Test-time image-to-image translation ensembling improves out-of-distribution generalization in histopathology
This addresses domain generalization for histopathology algorithms, improving robustness to protocol variations across medical centers, though it is incremental as it builds on existing image-to-image translation and ensembling methods.
The paper tackles the problem of inter-hospital variability in histopathology images harming algorithm generalization by proposing a test-time data augmentation method using multi-domain image-to-image translation, which significantly boosts performance on unseen protocols, as demonstrated by outperforming existing techniques on two histopathology tasks.
Histopathology whole slide images (WSIs) can reveal significant inter-hospital variability such as illumination, color or optical artifacts. These variations, caused by the use of different scanning protocols across medical centers (staining, scanner), can strongly harm algorithms generalization on unseen protocols. This motivates development of new methods to limit such drop of performances. In this paper, to enhance robustness on unseen target protocols, we propose a new test-time data augmentation based on multi domain image-to-image translation. It allows to project images from unseen protocol into each source domain before classifying them and ensembling the predictions. This test-time augmentation method results in a significant boost of performances for domain generalization. To demonstrate its effectiveness, our method has been evaluated on 2 different histopathology tasks where it outperforms conventional domain generalization, standard H&E specific color augmentation/normalization and standard test-time augmentation techniques. Our code is publicly available at https://gitlab.com/vitadx/articles/test-time-i2i-translation-ensembling.