Label-free detection of visceral pleural invasion in lung adenocarcinoma using multiphoton microscopy.

e20073 Background: Visceral pleural invasion (VPI) constitutes a significant pathological factor in non-small cell lung cancer (NSCLC). The rapid and accurate intraoperative assessment of VPI might be important for determining the appropriate extent of surgical intervention and may contribute to improved disease-free and overall survival rates. However, current intraoperative diagnostic methods for VPI lack reliability. Multiphoton microscopy (MPM) offers promise due to its label-free imaging capabilities and potential for in vivo applications. Therefore, this study aims to investigate the potential of label-free MPM for imaging lung cancer tissues, pleura, and the status of VPI. Methods: This study collected 82 paraffin-embedded lung adenocarcinoma tissue samples. Each encompassing both tumor regions and adjacent pleural areas for comprehensive analysis. Each specimen was cut into two serial 10-μm-thick slices. One slice was used for label-free MPM imaging, while the adjacent slice was stained with Elastica van Gieson (EVG). A lung cancer pathologist marked the region of interest (ROI) containing tumor and pleural tissues in the EVG-stained images. Then, label-free images that collected TPEF and SHG signals were obtained for all marked ROIs accordingly. Two additional surgeons and pathologists then assessed the position of the elastic layer and the tumor invasion in the MPM images. Results: Our study demonstrated that TPEF signals exhibit significant specificity for collagen fibers, facilitating the identification of pleural locations. The figures present MPM images of lung tissues alongside their corresponding EVG-stained images. TPEF signals from elastin (color-coded red) were detected, revealing details of the pleural structure, with the dense, curved elastin architecture (indicated by yellow region) easily to recognize. MPM images also depicted lung adenocarcinoma cells featuring bright cytoplasm, based on the TPEF signals of cellular NADH and FAD, and dark nuclei because of their non-fluorescence (indicated by pink region). Compared to EVG-stained images, MPM exhibited comparable ability to identify elastic layer in the lung and pleural tissues. MPM facilitated straightforward delineation of the lung structure without extrinsic labeling. Surgeons and pathologists can utilize MPM images to distinguish the position of the elastic layer and the tumor invasion in the MPM images for rapid, stain-free diagnosis of VPI in lung adenocarcinoma with an accuracy rate of 100%. And the decision time is less than 2 minutes per case. The accuracy of the deep learning model in diagnosing VPI by MPM images reached 92.86%. Conclusions: Our findings demonstrate that MPM technology can rapidly and accurately detect VPI in lung cancer. This capability may assist thoracic surgeons in making quick intraoperative decisions regarding the extent of the surgery for patients.