RestainNet: a self-supervised digital re-stainer for stain normalization

Abstract

Color inconsistency is an inevitable challenge in computational pathology, which generally happens because of stain intensity variations or sections scanned by different scanners. It harms the pathological image analysis methods, especially the learning-based models. A series of approaches have been proposed for stain normalization. However, most of them are lack of flexibility in practice. In this paper, we formulated stain normalization as a digital re-staining process and proposed a self-supervised learning model, which is called RestainNet. Our network is regarded as a digital re-stainer which learns how to re-stain an unstained (grayscale) image. Two digital stains, Hematoxylin (H) and Eosin (E), were extracted from the original image by Beer-Lambert's Law. We proposed a staining loss to maintain the correctness of stain intensity during the restaining process. Thanks to the self-supervised nature, paired training samples are no longer necessary, which demonstrates great flexibility in practical usage. Our RestainNet outperforms existing approaches and achieves state-of-the-art performance with regard to color correctness and structure preservation. We further conducted experiments on the segmentation and classification tasks and the proposed RestainNet achieved outstanding performance compared with SOTA methods. The self-supervised design allows the network learn any staining style with no extra effort.

A Standardized Pipeline for Colon Nuclei Identification and Counting Challenge

“…Since this challenge does not allow to use other public pathology datasets. We introduced our previously proposed self-supervised stain normalization approach RestainNet [10]. We used the CoNSeP dataset, which is a part of Lizard dataset, to train our stain normalization model.…”

“…First, we applied a GAN-based model [9] to generate paired pseudo masks and images to extend the training set. Next, we introduced a self-supervised stain normalization model [10] to make the color style consistent. Finally, we applied a strong baseline model HoVer-Net [4] as the backbone with an additional costsensitive loss [11] to tackle the class imbalanced problem.…”