Assessing preterm risk via label-free, multiparametric imaging of collagen fiber remodeling in the cervix

Preterm birth (PTB) continues to pose a critical global health challenge. Current clinical assessments of PTB risk are invasive or lack sensitivity, leading to PTB identification at a relatively late stage. Herein, we developed an automated diagnostic framework to identify PTB risk via label-free imaging of collagen fibers integrated with quantitative characterization. Specifically, we obtained structural features (alignment disorder, curvature, and local density) through second-harmonic generation imaging and biochemical features (collagen crosslinking properties) through two-photon excitation fluorescence imaging in normal and preterm murine cervical tissues at distinct time points throughout gestation. Earlier turning points in remodeling dynamics were observed for PTB mice. Further, we established a preterm risk index (PRI) by integrating complementary insights from different features using a linear support vector machine model, and achieved automatic preterm diagnosis with an accuracy higher than 90%. Notably, PRI was able to identify PTB as early as day 3 of mouse gestation. In addition, a high correlation between collagen fiber structure and crosslinking during pregnancy was observed, with these findings consistent with potential cervical remodeling and biochemical adjustment before birth.