Unsupervised determination of lung tumor margin with widefield polarimetric second-harmonic generation microscopy

Abstract

The extracellular matrix (ECM) is amongst many tissue components affected by cancer, however, morphological changes of the ECM are not well-understood and thus, often omitted from diagnostic considerations. Polarimetric second-harmonic generation (P-SHG) microscopy allows for visualization and characterization of collagen ultrastructure in the ECM, aiding in better understanding of the changes induced by cancer throughout the tissue. In this paper, a large region of hematoxylin and eosin-stained human lung section, encompassing a tumor margin, connecting a significant tumor portion to normal tissue was imaged with P-SHG microscopy. The resulting polarimetric parameters were utilized in principal components analysis and unsupervised k-means clustering to separate normal- and tumor-like tissue. Consequently, a pseudo-color map of the clustered tissue regions is generated to highlight the irregularity of the ECM collagen structure throughout the region of interest and to identify the tumor margin.