A novel MRI-based radiomics for preoperative prediction of lymphovascular invasion in rectal cancer.

BACKGROUND

To develop and validate a clinical-radiomics model for preoperative prediction of lymphovascular invasion (LVI) in rectal cancer.

METHODS

This retrospective study included data from 239 patients with pathologically confirmed rectal adenocarcinoma from two centers, all of whom underwent MRI examinations. Cases from the first center (n = 189) were randomly divided into a training set and an internal validation set at a 7:3 ratio, while cases from the second center (n = 50) constituted the external validation set. The clinical features and MRI imaging characteristics of the patients in the training set were analyzed. Univariate and multivariate logistic regression analyses were used to identify independent risk factors for LVI in rectal cancer, and these risk factors were then used to construct a clinical model. Regions of interest (ROIs) were delineated on T2-weighted imaging (T2WI) and diffusion-weighted imaging (DWI) sequences for feature extraction. After feature reduction and selection, the most strongly correlated features were identified, and their respective regression coefficients were calculated to construct the radiomics model. Finally, a combined clinical-radiomics model was built using a weighted linear combination of features and was visualized as a nomogram. The predictive performance of each model was quantified using receiver operating characteristics (ROC) curves and the area under the curve (AUC) in both training and validation sets, with DeLong analysis being used to compare model performance. Decision curve analysis (DCA) was used to evaluate the clinical utility of each model in the validation sets.

RESULTS

In the 239 patients, the combined model outperformed the clinical and radiomics models in predicting LVI in rectal cancer. The combined model showed excellent predictive performance in the training, internal validation, and external validation sets, with AUCs of 0.90 (0.88-0.97), 0.88 (0.78-0.99), and 0.88 (0.78-0.95), respectively. The sensitivity values were 75.9%, 68.8%, and 80.0%, respectively, and the specificity values were 90.3%, 92.7%, and 88.6%. DCA results indicated that the nomogram of the combined model had superior clinical utility compared with the clinical and radiomics models.

CONCLUSIONS

The clinical-radiomics nomogram serves as a valuable tool for non-invasive preoperative prediction of LVI status in patients with rectal cancer.