Prediction of Fibrous Adipose Tissue Infiltration in Rectal Cancer Using MRI-Based Radiomics

Abstract

Abstract Objective To evaluate the value of preoperative prediction of fibrous adipose tissue infiltration around rectal cancer using a magnetic resonance imaging (MRI) model. Methods The preoperative MRI images of 75 patients with rectal cancer confirmed by pathology were retrospectively analysed. According to the pathological results, the patients were divided into two groups: the fibrous adipose tissue infiltration group (n = 33) and the fibrous adipose tissue noninfiltration group (n = 42). The image group features for T2-weighted imaging (T2WI) and arterial phase contrast-enhanced T1-weighted imaging (CE_T1WI) were extracted. The data were divided into a training group (n = 54) and a verification group (n = 21) at a ratio of 7:3. The training set data were reduced by maximum correlation minimum redundancy (mRMR) and least absolute shrinkage and selection operator (LASSO) regression, and feature dimension reduction was used to select the optimal imaging features. Three image group models (T2WI model, CE_T1WI model, and T2WI + CE_T1WI model) were constructed. Based on the receiver operating characteristic (ROC) curves and multivariate logistic regression analysis, a clinical + T2WI model was constructed, and its predictive efficiency and clinical benefits were evaluated. Results In total, 851 imaging features were extracted from T2WI and CE_T1WI, and 10 and 14 optimal features were obtained, respectively, which were used to construct the T2WI + CE_T1WI model. In the training group, the areas under the curve (AUCs) for the T2WI model, CE_T1WI model and T2WI + CE_T1WI model for predicting fibrous adipose tissue infiltration around rectal cancer were 0.87 (95% CI: 0.75 ~ 0.95), 0.82 (95% CI: 0.70 ~ 0.91) and 0.84 (95% CI: 0.71 ~ 0.92), respectively. In the verification group, the AUCs were 0.76 (95% CI: 0.53 ~ 0.92), 0.73 (95% CI: 0.50 ~ 0.90) and 0.76 (95% CI: 0.52 ~ 0.92), respectively. For the clinical + T2WI model, the AUC for predicting fibrous adipose tissue infiltration of rectal cancer was 0.91 (95% CI: 0.80 ~ 0.97) in the training group and 0.80 (95% CI: 0.57 ~ 0.94) in the verification group. Decision curve analysis (DCA) suggested that when the threshold was 0:1, the net return of the clinical + T2WI model was higher than that of the T2WI model. Conclusion The clinical + T2WI model is effective in the preoperative prediction of fibrous adipose tissue infiltration around rectal cancer.