动态增强磁共振成像纹理分析对乳腺癌新辅助化疗效果的预测与评估

Efficacy prediction and evaluation of dynamic contrast-enhanced magnetic resonance imaging texture analysis in the neoadjuvant chemotherapy for breast cancer

目的探讨基于动态增强磁共振成像(DCE-MRI)的纹理分析对乳腺癌新辅助化疗疗效的预测与评估价值。方法回顾性分析山西省肿瘤医院2014年9月至2018年10月在新辅助化疗前后均行DCE-MRI且经病理确诊的63例乳腺癌患者的临床资料,根据病理检查结果,将患者分为治疗有效组(40例)和治疗无效组(23例),使用Omni-Kinetics软件测量新辅助化疗前及新辅助化疗4~8个周期后DCE-MRI容量转移常数(Ktrans)图的纹理参数,参数采用(±s)或中位数(P25,P75)表示,采用独立样本t检验或Mann-Whitney U检验对两组纹理参数进行统计分析,并绘制受试者工作特征曲线,根据曲线下面积(AUC)评估DCE-MRI Ktrans图的纹理参数对乳腺癌新辅助化疗疗效的预测效能。结果拟采用33个纹理参数,最终保留了29个。63例患者新辅助化疗前后有22个纹理参数差异均有统计学意义(均P<0.05);新辅助化疗前治疗有效组与治疗无效组间有9个纹理参数差异均有统计学意义(均P<0.05),包括均匀性[0.17(-0.06,0.34),0.39(0.22,0.48),Z=-2.955,P<0.01]、直方图能量[169.88(129.36,288.77),116.22(93.77,151.95),Z=3.241,P<0.01]及直方图熵[6.33(5.71,6.69),6.68(6.52,6.97),Z=-2.991,P<0.01]等;新辅助化疗后治疗有效组与治疗无效组间有8个纹理参数差异均有统计学意义(均P<0.05),包括直方图熵(6.00±0.71,6.46±0.49,t=-2.720,P<0.01)、熵(6.81±1.40,8.02±1.48,t=-3.238,P<0.01)、Haralick熵[0.49±0.10,0.55±0.10,Z=-2.613,P<0.01]、灰度不均匀性(GLN)[1.68(1.42,3.37),4.92(3.58,8.50),Z=-3.897,P<0.01]及游程长度不均匀性(RLN)[100.38(65.31,305.75),359.75(176.75,655.00),Z=-4.033,P<0.01]等;治疗有效组与治疗无效组新辅助化疗前后有8个参数变化率的差异均有统计学意义(均P<0.05),主要包括ΔGLN[-0.72(-0.78,-0.60),-0.23(-0.55,0.36),Z=-4.554,P<0.01]、ΔRLN[-0.71(-0.85,-0.52),-0.33(-0.48,-0.10),Z=-4.454,P<0.01]及Δ高灰度游程强度(HGLRE)[1.28(0.39,3.46),0.11(-0.24,0.86),Z=3.184,P<0.01]等。新辅助化疗后GLN、RLN、ΔGLN及ΔRLN的AUC分别为0.80、0.81、0.85及0.84。结论基于DCE-MRI Ktrans图的部分纹理参数对乳腺癌新辅助化疗的疗效有预测及评估作用。

Objective To investigate the efficacy prediction and evaluation value of neoadjuvant chemotherapy for breast cancer by using dynamic contrast-enhanced magnetic resonance imaging(DCE-MRI)texture analysis.Methods The clinical data of 63 patients with pathologically confirmed breast cancer in the Shanxi Provincial Cancer Hospital from September 2014 to October 2018 were retrospectively analyzed.All the patients underwent DCE-MRI before and after neoadjuvant chemotherapy and they were divided into the treatment-effective group(40 cases)and the treatment-ineffective group(23 cases)according to the postoperative pathological results.Texture parameters from volume transfer(Ktrans)maps of DCE-MRI before neoadjuvant chemotherapy and after 4-8 cycles of neoadjuvant chemotherapy were measured by using Omni-Kinetics software.The comparison of texture parameters between the two groups was performed by using independent sample t test or Mann-Whitney U test.The receiver operating characteristic curve was drawn and the prediction efficiency of these texture parameters in the therapeutic efficacy of neoadjuvant chemotherapy for breast cancer according to the corresponding area under the curve(AUC)was evaluated.Results A total of 33 texture parameters were enrolled,and finally 29 texture parameters were retained.Before and after neoadjuvant chemotherapy 22 texture parameters had statistically significant difference in 63 patients(all P<0.05).There was a statistically significant difference in 9 texture parameters between the two groups before neoadjuvant chemotherapy(all P<0.05),including uniformity[0.17(-0.06,0.34),0.39(0.22,0.48),Z=-2.955,P<0.01],histogram energy[169.88(129.36,288.77),116.22(93.77,151.95),Z=3.241,P<0.01]and histogram entropy[6.33(5.71,6.69),6.68(6.52,6.97),Z=-2.991,P<0.01].After neoadjuvant chemotherapy,8 of the 29 texture parameters between the two groups had statistically significant differences(all P<0.05),including histogram entropy(6.00±0.71,6.46±0.49,t=-2.720,P<0.01),entropy(6.81±1.40,8.02±1.48,t=-3.238,P<0.01),Haralick entropy[0.49±0.10,0.55±0.10,Z=-2.613,P<0.01],grey level non-uniformity(GLN)[1.68(1.42,3.37),4.92(3.58,8.50),Z=-3.897,P<0.01],run length non-uniformity(RLN)[100.38(65.31,305.75),359.75(176.75,655.00),Z=-4.033,P<0.01].There were statistical differences in 8 parameters change rate before and after neoadjuvant chemotherapy between the two groups(all P<0.05),mainly includingΔGLN[-0.72(-0.78,-0.60),-0.23(-0.55,0.36),Z=-4.554,P<0.01],ΔRLN[-0.71(-0.85,-0.52),-0.33(-0.48,-0.10),Z=-4.454,P<0.01],Δhigh grey level run emphasis(HGLRE)[1.28(0.39,3.46),0.11(-0.24,0.86),Z=3.184,P<0.01].According to the ROC curve,AUC of GLN,RLN,ΔGLN andΔRLN after neoadjuvant chemotherapy was 0.80,0.81,0.85 and 0.84,respectively.Conclusion Some texture parameters obtained from DCE-MRI Ktrans map can predict and evaluate the efficacy of neoadjuvant chemotherapy in breast cancer.