Prediction of Preoperative Synchronous Distant Metastasis of Rectal Cancer Based on MRI Radiomics Model

Purpose The objective of this study was to develop and validate a new non-invasive artificial intelligence (AI) model based on preoperative magnetic resonance imaging (MRI) data to predict the presence of synchronous distant metastasis (SDM) in rectal cancer (RC). Methods 169 eligible RC patients were enrolled, and T2WI and DWI sequence images were collected. The radiomics features were extracted through the PyRadiomics package of Python language, and a total of 1688 radiomics features were extracted, including first-order features, shape features, texture features, and Baud signs. One clinical model and three comprehensive models of clinical imaging were constructed. Five indexes including receiver operating characteristic (ROC), area under curve (AUC), accuracy, sensitivity, specificity, and 95% confidence interval (CI) were selected to evaluate the model. The clinical model using four independent risk factors (CEA, age, CA199, and T stage). Combining the clinical factors and imaging characteristics of different sequences, we established three clinically-imaging models: the DWI + clinical model, the T2W + clinical model, and the nomogram (radiomics + clinical) model. Results This nomogram model performed the best in predicting rectal cancer SDM. In the training set, the AUC, accuracy, sensitivity, specificity and 95%CI of the nomogram model were 0.93, 0.85, 0.85, 0.86, 0.89–0.96, respectively. In the test set, five indexes of the nomogram model were 0.94, 0.89, 0.88, 0.89, and 0.79 ~ 0.97, respectively. The correction plots were consistent between the predictions of the clinical radiomics model and the actual observed probabilities. Decision curve analysis showed that the nomogram model achieved the highest net benefit on the training set and the test set compared to the clinical model and the radiomics model. Conclusion Our predictive model is valuable for guiding and managing patients with rectal cancer SDM, providing options for improving patient treatment decisions and guiding personalized treatment regimens.