Hard Negative Sample Mining for Whole Slide Image Classification

TL;DR

This paper introduces a novel hard negative sample mining approach and a patch-wise ranking loss for weakly supervised whole slide image classification, improving feature representation and reducing training costs.

Contribution

It proposes a new method for mining hard negative samples and a ranking loss in MIL, enhancing WSI classification performance.

Findings

Improved classification accuracy on public datasets.

Reduced training time and computational costs.

Enhanced feature representations through hard negative mining.

Abstract

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