Adaptive Domain Generalization for Digital Pathology Images

TL;DR

This paper introduces reactive domain generalization methods for digital pathology images, enabling models to adapt to unseen domain shifts at test time without needing prior annotations or explicit shift predictions.

Contribution

It presents a review of existing techniques and investigates test-time training as a novel approach for adaptive domain generalization in histopathology.

Findings

Test-time training improves model robustness to unseen domain shifts.

Reactive adaptation reduces the need for annotated data in new domains.

The approach enhances generalization without prior knowledge of domain variations.

Abstract

In AI-based histopathology, domain shifts are common and well-studied. However, this research focuses on stain and scanner variations, which do not show the full picture -- shifts may be combinations of other shifts, or "invisible" shifts that are not obvious but still damage performance of machine learning models. Furthermore, it is important for models to generalize to these shifts without expensive or scarce annotations, especially in the histopathology space and if wanting to deploy models on a larger scale. Thus, there is a need for "reactive" domain generalization techniques: ones that adapt to domain shifts at test-time rather than requiring predictions of or examples of the shifts at training time. We conduct a literature review and introduce techniques that react to domain shifts rather than requiring a prediction of them in advance. We investigate test time training, a technique for domain generalization that adapts model parameters at test-time through optimization of a secondary self-supervised task.