Abstract
The identification of tumor capsular invasion as a sign of malignancy is currently employed in traditional histopathology routines for thyroid nodules. However, its limitations are associated with the assessment criteria for invasion, which often lead to disagreements among observers. The aim of this paper is to introduce a widefield imaging technique combined with quantitative collagen analysis to identify areas of capsular invasion in thyroid neoplasms. In this study, we introduce the application of widefield polarization-resolved second harmonic generation microscopy for imaging entire thyroid nodule sections on histological slides. We employ a cylindrical collagen model to extract parameters associated with the ultrastructure and orientation of collagen within the entire capsule of the thyroid nodule. We showcase the effectiveness of these parameters in distinguishing between areas of nodule capsule invasion and unaffected regions of the capsule through statistical analysis of individual parameters and employing a machine learning technique that involves generating maps via cluster analysis. Our results suggest that quantitative analysis facilitated by polarization-resolved widefield second harmonic generation microscopy could prove beneficial for the automated evaluation of capsular invasion sites in thyroid pathology.