Multiphoton microscopy combined with machine learning shows promise for localizing pancreatic neuroendocrine tumors

Pancreatic neuroendocrine tumors (PNETs) are a relatively rare type of cancer whose preferred method of treatment is surgery, however current intraoperative guidance techniques have poor contrast. Multiphoton microscopy (MPM) is an imaging technique capable of capturing many biomarkers indicative of cancer; this project examines whether MPM images may provide a basis for a robust method of PNET localization. 14 fixed frozen and 57 formalin-fixed paraffin-embedded samples were imaged using MPM and classified using linear discriminant analysis. The model performed well across both sample preparations, indicating our approach could be applied to improve surgical localization of PNETs.