabling effective learning from limited bag labels. Empirical results on synthetic and real-pathology datasets show CroCo consistently outperforms baselines in bag-level and, often, instance-level tasks, while reducing required labeling by half in practice. The approach advances practical WSI diagnosis by maximizing information from unlabeled data and instance-level cues, though it notes potential improvements via pseudo-label filtering to further enhance robustness.

Abstract

Computer-aided Whole Slide Image (WSI) classification has the potential to enhance the accuracy and efficiency of clinical pathological diagnosis. It is commonly formulated as a Multiple Instance Learning (MIL) problem, where each WSI is treated as a bag and the small patches extracted from the WSI are considered instances within that bag. However, obtaining labels for a large number of bags is a costly and time-consuming process, particularly when utilizing existing WSIs for new classification tasks. This limitation renders most existing WSI classification methods ineffective. To address this issue, we propose a novel WSI classification problem setting, more aligned with clinical practice, termed Weakly Semi-supervised Whole slide image Classification (WSWC). In WSWC, a small number of bags are labeled, while a significant number of bags remain unlabeled. The MIL nature of the WSWC problem, coupled with the absence of patch labels, distinguishes it from typical semi-supervised image classification problems, making existing algorithms for natural images unsuitable for directly solving the WSWC problem. In this paper, we present a concise and efficient framework, named CroCo, to tackle the WSWC problem through two-level Cross Consistency supervision. CroCo comprises two heterogeneous classifier branches capable of performing both instance classification and bag classification. The fundamental idea is to establish cross-consistency supervision at both the bag-level and instance-level between the two branches during training. Extensive experiments conducted on four datasets demonstrate that CroCo achieves superior bag classification and instance classification performance compared to other comparative methods when limited WSIs with bag labels are available. To the best of our knowledge, this paper presents for the first time the WSWC problem and gives a successful resolution.

Figures (3)

• Figure 1: A. The current paradigm for weakly supervised WSI classification assumes the availability of labels for all training bags. B. In our proposed weakly semi-supervised WSI classification paradigm, only a limited number of labeled training bags are available, while there are numerous unlabeled bags. C. The traditional semi-supervised learning paradigm relies on pseudo-labels and consistency in the domain of natural images. D. We introduce CroCo, a concise and efficient framework for bag-level and instance-level Cross Consistency Supervision in WSI classification.
• Figure 2: Pipeline of our Framework CroCo in processing unlabeled bags.
• Figure 3: The AUC curves of CroCo's bag-level pseudo labels (A) and instance-level pseudo labels (B) during training. (C) SVM and linear evaluation of extracted features on the Camelyon16 Dataset.