Radiomics feature as a preoperative predictive of lymphovascular invasion in early-stage endometrial cancer: A multicenter study

BackgroundThe presence of lymphovascular space invasion (LVSI) has been demonstrated to be significantly associated with poor outcome in endometrial cancer (EC). No effective clinical tools could be used for the prediction of LVSI preoperatively in early-stage EC. A radiomics nomogram based on MRI was established to predict LVSI in patients with early-stage EC.MethodsThis retrospective study included 339 consecutive patients with early-stage EC with or without LVSI from five centers. According to the ratio of 2:1, 226 and 113 patients were randomly assigned to a training group and a test group, respectively. Radiomics features were extracted from T1-weighted imaging (T1WI), T2-weighted imaging (T2WI), contrast-enhanced (CE), diffusion-weighted imaging (DWI), and apparent diffusion coefficient (ADC) maps. The radiomics signatures were constructed by using the Least Absolute Shrinkage and Selection Operator (LASSO) algorithm in the training group. The radiomics nomogram was developed using multivariable logistic regression analysis by incorporating radiomics signatures and clinical risk factors. The sensitivity, specificity, and AUC of the radiomics signatures, clinical risk factors, and radiomics nomogram were also calculated.ResultsThe individualized prediction nomogram was constructed by incorporating the radiomics signatures with the clinical risk factors (age and cancer antigen 125). The radiomics nomogram exhibited a good performance in discriminating between negative and positive LVSI patients with AUC of 0.89 (95% CI: 0.83–0.95) in the training group and of 0.85 (95% CI: 0.75–0.94) in the test group. The decision curve analysis indicated that clinicians could be benefit from the using of radiomics nomogram to predict the presence of LVSI preoperatively.ConclusionThe radiomics nomogram could individually predict LVSI in early-stage EC patients. The nomogram could be conveniently used to facilitate the treatment decision for clinicians.

Abstract licence: CC BY

A person counts once a day per paper. We keep no cookie and no address: a reader is a hash that changes every day. Any month under five reads shows no number.