STAIN: Smooth Tile-Aware Instance Normalisation for Virtual Staining


David Armstrong (The Queen's University Belfast), Iain B Styles (The Queen's University Belfast), Niall McLaughlin (The Queen's University Belfast)
The 35th British Machine Vision Conference
PDFPosterVideo (Right click to download)Supplementary

Abstract

We propose STAIN, a method for eliminating tiling artefacts when virtual staining models are applied tile-wise to very high resolution whole slide images (WSIs). Cancer diagnosis depends on examination of tissue samples under a microscope. The samples are naturally transparent, so chemical staining agents are applied to enhance their visibility - a costly and time consuming process. While virtual staining models have been developed to simulate the staining process, they must be applied tile-wise due to the very high resolution of scanned whole slide images. The tiling process introduces artefacts, such as colour and contrast differences between tiles and distorted objects at tile boundaries. To address the tiling artefact problem we combine a pixel-wise instance normalisation layer, with an appropriate padding technique, ensuring smooth colour and contrast transitions between tiles and preserving the structure of objects crossing tile boundaries. The resulting WSI is visually consistent and artefact free. Furthermore, we propose a new method for detecting tiling artefacts using gradient information at the seams between tiles. Experiments on three common virtual staining datasets show that STAIN significantly reduces artefacts compared with existing methods.

Citation

@inproceedings{Armstrong_2025_BMVC,
author    = {David Armstrong and Iain B Styles and Niall McLaughlin},
title     = {STAIN: Smooth Tile-Aware Instance Normalisation for Virtual Staining},
booktitle = {36th British Machine Vision Conference 2025, {BMVC} 2025, Sheffield, UK, November 24-27, 2025},
publisher = {BMVA},
year      = {2025},
url       = {https://bmva-archive.org.uk/bmvc/2025/assets/papers/Paper_102/paper.pdf}
}


Copyright © 2025 The British Machine Vision Association and Society for Pattern Recognition
The British Machine Vision Conference is organised by The British Machine Vision Association and Society for Pattern Recognition. The Association is a Company limited by guarantee, No.2543446, and a non-profit-making body, registered in England and Wales as Charity No.1002307 (Registered Office: Dept. of Computer Science, Durham University, South Road, Durham, DH1 3LE, UK).

Imprint | Data Protection