Researchers at the University of California, Los Angeles (UCLA), along with international collaborators, have developed a deep learning-based virtual multiplexed immunostaining (mIHC) method. This new approach allows for the simultaneous creation of ERG, PanCK, and H&E images from label-free tissue sections. The technology aims to improve the accuracy and efficiency of assessing vascular invasion in thyroid cancer.
Traditional immunohistochemistry (IHC) methods require separate tissue sections for each stain, which can increase costs and labor while risking tissue loss. These conventional techniques also suffer from variability between sections that may affect diagnostic precision. Multiplexed IHC technologies can apply multiple antibodies at once but are complex and not commonly used in routine pathology laboratories.
The research team led by Aydogan Ozcan and Nir Pillar introduced a virtual mIHC framework that uses deep learning algorithms to address these issues. Their technique employs autofluorescence microscopy images of unstained tissue sections to generate virtual stains resembling those produced by standard histochemical staining. The generated stains include ERG for endothelial cells, PanCK for epithelial cells, and H&E for general tissue structure.
The development process involved training the virtual mIHC method on paired datasets of autofluorescence and histochemically stained images from thyroid tissue microarrays. By using conditional generative adversarial networks (cGANs) and a digital staining matrix, the system converted label-free images into virtually stained ones that closely matched traditional results.
Board-certified pathologists conducted blind evaluations of the virtual mIHC staining outcomes. They found strong agreement with traditional methods regarding staining patterns, intensity, and cellular localization. The virtual stains were able to accurately highlight both epithelial and endothelial cells, supporting identification and localization of vascular invasion—a key factor in cancer metastasis.
According to the research team, "The virtual mIHC technique represents a significant advancement in histopathological evaluation, offering a cost-effective, efficient, and accurate alternative to conventional IHC and mIHC methods." They added: "By streamlining the diagnostic workflow and preserving valuable tissue samples, this innovation has the potential to transform clinical practice, improving patient outcomes in thyroid cancer and beyond."
Future work will focus on validating this technology across various tissue types and multi-site cohorts with an aim toward broader clinical adoption.
