Significant efforts are being made to reduce histology turnaround times. Total-Absorption Photoacoustic Remote Sensing (TA-PARS) is the first independent all-optical, label-free optical microscope to provide radiative and non-radiative absorption and optical scattering in a single acquisition. Such array of contrasts enables TA-PARS to rapidly capture most diagnostic elements. Here, a deep learning model, Pix2Pix, is trained within an end-to-end virtual staining framework, utilizing such contrasts. Virtually stained thin and fresh tissue exhibit high concordance when compared against histochemical staining. The proposed work paves the way for developing TA-PARS slide-free histology, which may revolutionize intraoperative microscopic diagnosis and margin assessment.
|