Table 1.

Disambiguation of Pathological Application Scenarios

Application	Setting	Source Domain	Target Domain
Stain normalization	I2I	image data showing (high) variability	image data with lower variability (potentially subset of source domain); unchanged underlying tissue characteristics
Stain and domain adaptation	I2I	image data acquired with a specific imaging setting (staining, scanner)	imaging setting different from source domain (mainly different staining); unchanged underlying tissue characteristics
Segmentation with supervised models	I2I	image data	label masks corresponding to the input image data
Synthesis enabling weakly supervised and unsupervised learning	I2I	label masks (standard case)	image data corresponding to the label masks
		image data	label masks corresponding to the image data
	I2I and I2Z	label masks	image data corresponding to the label masks and image representations
Data generation and augmentation classification	Z2I	latent vector	image data
	I2L	image data	classification labels