Magnetic Resonance Imaging Radiomics-Based Model to Identify the Pathological Features and Lymph Node Metastasis in Rectal Cancer

Abstract

Purpose Pathological features and lymph node staging plays an important role in treatment decision-making. Yet, the preoperative accurate prediction of pathological features and lymph node metastasis (LNM) is challenging. Methods In this prospective study, a total of 37 patients diagnosed with histologically confirmed rectal cancer who underwent pelvic 3.0T magnetic resonance imaging (MRI) were enrolled. MRI images of both the primary tumor alongside the lymph nodes and specimens were performed with a node-to-node match and labeling. The correlation analysis, least absolute shrinkage, and selection operator (LASSO) logistic regression (LR), backward stepwise LR were mainly used for radiomics feature selection and modeling. The univariate and multivariate backward stepwise LR were used for preoperative clinical predictors selection and modeling. Results A total of 487 lymph nodes including 39 metastatic lymph nodes and 11 tumor deposits were harvested from 37 patients. The texture features of the primary tumors could successfully predict tumor differentiation using a well-established model (area under the curve (AUC) = 0.798). Sixty-nine matched lymph nodes were randomly divided into a training cohort (n = 39) and a validation cohort (n = 30). Three independent risk factors were obtained from 56 texture parameters closely related to LNM. A prediction model was then successfully developed, which provided AUC values of 0.846 and 0.733 in the training and test cohort, respectively. Further, tumor deposits produced a higher radiomics score (Rad-score) compared with LNM (P = 0.042). Conclusion The study provides two non-invasive and quantitative methods, which respectively predict the tumor differentiation and regional LNM for rectal cancer preoperatively. Ultimately, these are favorable when producing treatment protocols for rectal cancer patients.