Multiparameter Magnetic Resonance Imaging-based Radiomics Model For The Preoperative Prediction Of Lymphovascular Invasion In Rectal Cancer

Objective:This study aims to investigate the application value of a fusion model based on the multiparameter magnetic resonance imaging（MRI）for preoperative prediction of lymphovascular invasion（LVI）in rectal cancer.Methods:Retrospective study.A total of 224 patients with rectal cancer who underwent radical resection for rectal cancer and met the criteria in Shanxi Cancer Hospital from January2016 to December 2019 were enrolled,including 129 males and 95 females,aged 28 to 83（61.3±9.7）years.All patients underwent multi-parameter MRI examination within two weeks before surgery,and were randomly divided into two groups,the training group（n=157）and the validation group（n=67）according to a ratio of 7:3.Clinicopathologic data and multi-parameter MRI data of patients were collected.ITK-SNAP image segmentation software was used to manually delineate the ROI of tumor slice by slice on the images of T₂ weighted imaging（T₂WI）,diffusion-weighted imaging（DWI）and enhanced T₁weighted imaging（cT₁WI）sequences to obtain the volume of interest（VOI）.The delineation information of the DWI is copied onto the apparent diffusion coefficient（ADC）map.Pyradiomics software was used to extract 3948 radiomics features from each patient.The features were screened by the three-step dimensionality reduction method of maximum relevance minimum redundancy（m RMR）,minimum absolute shrinkage and selection operator（LASSO）regression and multiple logistic regression,and radiomics signature were constructed.Independent predictors of clinicopathologic features and MRI features were screened by univariate and multivariate logistic regression analysis.A total of 5 prediction models were constructed based on the single sequence of T₂WI,ADC and cT₁WI and combined with the above three sequences,as well as the fusion model with clinicopathological features,and the corresponding nomogram was constructed.Area under receiver operating characteristic（ROC）curve（AUC）,calibration curve and decision curve were used to evaluate the efficacy and clinical benefit of the model.Results:Postoperative pathological examination confirmed LVI positive in 70 patients and negative in 154 patients.There were no significant differences in clinicopathologic features and MRI findings between the training and validation groups（all P values>0.05）.Six key features were identified after screening of 3948 radiomics features extracted from each patient,all of which were associated with colorectal cancer LVI（all P values<0.05）.Carcinoembryonic antigen（CEA）was an independent predictor of colorectal cancer（P<0.05）.The AUC of the radiomics model based on the single and combined sequences of T₂WI,ADC and cT₁WI were 0.765,0.772,0.776 and 0.878 in the training group,and0.741,0.739,0.764 and 0.846 in the validation group,respectively.The AUC of the fusion model training group and validation group constructed by CEA were 0.899 and 0.876,respectively,which showed the best prediction efficiency.The calibration curve shows that the fusion model has good calibration performance.The decision curve of the verification group shows that the fusion model has the maximum net benefit when the threshold probability ranges from 0.10 to 0.20 and from 0.35 to 0.90.Conclusion:The fusion model constructed based on the radiomics features of multi-parameter MRI and CEA has high diagnostic efficacy in predicting LVI of rectal cancer before surgery.Its visual nomogram can be used as an effective tool for predicting LVI before surgery,so as to help clinical evaluation of patients’prognosis and make treatment decisions.