Abstract
Stain normalization is crucial in addressing color variability caused by differences in staining protocols and equipment, toward enabling computational methods to learn tissue patterns while eliminating confounding from stain variations. Traditionally, such methods relied on a single reference image, failing to capture variability across slides from different organizations. More recently, reference aggregates from multiple slides are used, notwithstanding their dependency on tedious annotations of distinct histologic regions. Here we introduce the Unsupervised Population-based morphology-aware STAIn Normalization (UPStaiN), to mitigate the need for manual annotations, while still creating a reference standard informed by multiple aggregated slides. UPStaiN gains its morphology awareness in an unsupervised manner via K-means clustering of deep feature embeddings of reduced dimensionality from tissue patches across numerous slides, thereby capturing tissue-specific staining characteristics while accounting for stain variations, and hence enabling robust population-based reference aggregation. We evaluate UPStaiN on 2,176 H&E-stained whole slide images (WSIs) from the Ivy Glioblastoma Atlas Project (IvyGAP) dataset, and compare its results with stain normalization using manually annotated masks of morphologically distinct regions. Quantitative evaluation reveals UPStaiN achieving statistically significant (p-value<0.0157) and consistently lower standard deviation across every morphologically distinct region and across all testing WSIs. Our findings support UPStaiN’s unsupervised mechanism of morphology awareness from population-based aggregated reference slides as an annotation-free solution for stain normalization. Our code is available at https://github.com/IUCompPath/unsupervised-population-based-stain-norm
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/BrainWorks_030.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
Open Review Page: https://openreview.net/profile?id=~Spyridon_Bakas3
BibTex
@InProceedings{AdaSan_UPStaiN_MICCAISAT2026,
author = { Adap, Sanyukta AND Baheti, Bhakti AND Bakas, Spyridon},
title = { { UPStaiN: Unsupervised Population-Based Morphology-Aware Stain Normalization } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17254},
month = {pending},
page = {pending}
}