Short-pulsed AI-augmented third harmonic generation imaging visualizes negative-contrast cell nuclei in fresh mouse liver tissue without optical-induced damage.

Third harmonic generation (THG) microscopy has been widely applied to image various biological tissues, because it reveals cellular structures in a label-free manner. THG also shows great potential in clinical use for rapid intraoperative tumor assessment. However, whether THG imaging can provide cellular contrast in fresh liver tissue remains controversial. In this study, we present the detailed implementation of a home-built higher harmonic generation (HHG) imaging system, using THG as the major cellular contrast modality. By using ultra-short laser pulses (43 fs, 40.8 MHz repetition), we demonstrated for the first time that hepatocyte nuclei in fresh mouse liver tissue appear as negative contrast, surrounded by a brighter cytoplasmic background. This contrast mechanism is different from previous studies using longer laser pulses, fixed and frozen samples. We further investigated the potential tissue damage induced by the short laser pulses. We found that fresh liver tissue can only tolerate a laser power at ∼ 20 mW, without showing obvious accumulated photothermal damage, while at least 40 mW is needed to generate decent THG signals of negative-contrast nuclei of hepatocytes. To avoid this issue, we then developed a deep learning framework that maps THG imaging data acquired with 20 mW laser power to the 40 mW equivalents, allowing visualization of hepatocyte nuclei having negative THG contrast, without optical-induced tissue damage. Collectively, our findings reveal a contrast mechanism of THG in liver tissue that has not been reported previously, which will open new applications for THG and HHG imaging in liver tumor diagnosis.