PDF(Pubmed)
Abstract：
UNASSIGNED: This study aimed to develop and validate a radiomics nomogram based on multiparameter MRI for preoperative differentiation of type II and type I endometrial carcinoma (EC).
UNASSIGNED: A total of 403 EC patients from two centers were retrospectively recruited (training cohort, 70 %; validation cohort, 30 %). Radiomics features were extracted from T2-weighted imaging, dynamic contrast-enhanced T1-weighted imaging at delayed phase(DCE4), and apparent diffusion coefficient (ADC) maps. Following dimensionality reduction, radiomics models were developed by logistic regression (LR), random forest (RF), bootstrap aggregating (Bagging), support vector machine (SVM), artificial neural network (ANN), and naive bayes (NB) algorithms. The diagnostic performance of each radiomics model was evaluated using the ROC curve. A nomogram was constructed by incorporating the optimal radiomics signatures with significant clinical-radiological features and immunohistochemistry (IHC) markers obtained from preoperative curettage specimens. The diagnostic performance and clinical value of the nomogram were evaluated using ROC curves, calibration curves, and decision curve analysis (DCA).
UNASSIGNED: Among the radiomics models, the NB model, developed from 12 radiomics features derived from ADC and DCE4 sequences, exhibited strong performance in both training and validation sets, with the AUC values of 0.927 and 0.869, respectively. The nomogram, incorporating the radiomics model with significant clinical-radiological features and IHC markers, demonstrated superior performance in both the training (AUC = 0.951) and the validation sets (AUC = 0.915). Additionally, it exhibited excellent calibration and clinical utility.
UNASSIGNED: The radiomics nomogram has great potential to differentiate type II from type I EC, which may be an effective tool to guide clinical decision-making for EC patients.
UNASSIGNED: A total of 403 EC patients from two centers were retrospectively recruited (training cohort, 70 %; validation cohort, 30 %). Radiomics features were extracted from T2-weighted imaging, dynamic contrast-enhanced T1-weighted imaging at delayed phase(DCE4), and apparent diffusion coefficient (ADC) maps. Following dimensionality reduction, radiomics models were developed by logistic regression (LR), random forest (RF), bootstrap aggregating (Bagging), support vector machine (SVM), artificial neural network (ANN), and naive bayes (NB) algorithms. The diagnostic performance of each radiomics model was evaluated using the ROC curve. A nomogram was constructed by incorporating the optimal radiomics signatures with significant clinical-radiological features and immunohistochemistry (IHC) markers obtained from preoperative curettage specimens. The diagnostic performance and clinical value of the nomogram were evaluated using ROC curves, calibration curves, and decision curve analysis (DCA).
UNASSIGNED: Among the radiomics models, the NB model, developed from 12 radiomics features derived from ADC and DCE4 sequences, exhibited strong performance in both training and validation sets, with the AUC values of 0.927 and 0.869, respectively. The nomogram, incorporating the radiomics model with significant clinical-radiological features and IHC markers, demonstrated superior performance in both the training (AUC = 0.951) and the validation sets (AUC = 0.915). Additionally, it exhibited excellent calibration and clinical utility.
UNASSIGNED: The radiomics nomogram has great potential to differentiate type II from type I EC, which may be an effective tool to guide clinical decision-making for EC patients.
摘要：
■本研究旨在开发和验证基于多参数MRI的放射组学列线图，用于II型和I型子宫内膜癌（EC）的术前分化。
■回顾性招募了来自两个中心的403例EC患者（培训队列，70%；验证队列，30%）。从T2加权成像中提取影像组学特征，延迟相位动态对比增强T1加权成像（DCE4），和表观扩散系数(ADC)图。降维之后，通过逻辑回归(LR)开发了影像组学模型，随机森林(RF)，引导聚合(Bagging)，支持向量机(SVM)，人工神经网络(ANN)，和朴素贝叶斯(NB)算法。使用ROC曲线评估每个影像组学模型的诊断性能。通过将最佳的影像组学特征与从术前刮宫标本获得的重要临床放射学特征和免疫组织化学（IHC）标记相结合，构建了列线图。使用ROC曲线评估列线图的诊断性能和临床价值，校正曲线,和决策曲线分析(DCA)。
■在影像组学模型中，NB模型,从ADC和DCE4序列衍生的12个影像组学特征开发而成，在训练集和验证集中均表现出强劲的性能，AUC值分别为0.927和0.869。列线图，整合具有重要临床放射学特征和IHC标记的影像组学模型，在训练（AUC=0.951）和验证集（AUC=0.915）中都表现出优异的性能。此外,它表现出优异的校准和临床实用性。
■放射组学列线图具有区分II型和I型EC的巨大潜力，这可能是指导EC患者临床决策的有效工具。
■回顾性招募了来自两个中心的403例EC患者（培训队列，70%；验证队列，30%）。从T2加权成像中提取影像组学特征，延迟相位动态对比增强T1加权成像（DCE4），和表观扩散系数(ADC)图。降维之后，通过逻辑回归(LR)开发了影像组学模型，随机森林(RF)，引导聚合(Bagging)，支持向量机(SVM)，人工神经网络(ANN)，和朴素贝叶斯(NB)算法。使用ROC曲线评估每个影像组学模型的诊断性能。通过将最佳的影像组学特征与从术前刮宫标本获得的重要临床放射学特征和免疫组织化学（IHC）标记相结合，构建了列线图。使用ROC曲线评估列线图的诊断性能和临床价值，校正曲线,和决策曲线分析(DCA)。
■在影像组学模型中，NB模型,从ADC和DCE4序列衍生的12个影像组学特征开发而成，在训练集和验证集中均表现出强劲的性能，AUC值分别为0.927和0.869。列线图，整合具有重要临床放射学特征和IHC标记的影像组学模型，在训练（AUC=0.951）和验证集（AUC=0.915）中都表现出优异的性能。此外,它表现出优异的校准和临床实用性。
■放射组学列线图具有区分II型和I型EC的巨大潜力，这可能是指导EC患者临床决策的有效工具。
