双能CT影像组学在甲状腺结节定性诊断中的价值研究  

Value of Dual-Energy CT Radiomics in the Qualitative Diagnosis of Thyroid Nodules

作　　者：韦启君[1] 林盛才[1] 黄福灵[1] 周春燚 WEI Qijun;LIN Shengcai;HUANG Fuling;ZHOU Chunyi(Department of Radiology,The First Affiliated Hospital of Guangxi Medical University,Nanning Guangxi 530021,China)

机构地区：[1]广西医科大学第一附属医院放射科,广西南宁530021

出　　处：《中国医疗设备》2022年第12期87-90,共4页China Medical Devices

基　　金：广西医疗卫生适宜技术开发与推广应用项目(S2021091)。

摘　　要：目的探讨基于双能量CT图像的影像组学预测模型对甲状腺结节良恶性的鉴别诊断价值。方法回顾性分析我院经病理证实的55例甲状腺结节患者的临床和影像资料,其中良性结节28例,恶性结节27例。患者术前均行甲状腺双能量CT平扫和增强扫描。通过双能量分析得到各期的Mixed图、VNC图及iodine图,分别对各期图像进行3D感兴趣区(Region of Interest,ROI)勾画及组学特征提取,采用单因素分析与最小绝对收缩算子(Least Absolute Shrinkage and Selection Operator,LASSO)算法进行特征筛选,使用多因素逻辑回归分别构建混合能量模型及双能量模型。应用受试者操作特征(Receiver Operating Characteristic,ROC)曲线及曲线下面积(Area Under Curve,AUC)对模型进行验证,评价影像组学特征预测良恶性甲状腺结节的效能。结果经过特征筛选得到8个影像组学特征用于构建甲状腺结节良恶性鉴别模型。混合能量预测模型的AUC为0.86(95%CI:0.75~0.96),特异度和敏感度分别为72.4%、92.3%,诊断准确率为81.8%;双能量预测模型的AUC为0.95(95%CI:0.89~1),特异度和灵敏度分别为86.2%、100%,诊断准确率为92.7%。结论基于双能量CT的影像组学预测模型对良恶性甲状腺结节有较高的诊断效能。Objective To investigate the value of a radiomics prediction model based on dual-energy CT images in the differential diagnosis of benign and malignant thyroid nodules.Methods The clinical and imaging data of 55 patients with thyroid nodules confirmed by pathology in our hospital were retrospectively analyzed,including 28 benign nodules and 27 malignant nodules.All patients underwent the dual-energy CT of thyroid routine scan and enhanced scan.The Mixed maps,VNC maps and iodine maps of each phase were obtained by dual energy analysis,3D region of interest(ROI)delineation and omics feature extraction were performed on the images of each phase respectively.Unifactor analysis and least absolute shrinkage and selection operator(LASSO)algorithm were used for feature screening.Multi-factor Logistic regression was used to construct a mixed-energy model and a dual-energy model.Receiver operating characteristic(ROC)curve and the area under curve(AUC)were used to verify the model,and the efficacy of imaging omics characteristics in predicting benign and malignant thyroid nodules was evaluated.Results After feature screening,8 radiomics features were obtained to establish the differential model of benign and malignant thyroid nodules.The AUC of the mixed energy prediction model was 0.86(95%CI:0.75~0.96),the specificity and sensitivity were 72.4%and 92.3%,respectively,and the ratio of diagnostic accuracy was 81.8%.The AUC of the dual-energy prediction model was 0.95(95%CI:0.89~1),the specificity and sensitivity were 86.2%and 100%,respectively,and the ratio of diagnostic accuracy was 92.7%.Conclusion The radiomics prediction model based on dual-energy CT has high diagnostic efficacy for benign and malignant thyroid nodules.

关 键 词：甲状腺结节 影像组学 双能量 计算机断层成像 鉴别诊断 

分 类 号：R581[医药卫生—内分泌]