Neural Stain-Style Transfer Learning using GAN for Histopathological Images

TL;DR

This paper introduces a GAN-based stain-style transfer model for histopathological images that preserves tumor features and improves classification consistency across different stain styles.

Contribution

The proposed SST model uniquely combines feature-preserving loss with GANs to transfer stain styles while maintaining histopathological patterns.

Findings

Effective stain-style transfer demonstrated on CAMELYON16 dataset.

Improved tumor classification stability across different stain styles.

Preservation of histopathological features during style transfer.

Abstract

Performance of data-driven network for tumor classification varies with stain-style of histopathological images. This article proposes the stain-style transfer (SST) model based on conditional generative adversarial networks (GANs) which is to learn not only the certain color distribution but also the corresponding histopathological pattern. Our model considers feature-preserving loss in addition to well-known GAN loss. Consequently our model does not only transfers initial stain-styles to the desired one but also prevent the degradation of tumor classifier on transferred images. The model is examined using the CAMELYON16 dataset.