Magnetic resonance imaging-based lymph node radiomics for predicting the metastasis of evaluable lymph nodes in rectal cancer.

Background: Lymph node (LN) staging in rectal cancer (RC) affects treatment decisions and patient prognosis. For radiologists, the traditional preoperative assessment of LN metastasis (LNM) using magnetic resonance imaging (MRI) poses a challenge.

Aim: To explore the value of a nomogram model that combines Conventional MRI and radiomics features from the LNs of RC in assessing the preoperative metastasis of evaluable LNs.

Methods: In this retrospective study, 270 LNs (158 nonmetastatic, 112 metastatic) were randomly split into training ( = 189) and validation sets ( = 81). LNs were classified based on pathology-MRI matching. Conventional MRI features [size, shape, margin, T2-weighted imaging (T2WI) appearance, and CE-T1-weighted imaging (T1WI) enhancement] were evaluated. Three radiomics models used 3D features from T1WI and T2WI images. Additionally, a nomogram model combining conventional MRI and radiomics features was developed. The model used univariate analysis and multivariable logistic regression. Evaluation employed the receiver operating characteristic curve, with DeLong test for comparing diagnostic performance. Nomogram performance was assessed using calibration and decision curve analysis.

Results: The nomogram model outperformed conventional MRI and single radiomics models in evaluating LNM. In the training set, the nomogram model achieved an area under the curve (AUC) of 0.92, which was significantly higher than the AUCs of 0.82 ( < 0.001) and 0.89 ( < 0.001) of the conventional MRI and radiomics models, respectively. In the validation set, the nomogram model achieved an AUC of 0.91, significantly surpassing 0.80 ( < 0.001) and 0.86 ( < 0.001), respectively.

Conclusion: The nomogram model showed the best performance in predicting metastasis of evaluable LNs.