Diagnostic Value Of MR-T2WI Radiomics For Lymph Node Metastasis Of Colorectal Cancer

Objective:To explore the diagnostic value of preoperative individualized clinical-imaging omics prediction model for lymph node metastasis in patients with colorectal cancer based on MR-T2 WI image extraction and analysis of the related imaging features of primary lesion and surrounding lymph nodes.Methods:A total of 115 patients with surgically diagnosed colorectal cancer in The Affiliated Hospital of Beihua University from January 2019 to December 2021 were selected and complete colorectal MRI images were collected before surgery.70% of patients were randomized to the training set and 30% to the validation set.The region of interest（ROI）of the primary tumor and lymph nodes around the primary lesion（lymph nodes within the mesocentric fascia and lymph nodes with a diameter of more than 2mm around rectal vessels）were mapped in 3D semi-automated t2-WI axial sequence.The corresponding imaging features were extracted from ROI mapped using Philips Pyradiomics tool.T test,minimum absolute contraction and selection operator algorithm（LASSO）,multi-factor logistic regression method are used to reduce dimension and filter feature.Multiple clinical-imaging omics models for preoperative prediction of lymph node metastasis were established by multivariate logistic regression analysis.Receiver operating characteristic curve（ROC）,area under curve（AUC）,sensitivity（SEN）,specificity（SPE）,positive predictive value（PPV）and negative predictive value（NPV）were used to compare the performance of prediction models in training sets and test sets.The nomogram was used to visualize the data,and the decision curve analysis（DCA）and the calibration curve under the nomogram model were used to evaluate the value of the predictive model in clinical application.Results:1.There was no significant difference in clinical characteristics between the training set and the test set（P>0.05）.Multivariate logistic regression showed that patient age,preoperative CA19-9,and preoperative CEA were independent predictors in the clinical model.This is consistent with other studies.+2.1200 radiographic-related features were extracted from each region of interest,and finally 8 primary lesions and 2 surrounding lymph nodes were screened out for the establishment of clinical-radiographic-histological model.3.The AUC of Clinical+Lesion+Lymph model is the best in both training and test sets.The AUC value was 0.822 in the training set and 0.850 in the validation set.4.The calibration curve analysis of the line graph constructed by the optimal prediction model showed that the prediction effect was in good agreement with the actual situation of lymph node metastasis.DCA suggests that this Nomogram model to predict lymph node metastasis of colorectal cancer could be useful in clinical practice.Conclusions:A clinical-imaging map based on MR-T2 WI imaging features and clinical risk factors is proposed to individually predict lymph node metastasis in patients with colorectal cancer before surgery.