摘要
目的探讨基于CT增强的放射组学特征直方图参数对鉴别腮腺多形性腺瘤(PA)与腺淋巴瘤(AL)的应用价值。方法收集经病理确诊的23例AL与21例PA,运用MaZda软件提取并分析CT增强静脉期图像中的肿瘤放射组学直方图参数,包括均值、方差、偏度、峰度和第1、10、50、90、99百分位数,运用受试者工作特征曲线(ROC)对组间有统计学意义的参数进行分析并评价诊断效能,利用多变量Logistic回归分析对组间有统计学意义的参数进行建模并运用ROC曲线评价其模型效能。结果两组中5个直方图参数(均值与方差、第50、90、99百分位数)间的差异有统计学意义(P均<0.05),其中第99百分位数在两组中具有最高的鉴别诊断效能,曲线下面积(AUC)达0.85,对应的特异度及灵敏度均为80%。利用这5个直方图参数建立多参数Logistic回归诊断模型的AUC、特异度及灵敏度分别为0.945、92.7%、86.3%。结论基于CT增强的放射组学特征直方图参数能够有效的对PA与AL进行鉴别。
ObjectiveTo explore the value of radiomics characteristic histogram parameters based on CT enhancement in distinguishing parotid pleomorphic adenoma(PA)from adenolymphoma(AL).MethodsA total of 23 patients with AL and 21 patients with PA confirmed by pathology were collected.The tumor radiomics histogram parameters in CT enhanced venous phase images were extracted and analyzed by MaZda software,including mean,variance,skewness,kurtosis and the 1st,10th,50th,90th and 99th percentile.Receiver operator characteristic curve(ROC)was used to analyze the statistically significant parameters between groups and evaluate the diagnostic efficiency.Multivariate logistic regression analysis was used to model the statistically significant parameters between groups,and ROC curve was used to evaluate the model efficiency.ResultsThere were statistically significant differences in five histogram parameters(mean and variance,50th,90th and 99th percentile)between the two groups(allP<0.05).The 99th percentile had the highest differential diagnostic efficacy in the two groups,the area under curve(AUC)was 0.85,and the corresponding specificity and sensitivity were 80%.The AUC,specificity and sensitivity of the multi parameter logistic regression diagnostic model established by using the five histogram parameters were 0.945,92.7%and 86.3%respectively.ConclusionCT enhanced radiomics feature histogram parameters can effectively distinguish PA from AL.
作者
于冬洋
韩雷
单奔
柳勇
赵正宇
马乐艳
Dongyang Yu;Lei Han;Ben Shan;Yong Liu;Zhengyu Zhao;Leyan Ma(Department of Radiology,Affiliated Huai′an Hospital of Xuzhou Medical University,Huai′an 223002,China;Department of Radiology,Affiliated Hospital of Xuzhou Medical University,Xuzhou 221002,China)
出处
《中华消化病与影像杂志（电子版）》
2022年第5期291-295,共5页
Chinese Journal of Digestion and Medical Imageology(Electronic Edition)
作者简介
通信作者:马乐艳,Email:304695554@qq.com
