A Closer Look at Domain Shift for Deep Learning in Histopathology

TL;DR

This paper investigates how domain shifts affect deep learning models in histopathology, analyzing the impact of data preparation and introducing a new measure to evaluate domain differences in learned representations.

Contribution

It provides a detailed analysis of domain shift effects in histopathology and introduces a novel measure for assessing domain similarity based on model representations.

Findings

Data preparation significantly influences learning outcomes.

Latent representations are sensitive to data distribution changes.

The proposed measure can detect domain shifts affecting model generalization.

Abstract

Domain shift is a significant problem in histopathology. There can be large differences in data characteristics of whole-slide images between medical centers and scanners, making generalization of deep learning to unseen data difficult. To gain a better understanding of the problem, we present a study on convolutional neural networks trained for tumor classification of H&E stained whole-slide images. We analyze how augmentation and normalization strategies affect performance and learned representations, and what features a trained model respond to. Most centrally, we present a novel measure for evaluating the distance between domains in the context of the learned representation of a particular model. This measure can reveal how sensitive a model is to domain variations, and can be used to detect new data that a model will have problems generalizing to. The results show how learning is heavily influenced by the preparation of training data, and that the latent representation used to do classification is sensitive to changes in data distribution, especially when training without augmentation or normalization.