This paper proposes a deep learning (DL)-based approach for generating virtual Hematoxylin and Eosin (H&E) staining from label-free fluorescence lifetime imaging microscopy (FLIM) images. The method combines an advanced DL model with a contemporary image quality metric to generate clinical-grade virtual H&E-stained images from unstained tissue samples. Experiments show that incorporating lifetime information improves accuracy compared to intensity-only images, enabling rapid and precise interpretation of FLIM images at the cellular level. Distinct lifetime signatures of seven cell types in the tumor microenvironment were identified, suggesting potential for biomarker-free tissue histology using FLIM across multiple cancer types.

Qiang Wang, Ahsan R. Akram, David A. Dorward, Sophie Talas, Basil Monks, Chee Thum, James R. Hopgood, Malihe Javid, Marta Vallejos