To Explore The Predictive Value Of MRI Radiomics In Lymph Node Metastasis Of Rectal Cancer

Objective: To explore the predictive value of MRI radiomics image feature extraction in lymph node metastasis of rectal cancer.Method:A retrospective study was conducted on a total of 105 patients(73 patients with lymph node metastasis and 32 patients with non-lymph node metastasis)who underwent radical resection of rectal cancer in Yijishan-Hospital,the First Affiliated Hospital of Wannan Medical College,from November 2018 to December 2020 and were pathologically diagnosed as malignant rectal cancer after surgery.All patients had complete pelvic MRI data and postoperative pathological data,and the images were clear and the structure was identifiable.Patients were stochastically split into training group and validation group,of which 70 cases were in the training group and 36 cases were in the validation group.The MRI images of patients were delineated by the software 3Dslicer(version 4.11.2),and the software was equipped with internal plug-ins for 3D reconstruction and data extraction of the delineated areas.140 radiologic features were proposed for each case,and univariate statistical test was performed on the extracted features to compare the metastatic group and the non-metastatic group.The features with low similarity were selected,and then the independent sample t-test was performed on the selected features to evaluate the univariate correlation between the features and lymph node metastasis,and finally 12 radiomics features were obtained.The radiomics model was established for the acquired features and the random forest algorithm was used to establish the radiomics model.Finally,the model was used as a validation set to judge the lymph node metastasis and evaluate the effectiveness of the established model.Results:The radiological model established in this study has good performance in the training cohort and validation cohort,and has a good predictive value for rectal cancer with lymph node metastasis.Through multivariate logic analysis,it is found that the radiological feature is an independent predictor of lymph node metastasis.The area AUC under the ROC of the subject working characteristic curve obtained by the model in the training queue and the verification queue is 0.817 and 0.735,respectively,and there is no difference in the effectiveness of the model between the training cohort and the validation cohort.Conclusion: This article presents a radiological model based on pre-treatment MRI data for the prediction of lymph node metastasis in patients with rectal cancer which has been validated and wieldy.This model has certain predictive value for lymph node metastasis in patients with rectal cancer.