Radiomics analysis is widely used to assess tumor prognosis.
To explore the value of computed tomography (CT) radiomics nomogram in predicting disease-free survival (DFS) of patients with colorectal cancer (CRC) after operation.
A total of 522 CRC patients from three centers were retrospectively included. Radiomics features were extracted from CT images, and the least absolute shrinkage and selection operator Cox regression algorithm was employed to select radiomics features. Clinical risk factors associated with DFS were selected through univariate and multivariate Cox regression analysis to build the clinical model. A predictive nomogram was developed by amalgamating pertinent clinical risk factors and radiomics features. The predictive performance of the nomogram was evaluated using the C-index, calibration curve, and decision curve. DFS probabilities were estimated using the Kaplan-Meier method.
Integrating the retained eight radiomics features and three clinical risk factors (pathological N stage, microsatellite instability, perineural invasion), a nomogram was constructed. The C-index for the nomogram were 0.819 (95% CI=0.794-0.844), 0.782 (95% CI=0.740-0.824), 0.786 (95% CI=0.753-0.819), and 0.803 (95% CI=0.765-0.841) in the training set, internal validation set, external validation set 1, and external validation set 2, respectively. The calibration curves demonstrated a favorable congruence between the predicted and observed values as depicted by the nomogram. The decision curve analysis underscored that the nomogram yielded a heightened clinical net benefit.
The constructed radiomics nomogram, amalgamating the radiomics features and clinical risk factors, exhibited commendable performance in the individualized prediction of postoperative DFS in CRC patients.