Autofluorescence Imaging and Virtual Histological Staining of Human Prostate Sections for Cancer Diagnosis

Autofluorescence imaging with virtual histological staining has shown promising potential to improve the efficiency of postoperative histopathology workflow. Here, a widefield deep‐ultraviolet light‐emitting diode‐based imaging system to generate autofluorescence images of human prostate tissue sections from 12 patients with a lateral resolution of ≈1 μm is used. Subsequently, the autofluorescence images are transformed into virtual hematoxylin and eosin‐stained images via a weakly supervised deep learning framework. The virtual staining results are assessed by four professional pathologists through statistical analysis, which shows high diagnostic accuracy (91%), and high tissue detail quality.