摘要
目的:探讨基于磁共振T_(2)WI影像组学联合内膜形态学特征对子宫内膜纤维化分级评估的价值。方法:前瞻性纳入经宫腔镜证实的轻中度子宫内膜纤维化患者43例、重度子宫内膜纤维化患者99例和健康对照80例。按照7∶3比例随机分为训练集和验证集,构建两个分类任务:①健康对照组vs.轻中度组;②轻中度组vs.重度组。采集子宫矢状面及冠状面T_(2)WI图像,测量子宫内膜厚度(ET)、宫腔长度(LUC)、子宫颈及峡部长度(LCI)、宫腔上端宽度(WUUC)。采用单因素和多因素Logistic回归筛选子宫内膜纤维化分级相关的内膜形态学参数,建立形态学模型。从子宫矢状面T_(2)WI图像上应用3D-Slicer软件独立手动逐层勾画子宫体内膜区域并提取影像组学特征,通过最大相关最小冗余(mRMR)和最小绝对收缩和选择算子(LASSO)进行特征降维和筛选,建立影像组学模型并计算影像组学评分。联合影像组学评分和内膜形态学特征构建联合模型。采用受试者操作特征(ROC)曲线评估模型的诊断性能。采用Delong检验比较ROC曲线下面积(AUC)差异。绘制校正曲线以评估模型的拟合优度,决策曲线(DCA)评估模型的临床应用价值。结果:最终分别共筛选出8个、5个影像组学特征构建影像组学模型区分健康对照组与轻中度组、轻中度组vs.重度组。对于区分健康对照组与轻中度组,联合模型在训练集和测试集中的AUC分别为0.975和0.957,均优于ET+LUC模型(P均<0.05),而与影像组学模型差异无统计学意义(P均>0.05)。对于区分轻中度组与重度组,联合模型在训练集和测试集中的AUC分别为0.938和0.897,均优于ET+LCI模型(P均<0.05),而与影像组学模型差异无统计学意义(P均>0.05)。校准曲线提示各联合模型拟合度均较好,DCA曲线显示在较大的阈值概率范围内各联合模型均具有较好的临床净收益。结论:磁共振T_(2)WI影像组学联合内膜形态学特征在子宫内膜纤维化分级评估中具有较高价值。
Objective:To investigate the value of magnetic resonance T_(2)-weighted imaging(T_(2)WI)-based radiomics combined with endometrial morphological features for the grading of endometrial fibrosis.Methods:This prospective study included 43 patients with mild-to-moderate endometrial fibrosis,99 patients with severe endometrial fibrosis confirmed by hysteroscopy,and 80 healthy controls.All subjects were randomly divided into training set and test set at a 7:3 ratio for two classification tasks:healthy controls group vs.mild-to-moderate group;mild-to-moderate group vs.severe group.Sagittal and coronal T_(2)WI images of the uterus were acquired to measure endometrial thickness(ET),length of the uterine cavity(LUC),length of the cervix and isthmus(LCI),and width of upper uterine cavity(WUUC).Univariate and multivariate Logistic regression analyses were performed to identify endometrial morphological parameters associated with the grading of endometrial fibrosis,and morphologic models were constructed accordingly.The endometrial region of corpus uteri was manually delineated slice by slice on sagittal T_(2)WI images using 3D slicer software for radiomics feature extraction.Feature selection was performed using the maximum relevance minimum redundancy(mRMR)method followed by the least absolute shrinkage and selection operator(LASSO)regression to construct radiomics models and compute radiomics scores.Combined models integrating radiomics scores and morphological features were developed.The diagnostic performance of each model was assessed using receiver operating characteristic(ROC)curve analysis,with the area under the curve(AUC)compared by the DeLong test.Calibration curves were plotted to evaluate model fit,and decision curve analysis(DCA)was conducted to assess clinical utility.Results:A total of 8 and 5 radiomics features were ultimately selected to construct the radiomics models,respectively.For distinguishing between the healthy control group and the mild-to-moderate group,the combined model achieved AUCs of 0.975 and 0.957 in the training and test sets,respectively,outperforming the ET+LUC model(both P<0.05),while showing no statistically significant difference compared with the radiomics model(both P>0.05).For distinguishing between the mild-to-moderate group and the severe group,the combined model achieved AUCs of 0.938 and 0.897 in the training and test sets,respectively,outperforming the ET+LCI model(both P<0.05),while showing no statistically significant difference compared with the radiomics model(both P>0.05).Calibration curves indicated good fit for each combined models,and decision curve analysis(DCA)demonstrated favorable clinical net benefits across a wide range of threshold probabilities for each combined model.Conclusions:Magnetic resonance T_(2)WI combined with endometrial morphological features demonstrates high value in grading endometrial fibrosis.
作者
梁欢欢
周楠
朱慧
陈玉灿
王欢欢
姜佩佩
胡娅莉
周正扬
LIANG Huan-huan;ZHOU Nan;ZHU Hui(Department of Radiology,Nanjing Drum Tower Hospital,Affiliated Hospital of Medical School,Nanjing University,Nanjing 210008,China)
出处
《放射学实践》
北大核心
2026年第3期282-289,共8页
Radiologic Practice
基金
中国科学院战略性先导科技专项(XDA16040302)。
