“…This deep learning-based virtual staining technique has been extensively explored by multiple research groups and successfully applied to generate a range of histological stains, such as H&E 12,[19][20][21][22][23][24][25][26][27][28][29][30][31] , Masson's trichrome (MT) staining 12,20,22 and immunohistochemical (IHC) staining 32,33 . These previous works utilized images from various label-free microscopy modalities, including autofluorescence microscopy 12,22,[25][26][27]32 , quantitative phase imaging 20,34 , photoacoustic microscopy 29,31,35 and reflectance confocal microscopy 28 , among others 19,21,23,24,30,33,[36][37][38] . However, these earlier studies have primarily focused on standard biopsy samples, and there has been no virtual staining study on autopsy samples and other large specimens, which often demonstrate suboptimal staining quality with traditional histochemical approaches due to delayed fixation and autolysis.…”