弥散加权成像模型鉴别乳腺良恶性病变

Diffusion-weighted imaging models for distinguishing benign and malignant breast lesions

目的观察弥散加权成像(DWI)单指数模型、体素内不相干运动(IVIM)和弥散峰度成像(DKI)模型鉴别乳腺良、恶性病变的价值。方法对临床疑诊乳腺病变的202例患者(215个病灶)采集乳腺DWI,b值取0、30、50、80、120、160、200、500、1000、1500及2000 s/mm^(2)。采用单指数模型、IVIM模型、DKI模型分析原始数据,比较乳腺良、恶性病变表观扩散系数(ADC)、IVIM参数灌注分数(IVIM-FP)、真实弥散系数(IVIM-D)及灌注相关弥散系数(IVIM-DP)和DKI参数平均峰度(DKI-K)及平均弥散率(DKI-D)的差异。将单因素分析结果显示P<0.10的变量纳入多因素逐步logistic回归分析,筛选鉴别乳腺良、恶性病变的最佳参数,并建立联合模型。以病理结果为标准,采用受试者工作特征(ROC)曲线,计算曲线下面积(AUC),分析单一及联合参数鉴别乳腺良、恶性病变的效能。结果良性组54例(63个病灶),恶性组148例(152个病灶),组间ADC、IVIM-D、DKI-K和DKI-D差异均有统计学意义(P<0.001),IVIM-FP及IVIM-DP差异无统计学意义(P均>0.05)。ADC+DKI-K联合为鉴别乳腺良、恶性病变的最佳联合参数。ADC诊断乳腺良、恶性病变的效能最高,敏感度91.45%,特异度82.54%、准确率88.84%;其AUC 0.92,高于DKI-K及IVIM-DP(Z=4.72、6.78,P均<0.01),与IVIM-D、DKI-D、ADC+DKI-K差异均无统计学意义(Z=0.64、1.34、1.11,P=0.52、0.18、0.27)。结论DWI单指数模型、IVIM及DKI模型对鉴别乳腺良、恶性病变均有较高价值;DWI单一参数中,ADC诊断效能最佳,与联合参数模型相当。

Objective To explore the value of single index model,models of intravoxel incoherent motion(IVIM)and diffusional kurtosis imaging(DKI)of diffusion-weighted imaging(DWI)for differentiating benign and malignant breast lesions.Methods Totally 202 patients with 215 clinically suspicious breast lesions underwent breast MR DWI with b values of 0,30,50,80,120,160,200,500,1000,1500 and 2000 s/mm^(2).Single index model and IVIM and DKI models were used to analyze the original data.The apparent diffusion coefficient(ADC),perfusion fraction(IVIM-FP),diffusion coefficient(IVIM-D),pseudo-diffusion coefficient(IVIM-DP)of IVIM parameter,as well as mean kurtosis(DKI-K),mean diffusivity(DKI-D)of DKI parameter of benign and malignant lesions were compared.Variables with P<0.10 in univariate analysis were included in multivariate stepwise logistic regression analysis to select the optimal combined parameters for differentiating benign and malignant breast lesions,then the combined model was established.Based on pathological results,receiver operating characteristic(ROC)curves were used to analyze the diagnostic efficacy of single and combined parameters for differentiating benign and malignant breast lesions,and the areas under the curve(AUC)were calculated.Results There were 54 patients with 63 benign lesions(benign group)and 148 cases with 152 malignant lesions(malignant group).Significant differences of ADC,IVIM-D,DKI-K and DKI-D were detected between groups(all P<0.001),while no significant difference of IVIM-FP nor IVIM-DP was found between groups(both P>0.05).The combination of ADC and DKI-K was the best combined parameters for differentiating benign and malignant breast lesions.ADC was the most effective index for diagnosing benign and malignant breast lesions,with the sensitivity of 91.45%,specificity of 82.54%,accuracy of 88.84%and AUC of 0.92,higher than that of DKI-K and IVIM-DP(Z=4.72,6.78,both P<0.01).No statistical difference of AUC was found between ADC and IVIM-D,DKI-D,ADC+DKI-K parameters(Z=0.64,1.34,1.11,P=0.52,0.18,0.27).Conclusion DWI single index model,IVIM and DKI models all had high diagnostic value for differentiating benign and malignant breast lesions.Among single parameters of DWI,the diagnostic performance of ADC was the best,which was comparable to that of the combined parameters model.