Intraoperative biopsy imaging of lung cancer risk

Before surgical resection of lung tumor, intraoperative biopsy is needed for cancer diagnosis, while current techniques that guide biopsy have limited performance in tumor identification and boundary determination. Remodeling of extracellular matrix (ECM), mainly collagen and elastin fibers, is an emerging hallmark of tumorigenesis.

Herein, we establish a quantitative multiphoton microscopy (MPM) imaging method for time-efficient, highly-sensitive lung cancer detection via characterization of ECM remodeling. From label-free images of collagen and elastin fibers acquired simultaneously, we construct a similarity coefficient (SC) metric to describe their interaction, and further develop an artificial intelligence (AI)-ECM framework by producing a fiber voxel dictionary via unsupervised learning of morpho-structural features for explainable and visible assessments of cancer risk.

The application is demonstrated by ex vivo human lung cancer diagnosis with a sensitivity of 99.37%, and recognizing the tumor boundary. The translational potential is further revealed via in vivo imaging of a murine model harboring human lung cancer.

This technology can help surgeons perform more precise biopsies and surgeries by providing explainable visual cues, thus leading to better outcomes for lung cancer patients.