基于MRI影像组学鉴别卵巢卵泡膜细胞瘤与阔韧带肌瘤  

Identification of Ovarian Thecoma and Broad Ligament Myoma Based on MRI Radiomics

作　　者：夏旭东 李铭[1] 王海彬[1] 王功夏 王亚龙 周小山 XIA Xudong;LI Ming;WANG Haibin;WANG Gongxia;WANG Yalong;ZHOU Xiaoshan(Department of Radiology,Anyang Tumor Hospital,Anyang 455001,China)

机构地区：[1]河南省安阳市肿瘤医院医学影像科,河南安阳455001

出　　处：《中国医学影像学杂志》2022年第2期145-152,共8页Chinese Journal of Medical Imaging

摘　　要：目的探讨基于MRI影像组学对卵巢卵泡膜细胞瘤(OTCA)与阔韧带肌瘤(BLM)的鉴别诊断价值。资料与方法回顾性分析安阳市肿瘤医院2016年1月—2021年3月经病理证实的76例OTCA和58例BLM的MRI图像,比较两组疾病的MRI特征。于肿瘤最大层面勾画感兴趣区提取T2WI脂肪抑制序列图像纹理特征,采用分层抽样方式按照7∶3分为训练组104例和测试组30例,根据病理结果分为OTCA亚组和BLM亚组。基于训练组,使用最小绝对收缩和选择算子回归分析筛选关键特征,根据回归模型中变量的回归系数,建立线性方程计算影像组学标签评分。采用受试者工作特征(ROC)曲线评价基于MRI图像特征、影像组学及其组合区分两种疾病的能力。结果共4个MRI特征为鉴别两组疾病的独立特征,包括同侧卵巢可见性(χ^(2)=5.503,P<0.05)、外周囊性区(χ^(2)=7.693,P<0.05)、动脉期强化程度(P<0.05)及表观扩散系数(t=3.310,P<0.05);训练组和测试组OTCA、BLM亚组的影像组学标签评分比较,差异均有统计学意义(P<0.05)。联合MRI图像特征和影像组学标签评分诊断效能最高,在训练组和验证组中的ROC曲线下面积分别为0.900和0.891,敏感度分别为95.8%和99.8%,特异度分别为82.6%和75.0%,准确度分别为83.8%和80.1%。与传统MRI特征(训练组和验证组ROC曲线下面积分别为0.800、0.767,敏感度分别为66.0%、68.8%,特异度分别为84.2%、75.0%,准确度分别为76.9%、71.1%)相比,联合MRI图像特征和影像组学标签评分训练组和测试组诊断效能均有提高(P<0.05)。结论基于MRI影像组学可以提高MRI鉴别OTCA与BLM的能力。Purpose To investigate the value of MR image in the differential diagnosis of ovarian thecoma(OTCA)and broad ligament fibroid(BLM).Materials and Methods Patients pathologically confirmed OTCA(n=76)and BLM(n=58)in Anyang Tumor Hospital from January 2016 to March 2021 were retrospectively enrolled,and MRI features between the two groups were analyzed and compared.The texture features of T2WI fat suppression sequence image were extracted by outlining the region of interest(ROI)at the maximum level of the tumor and were divided into training group(n=104)and test group(n=30)via stratified sampling in a 7∶3 ratio,and were further divided into OTCA subgroup and BLM subgroup according to pathological findings.Based on the training group,key features were screened by using the least absolute shrinkage and selection operator regression analysis.According to the regression coefficients of the variables in the regression model,a linear equation was established to calculate the Radiomic score.Receiver operating characteristic curve was used to evaluate the ability to distinguish two diseases based on MR image features,radiomics,combination of features and radiomics.Results A total of 4 MRI features,including ipsilateral ovarian visibility(χ^(2)=5.503,P<0.05),peripheral cystic area(χ^(2)=7.693,P<0.05),arterial enhancement(P<0.05),and ADC value(t=3.310,P<0.05),were the independent features for differentiating OTCA group from BLM group.There were statistically significant differences in radiomics score of OTCA subgroup and BLM subgroup between the training group and test group(P<0.05).Combination of MRI features and radiomics score showed the highest diagnostic efficacy,with area under the curve(AUC)of 0.900 and 0.891,sensitivity of 95.8%and 99.8%,specificity of 82.6%and 75.0%,accuracy of 83.8%and 80.1%in the training and validation groups,respectively.Compared with the conventional MRI features(AUC of 0.800 and 0.767,sensitivity 66.0%and 68.8%,specificity 84.2%and 75.0%,and accuracy 76.9%and 71.1%in the training and validat

关 键 词：卵巢肿瘤 肌瘤 阔韧带 磁共振成像 影像组学 病理学 外科 诊断 鉴别 

分 类 号：R730.42[医药卫生—肿瘤] R737.31[医药卫生—临床医学]