Virtually stained H&E images and nuclei segmentation combining neural networks and spectral phasor analysis

• DOI: 10.1117/12.2673774
• Publication Date: 2023-08-11T20:45:50Z
• Authors: Elena Pagani; Davide Panzeri; Riccardo Scodellaro; Margaux Bouzin; Laura D'Alfonso; Maddalena Collini; Giuseppe G. Chirico; Donato Inverso; Laura Sironi
• Keywords: Autofluorescence

H&E stained sections are the gold standard for disease diagnosis but, unfortunately, staining process is time-consuming and expensive. In an effort to overcome these problems, here, we propose a virtual algorithm, able predict Hematoxylin/Eosin (H&E) image, usually exploited during clinical evaluations, starting from autofluorescence signal of entire liver tissue acquired by confocal microscope. The color texture contents generated virtually images have been analyzed through phasor-based approach detect tumorous segment relevant biological structures (accuracy>90% compared expert manual analysis).