三种不同指数模型对乳腺良恶性病变鉴别诊断效能的比较:一项基于多b值MUSEDWI的研究

Comparison of the Diagnostic Efficacy of Three Different Exponential Models in Discriminating Malignant from Benign Breast Lesions:A Study Based on Multi-bvalue MUSE-DWI

目的比较基于多b值MUSE-DWI的单指数模型、双指数模型、拉伸指数模型鉴别乳腺良恶性病变的诊断效能。方法回顾性分析62例经病理证实的乳腺疾病患者的MRI检查资料。根据病理结果,将病灶分为良恶性两组,采用单因素分析或非参数检验比较三种指数模型定量参数的组间差异。采用二元Logistic回归分析有组间差异的参数。采用DeLong检验比较三种指数模型的受试者工作特征(ROC)曲线下面积(AUC),评估三种模型诊断效能的差异。结果三种指数模型定量参数的值在恶性组均低于良性组(P<0.05)。单指数模型、双指数模型参数联合模型、拉伸指数模型参数联合模型鉴别乳腺良恶性病变的AUC分别为0.701,0.859,0.828;双指数模型参数联合模型与单指数模型的AUC有统计学差异(P=0.023),拉伸指数模型参数联合模型的AUC与双指数模型参数联合模型(P=0.52)及单指数模型(P=0.08)无统计学差异。结论三种不同指数模型有助于鉴别乳腺良恶性病变;双指数模型参数联合模型优于单指数模型。

Objective The objective of this study was to compare the diagnostic efficacy of three different exponential models(mono-exponential,bi-exponential,and stretched exponential)in discriminating the malignant from benign breast lesions,based on multi-b-value MUSE-DWI.Methods MRI data of 62 patients with pathologically confirmed breast diseases were retrospectively analyzed.Based on the pathological results,the lesions were divided into benign or malignant groups,and the differences between groups in the quantitative parameters of three different exponential models were compared using univariate analysis or non-parametric tests as appropriate.Binary logistic regression analysis was performed on parameters with significant intergroup differences.The DeLong test was used to compare the area under the receiver operating characteristic(ROC)curve(AUC)of the three models to evaluate the significant difference in their diagnostic efficacy.Results Intergroup comparisons showed that the values of the quantitative parameters of three different exponential models in the malignant group were lower than those in the benign group(P<0.05).The AUC of the three models was 0.701,0.859,and 0.828,respectively.There was a significant difference in the AUC between the combined model of the parameters of the bi-exponential mode and the mono-exponential model.The diagnostic efficacy of the combined model of the parameters of the stretched-exponential model had no significant difference with the combined model of the parameters of the bi-exponential model(P=0.52)or the mono-exponential model(P=0.08).Conclusion The three exponential models(mono-exponential,bi-exponential,and stretched exponential)can help discriminate malignant from benign breast lesions.The combined model of the parameters of the bi-exponential model had better diagnostic efficacy than the mono-exponential model.