【摘要】 目的 探讨基于MRI影像组学、深度学习(DL)及临床特征建立联合模型预测子宫内膜癌(EC)淋巴结转移(LNM)的价值。方法 回顾性分析经术后病理证实为EC的临床及MRI资料。按7:3比例随机分为训练集(130例)和测试集(56例)。采用单因素及多因素Logistic回归分析筛选LNM的独立危险因素,建立临床模型;从MRI图像中提取影像组学和DL特征并建立影像组学、DL及DL-影像组学(DLradiomics,DLR)模型;最后联合影像组学、DL及临床特征建立联合模型。通过曲线下面积(AUC)、校准曲线和决策曲线评估模型性能。结果 年龄、组织学分级均为LNM(+)的独立危险因素(P<0.05)。联合模型在训练集和测试集中AUC是最高的,分别为0.964 95%Cl=0.934-0.994)和0.860 95%Cl=0.677-1.000)。联合模型的校准度较高,临床净收益更大。结论 基于MRI的影像组学特征、DL特征联合临床特征建立的联合模型预测EC LNM具有较高诊断效能。更多还原

【Abstract】 Objective To investigate the value of a combined model based on MRI radiomics features,deep learning(DL) features,and clinical features in predicting lymph node metastasis(LNM) in patients with endometrial carcinoma(EC).Methods A retrospective analysis was conducted on clinical and MRI data from patients pathologically confirmed with EC.The patients were randomly divided into a training set(n=130) and a test set(n=56) in a 7:3 ratio.U niva riate and multivariate logistic regression analyses were used to identify independent clinical risk factors for LN M,and a clinical model was esta blished.Radiomics and DL features were extracted from MRI images to construct radiomics,DL,and DLradiomics(DLR) models.Finally,a combined model integrating radiomics,DL,and clinical features was developed.Model performance was evaluated using the area under the curve(AUC),calibration curves,and decision curve analysis.Results Age and histological grade were identified as independent risk factors for LNM(P<0.05).The combined model achieved the highest AUC in both the training and test sets,with values of 0.964(95% Cl:0.934-0.994) and 0.860(95% Cl:0.677-1.000),respectively.The combined model demonstrated high calibration and greater clinical net benefit.Conclusion The com bined model based on MRI radiomics featu res,DL features,and clinical features demonstrates high diagnostic performance in predicting LNM in EC patients.更多还原