A CT-based radiomics model for preoperative prediction of lymphovascular invasion in colorectal cancer.

OBJECTIVE

This study aimed to develop and validate a computed tomography (CT)-based radiomics nomogram for the preoperative prediction of lymphovascular invasion (LVI) in colorectal cancer (CRC).

METHODS

In this retrospective study, CRC patients were randomly partitioned into training and validation sets at a 7:3 ratio, and were classified as LVI-positive or LVI-negative based on postoperative histopathology. In the training set, univariate analysis, Least Absolute Shrinkage and Selection Operator (LASSO), and multivariate Logistic regression analysis were used to identify the influencing factors of LVI. A combined nomogram integrating the predictors was developed. Model discrimination was evaluated using the area under the receiver operating characteristic curve (AUC). Calibration and clinical utility were assessed with calibration curves and decision curve analysis (DCA), respectively.

RESULTS

Among the 252 patients in the training set, 70 (27.78%) were LVI-positive. In the validation set of 108 patients, 31 (28.70%) were LVI-positive. Multivariate analysis identified the tumor volume, maximum tumor diameter, depth of tumor invasion, maximum short-axis diameter of regional lymph nodes, Carcinoembryonic Antigen level, Neutrophil-to-Lymphocyte Ratio (NLR), standard deviation of CT value of tumor parenchyma, and Rad-score as independent predictors of LVI (P<0.05). The nomogram incorporating these factors demonstrated good predictive performance, with an AUC of 0.776 (95% CI: 0.694-0.858) in the training set and 0.722 (95% CI: 0.575-0.869) in the validation set. Calibration curves indicated satisfactory agreement between predictions and observations. DCA confirmed the clinical net benefit of the nomogram across a wide range of risk thresholds.

CONCLUSION

The proposed CT-based radiomics nomogram, featuring the Rad-score, NLR, and intratumoral heterogeneity, provides a non-invasive and reliable tool for preoperative LVI prediction in CRC. This model holds promise for enhancing individualized clinical strategy and prognosis assessment.