Multiphoton imaging-based quantifiable collagen signatures for predicting outcomes in patients with pancreatic ductal adenocarcinoma

Pancreatic ductal adenocarcinoma (PDAC) presents a clinical challenge due to its poor prognosis and high mortality rate. Here, we aimed to enhance the prognostic prediction of patients with PDAC by studying collagen features in tumor microenvironment using multiphoton microscopy (MPM) combining with image processing technique. We identified eight distinct tumor-associated collagen signatures (TACS1-8) from multiphoton images of PDAC tissues and developed an optical biomarker, TACS-score, based on the TACS1-8 using ridge regression analysis. Additionally, we also extracted 142 microscopic TACS (M-TACS) from second-harmonic generation (SHG) images and constructed a new robust biomarker, M-TACS-score, using the least absolute shrinkage and selection operator (LASSO) regression analysis. Our statistical results demonstrate that as two new optical biomarkers, TACS- and M-TACS-score, are independent prognostic factors and have good discriminatory ability (high AUC) as well as risk stratification (high HR) comparing with traditional clinical model (combining seven clinical risk factors, age, sex, TNM stage, tumor location and differentiation, perineural and lymph-vascular invasion) in predicting overall survival (OS) of patients with PDAC, highlighting their potential prognostic and predictive value. A combination of label-free multiphoton imaging technique and computer-aided image processing method may offer a novel and promising approach for finding new biomarkers to improve prognosis prediction and thereby tailor treatment strategies more effectively.