Ensuring Accurate Stain Reproduction in Deep Generative Networks for Virtual Immunohistochemistry
Immunohistochemistry is a valuable diagnostic tool for cancer pathology. However, it requires specialist labs and equipment, is time-intensive, and is difficult to reproduce. Therefore, a method of virtualising IHC has been a long-term goal. Generative Adversarial Networks have shown promise at inferring immunostains from haematoxylin and eosin. However, they have a substantial weakness in this domain as they can fabricate structures that are not present in the original tissue. CycleGANs can mitigate invented structures but have a related disposition to generate areas of diffuse and inaccurate staining. In this talk, we shall describe a custom loss function to improve the CycleGAN mapping ability for pathology images by enforcing realistic stain replication while retaining tissue structure.
