Prognostic Value of Digital Pathological Features in Colorectal Cancer

Background: The traditional TNM staging system for colorectal cancer (CRC) is limited by the biological characteristics of tumors and their high heterogeneity. This study proposes a pathological signature of colorectal cancer (PScrc) based on digital pathology technology, aimed at assessing its prognostic value in overall survival (OS) and disease-free survival (DFS) through the analysis of various pathological features in hematoxylin and eosin (HE)-stained slides. Methods: A retrospective cohort analysis was conducted, including 149 patients who underwent surgery for colorectal cancer from January 2000 to December 2012, randomly divided into a training set and a validation set. High-resolution images of all HE stained slides were obtained through digital scanning, and pathological features were extracted using CellProfiler software. The PScrc was constructed using a LASSO-Cox regression model to evaluate its association with OS and DFS, and independent prognostic factors were identified through Cox regression analysis. All statistical analyses were performed using SPSS and R software. Results: A total of 149 colorectal cancer patients were included, with 75 in the training cohort and 74 in the validation cohort. LASSO-Cox regression analysis identified eight key pathological features for the construction of PScrc. Survival analysis demonstrated that patients in the high PScrc group had significantly worse survival outcomes compared to those in the low PScrc group. Univariate and multivariate Cox regression analyses confirmed PScrc and other clinicopathological features as independent prognostic factors. The C-index, AUROC, and decision curve analysis were used to evaluate the clinical value of the nomogram that was generated based on these criteria so as to predict OS and DFS. Additionally, the predictive capability of PScrc for response to adjuvant chemotherapy was analyzed, leading to the optimization of the PScrc_chemo score, which indicated that patients with low PScrc_chemo had poorer survival outcomes following adjuvant chemotherapy. Conclusion: PScrc adds predictive value to the TNM staging system for colorectal cancer by acting as a prognostic predictor for CRC patients.