Histopathology Image Normalization via Latent Manifold Compaction

TL;DR

This paper introduces Latent Manifold Compaction, an unsupervised method for histopathology image normalization that learns batch-invariant features, improving cross-batch generalization and outperforming existing normalization techniques.

Contribution

The paper proposes a novel unsupervised framework, Latent Manifold Compaction, for histopathology image normalization that effectively generalizes across unseen domains by explicitly compacting stain-induced latent manifolds.

Findings

LMC significantly reduces batch effects in histopathology images.

LMC outperforms state-of-the-art normalization methods in classification and detection tasks.

LMC demonstrates strong generalization to unseen target domains.

Abstract

Batch effects arising from technical variations in histopathology staining protocols, scanners, and acquisition pipelines pose a persistent challenge for computational pathology, hindering cross-batch generalization and limiting reliable deployment of models across clinical sites. In this work, we introduce Latent Manifold Compaction (LMC), an unsupervised representation learning framework that performs image harmonization by learning batch-invariant embeddings from a single source dataset through explicit compaction of stain-induced latent manifolds. This allows LMC to generalize to target domain data unseen during training. Evaluated on three challenging public and in-house benchmarks, LMC substantially reduces batch-induced separations across multiple datasets and consistently outperforms state-of-the-art normalization methods in downstream cross-batch classification and detection tasks, enabling superior generalization.