Using GANs to transform the domain of large microscopic porcine liver images obtained from native samples into a different domain that represents decellularized liver tissue.

By erasing this domain gap, we can overcome some of the constraints of liver tissue analysis and unlock new possibilities for comprehensive image-based studies, for example, using proven analysis tools from the scaffan library in the desired domain and comparing the results of given selected metric with comparable data. This allows further higher quality evaluation of the sample.

This is a repository for a master's thesis work, developed by Václav Javorek under the supervision of Miroslav Jiřík.

AtoB = Decellularized to native domain transfer

BtoA = Native to decellularized domain transfer

Example colab - transformation inference

Example colab - splitting, processing and merging back a whole slide (.czi) image

AtoB

BtoA

AtoB + post-processing

BtoA + post-processing