Construction of a predictive model for pathological grading of rectal neuroendocrine tumors based on MRI features
10.3760/cma.j.cn112152-20220211-00092
- VernacularTitle:基于MRI特征构建直肠神经内分泌肿瘤病理分级的预测模型
- Author:
Wenjing PENG
1
;
Lijuan WAN
;
Hongmei ZHANG
;
Shuangmei ZOU
;
Han OUYANG
;
Xinming ZHAO
;
Chunwu ZHOU
Author Information
1. 国家癌症中心 国家肿瘤临床医学研究中心 中国医学科学院北京协和医学院肿瘤医院影像诊断科，北京 100021
- Keywords:
Rectal neoplasms;
Neuroendocrine tumors;
Magnetic resonance imaging;
Pathological grading
- From:
Chinese Journal of Oncology
2022;44(8):851-857
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To explore the value of MRI features in predicting the pathological grade of rectal neuroendocrine tumors and to develop a predicting model.Methods:A retrospective analysis was performed on 30 cases of rectal neuroendocrine tumors confirmed by surgery and pathology between 2013 and 2019. All of them underwent plain rectal MRI, DWI and dynamic contrast-enhanced MRI. The clinical features and MRI characteristics (ie. tumor location, maximum tumor diameter, boundary, growth pattern, enhancement of three-staged lesions, and the lymph node metastasis) were analyzed by statistical methods to evaluate the difference between different tumor pathologic grades (G1, G2 and G3). Characteristics with statistical significance were analyzed by collinearity diagnostics, and stepwise regression method was used to select independent predictors. Ordinal logistic regression analysis was then conducted to develop the predicting model.Results:Maximum tumor diameter, tumor boundary, growth pattern, mr-T, mr-N, EMVI, MRF, T2WI signal intensity, tumor enhancement degree in venous phase and distant metastasis were closely correlated with the pathological grade of rectal neuroendocrine tumors ( P<0.001, 0.001, 0.001, <0.001, 0.001, 0.004, 0.024, 0.015, 0.001, and <0.001, respectively). The mr-T and tumor enhancement degree in venous phase were identified as the independent predictors to construct the prediction model. The model got ideal performance in predicting the grades, with the areas under the receiver operating characteristic (ROC) curves (AUCs) of 0.945, 0.624 and 0.896, the sensitivities were 75.0%, 85.7%, and 90.9% and corresponding specificities were 88.9%, 52.6% and 93.3% for G1, G2 and G3 rectal neuroendocrine tumors, respectively. Conclusion:The model based on mr-T and tumor enhancement degree in venous phase can serve as a clinical tool for predicting the pathological grade of rectal neuroendocrine tumors.
