Multi-Scale Prototypical Transformer for Whole Slide Image Classification

TL;DR

This paper introduces a multi-scale prototypical Transformer that improves whole slide image classification by effectively fusing multi-scale features and reducing redundant instances, outperforming existing methods.

Contribution

The paper presents a novel MSPT model combining prototypical learning with Transformer architecture and multi-scale feature fusion for enhanced WSI classification.

Findings

MSPT outperforms existing algorithms on public datasets.

Prototypical Transformer reduces redundant instances effectively.

Multi-scale feature fusion improves classification accuracy.

Abstract

Whole slide image (WSI) classification is an essential task in computational pathology. Despite the recent advances in multiple instance learning (MIL) for WSI classification, accurate classification of WSIs remains challenging due to the extreme imbalance between the positive and negative instances in bags, and the complicated pre-processing to fuse multi-scale information of WSI. To this end, we propose a novel multi-scale prototypical Transformer (MSPT) for WSI classification, which includes a prototypical Transformer (PT) module and a multi-scale feature fusion module (MFFM). The PT is developed to reduce redundant instances in bags by integrating prototypical learning into the Transformer architecture. It substitutes all instances with cluster prototypes, which are then re-calibrated through the self-attention mechanism of the Trans-former. Thereafter, an MFFM is proposed to fuse the clustered prototypes of different scales, which employs MLP-Mixer to enhance the information communication between prototypes. The experimental results on two public WSI datasets demonstrate that the proposed MSPT outperforms all the compared algorithms, suggesting its potential applications.