Deep Learning for Generating Computational PIN-4 Immunohistochemistry Staining from Prostate Biopsy H&E Images
Researchers have developed a deep learning model capable of generating immunohistochemistry (IHC) staining patterns from standard hematoxylin and eosin (H&E) images of prostate biopsies. This method uses a conditional generative adversarial network (cGAN) trained on a dataset of paired H&E and PIN-4 IHC images. The generated images accurately capture diagnostically relevant staining patterns, addressing the current limitation of spatial misalignment between H&E morphology and IHC signals. AI
IMPACT Enables direct interpretation of predicted IHC markers alongside H&E morphology, improving diagnostic accuracy in prostate cancer biopsies.