stAIns

2024-2025

Computational pathology has been transforming the traditional pathology practice by integrating AI and WSI analysis techniques. Specifically regarding the assessment of different tissue dyes, virtual staining is a promising approach to simulate histological staining on WSI. It leverages advanced image processing and machine learning models to digitally replicate conventional histological stains, such as IHC.

By virtually mimicking the staining process, this technique generates multiple digital slides from a single H&E sample, promoting the efficient use of resources, preservation of tissue specimens, as well as reduction of lab workload and costs. Additionally, virtual staining reduces the turnaround time for diagnoses, enabling faster and more accurate assessments of patient cases. Moreover, virtual staining allows for the customisation of staining patterns, tailoring the visualisation to emphasise specific tissue features and potentially uncovering hidden details that might be otherwise challenging to visualise with conventional staining. Therefore, this approach has the potential to revolutionize how pathologists analyze tissue samples. It is expected to play an increasingly significant role in research, diagnostics, and personalized medicine, ultimately enhancing patient care and improving healthcare outcomes.

We aim to leverage publicly available datasets and in-house data to train a deep generative model capable of synthesizing an IHC sample (stained for HER2) from the standard H&E WSI.

This project is funded by the Dutch Research Council (NWO), through the NGF AiNed XS Europe program.