钼靶及MRI纹理分析技术在乳腺良恶性病灶的诊断价值

目的:探讨综合乳腺MRI多个序列、钼靶联合MRI纹理分析对乳腺良、恶性病灶的诊断价值.方法:搜集本院2014年6月-2019年12月116例乳腺钼靶图像(恶性57例,良性59例)及96例乳腺MRI(恶性45例,良性51例)的横轴面T2 WI脂肪抑制序列、DWI及增强后的第2、3期(C2、C3)图像进行回顾性纹理分析,其...

Diagnostic value of mammography and MRI texture analysis technique in benign and malignant breast lesions

Objective:To investigate the diagnostic value of multiple breast MRI sequences combined with mammography in breast benign and malignant lesions using texture analysis.Methods:One hundred and sixteen cases performed with mammography(57 malignant cases,29 benign cases)and 96 cases performed with breast MRI(45 malignant cases,51 benign cases)from June 2014 to December 2019 were included and analyzed in this study.Thirty patients(14 benign cases,16 malignant cases)underwent both mammography and MRI.The axial T 2WI with fat suppression,DWI and phase 2 and phase 3(C2 and C3)enhanced images were collected for MRI analysis.C.K.(CT kinetics,GE Healthcare)software was used to extract texture features.Rank sum test and independent sample t-test were used to explore the texture features with statistical difference between the two groups.The area under the curve(AUC)was calculated by receiver operating characteristic curve(ROC)analysis for each sequence.The threshold,sensitivity and specificity for diagnosing benign and malignant lesions were calculated using the maximum Youden index.Features with the highest AUC value extracted from mammography and each sequence of breast MRI were selected to construct a combined diagnostic model.The ROC curve of combined predictor obtained by logistic regression was also performed,and the AUC value was compared with former models.Results:The short-range gray-scale feature in the third-order texture features showed statistically significant between the two groups,with an AUC value of 0.66.The sensitivity and specificity were 86%and 35%respectively,using the cot-off value of short-range gray-scale parameters≤222264.00.The correlation in the second-order texture features extracted from T 2WI,C2 and C3 images and Haralick Prominence extracted from DWI images showed statistically significant between the two groups,with AUC values of 0.68,0.72,0.74 and 0.70,respectively.The Youden indexes were largest when choosing T 2WI≥3.55×10-5,the correlation of C2≥3.21×10-5,the correlation of C3≥5.30×10-5,and the Haralick correlation of DWI≥4.15×109,and the sensitivity and specificity were 76%and 55%,87%and 51%,80%and 65%,91%and 47%,respectively.The AUC values of multi-sequences based MRI model and mammography combined with multi-sequences based MRI model were 0.87(95%CI:0.63~1.00)and 0.91(95%CI:0.81~1.00),respectively.The sensitivity and specificity were 81%and 86%,88%and 86%respectively when the Yoden indexes were the largest with the cut-off values of 0.64 and 0.51.Conclusion:Combing the texture features of multiple sequences of breast MRI can improve the diagnostic efficiency for benign and malignant lesions.Mammography combined with multiple sequences breast outperformed mammography or breast MRI alone.