Objective:

To establish and validate a radiomics nomogram in the non-invasive prediction of extramural venous invasion（EMVI） in rectal cancer.

Methods:

This retrospective study included preoperative images of 147 rectal cancer patients in The First Affiliated Hospital of Nanjing Medical University. All patients were randomly divided into training cohort and validation cohort with a ratio of 7∶3. Five clinical factors

including age

sex

preoperative histological grade by colonoscopy

carcinoembryonic antigen (CEA) and carbohydrate antigen (CA) 19-9 levels and six high revolution magnetic resonance imaging (MRI) features

including tumor site and length

T stage

N stage

circumferential resection margin (CRM) and MRI defined-EMVI (mrEMVI) were recorded. Image segmentation was performed by manually delineating the whole tumors on readout segmentation of long variable echo-trains (RESOLVE) diffusion-weighted imaging (DWI) and then copied onto the corresponding apparent diffusion coefficient (ADC) maps. The maximum-relevance and minimum-redundancy (mRMR) and the least absolute shrinkageand selection operator (LASSO) methods were used to select radiomics features. Logistic regression analysis was employed to construct models based on clinical factors and high revolution MRI features (clinical model)

tumor radiomics features (radiomics model)

and clinical model combined with tumor radiomics features (combined model nomogram). Then calibration curve was used to evaluate the calibration efficiency of nomogram. Receiver operating characteristic (ROC) curve was performed to assess the diagnostic efficacy of each model in two cohorts. The area under curve (AUC) was calculated for each ROC curve. The DeLong test was conducted to compare AUCs between models. Decision curve analysis (DCA) was performed to assess the clinical usefulness of the three models

by quantifying the net benefits at different threshold probabilities in the validation cohort.

Results:

The combined model showed higher AUCs (training cohort 0.928

validation cohort 0.891) than clinical model and radiomics model. The result of DCA indicated that using nomograms of combined model to predict EMVI gains more benefits than clinical model and radiomics model at any threshold probabilities in the validation cohort.

Conclusion:

The radiomics nomogram

incorporating RESOLVE ADC based-radiomic features from the original tumor with clinical predict factors was promising to serve as a reliable clinical tool for preoperative non-invasive prediction of EMVI in rectal cancer.