Association of the Pathomics-Collagen Signature with Lymph Node Metastasis in Colorectal Cancer

INTRODUCTION: Lymph node status is a prognostic marker and strongly influences therapeutic decisions in colorectal cancer (CRC). However, traditional pathological images cannot analyze collagen in the tumor microenvironment. This study aims to combine pathological and collagen information to develop and validate a pathomics-collagen signature (PCS) for predicting CRC lymph node metastasis (LNM).

METHODS: This retrospective study included 3 independent cohorts of 973 CRC patients. Fully quantitative pathomics and collagen features were extracted from hematoxylin and eosin slides and multiphoton images of the specimen, respectively. Lasso regression was applied to construct the PCS and integrated with clinicopathological risk factors to develop a nomogram for LNM prediction. Performance of the nomogram was assessed through calibration, discrimination, and clinical usefulness.

RESULTS: The PCS, which integrated 11 pathomics and 9 collagen features, was significantly associated with LNM in the training, internal, and external validation cohorts (p < 0.001). The nomogram based on PCS yielded satisfactory discrimination and calibration, with an AUC of 0.939, 0.895, and 0.893 in the 3 cohorts. Decision curve analysis demonstrated that the PCS-nomogram was clinically useful. Moreover, PCS was an independent predictor of LNM at No.1, No. 2, and No. 3 stations. The PCS-nomograms demonstrated good prediction performances, with AUCs of 0.849-0.939 in the training cohort, 0.837-0.902 in the internal validation cohort, and 0.851-0.895 in the external validation cohort for 3 nodal stations.

CONCLUSION: The PCS based on pathomics and collagen features is associated with LNM, and the PCS-nomogram integrating the PCS and clinicopathological risk factors is useful for individualized prediction of LNM in CRC patients.