Virtual staining model allows transforming autofluorescence images into H&E images that constitutes the gold standard in histology. Then, virtually stained images can be directly diagnosed by pathologists or automatically analysed by the H&E classification algorithm allowing real-time in-situ optical biopsy.

To this end, we followed a Domain transfer approach to map existing images from the novel imaging multi photon domains (Ds) into the target haematoxylin-eosin microscopy domain at pixel level (Dt). We generated a double Siamese network that simultaneously enforces that the generated transformed images both from the source and target domain can be transformed back to their original domain in a pixel-wise manner enforcing the invertibility We also imposed not only pixel-wise similarity on the reconstruction but also higher-level similarity by integrating image embedding vector reconstruction.

The details of this output will be openly accessible through publications in scientific journals at the scientific community´s disposal.

CONTACT

Artzai Picón

Tecnalia Research & Innovation (TECNALIA)

www.tecnalia.com

artzai.picon@tecnalia.com