基于FireVoxel软件MR-T_(2)WI纹理分析在前列腺癌诊断中的初步应用

Preliminary Application of MR-T_(2)WI Texture Analysis for Detection of Prostate Cancer Basing on Fire Voxel Software

目的基于FireVoxel软件探讨MR-T_(2)WI纹理分析在前列腺癌(PCa)诊断中的价值。方法搜集超声引导下前列腺靶点穿刺、临床病理确诊并于穿刺前后1个月内完成前列腺3.0 T MRI检查,且T_(2)WI图像上外周带存在低信号提示癌灶可能的51例患者。对前列腺外周带T_(2)WI低信号区域行基于FireVoxel软件纹理分析,分别计算Mean、Signal Stddev、Inhomogenity、Histogram Skewness、Histogram Kurtosis、Histogram Entropy这6个指标在诊断PCa及预测Gleason评分(GS)分组的受试者工作特征曲线(ROC)曲线下面积(AUC)。结果51例患者中外周带PCa 31例,良性增生20例;其中,中低GS组(Gleason评分≤8分)13例,高GS组(Gleason评分≥9分)18例。基于FireVoxel软件纹理分析Mean、Signal Stddev、Inhomogenity、Histogram Skewness、Histogram Kurtosis、Histogram Entropy指标在诊断PCa的AUC分别为0.711(P=0.011)、0.850(P=0.000)、0.405(P=0.255)、0.515(P=0.862)、0.640(P=0.093)、0.569(P=0.407)。进一步简化Signal Stddev指标诊断性能发现,当Signal Stddev<38诊断癌灶的AUC达0.846(P=0.000)。但各指标在预测中低GS组与高GS组的价值并不理想,AUC分别为0.483(P=0.107873)、0.581(P=0.447)、0.560(P=0.575)、0.675(P=0.101)、0.517(P=0.873)、0.577(P=0.471)。结论基于FireVoxel软件MR-T_(2)WI纹理分析Mean和Signal Stddev指标在没有增加扫描时间、机器配置等情况下能快速、简单、有效地提高T_(2)WI序列对PCa的诊断效能,值得临床特别是广大基层单位,推广应用。

Objective To explore the value of MR-T_(2)WI texture analysis for detection of prostate cancer(PCa)basing on FireVoxel software.Methods 51 patients with low signal on the peripheral zone of T_(2)WI image were involved,who completed ultrasound-guided prostate target puncture,clinicopathological diagnosis and prostate 3.0 T MRI examination which were completed within 1 month before and after the puncture.Basing on FireVoxel software texture analysis,the T_(2)WI low-signal area in the peripheral zone of the prostate was analysised.The Mean,Signal Stddev,Inhomogenity,Histogram Skewness,Histogram Kurtosis,Histogram Entropy were included into ayalysis.The area under the ROC curve of the six indicators in diagnosis PCa and predicted Gleason score group were calculated.Results There were 31 cases of PCa and 20 cases of benign hyperplasia in the 51 patients,including 13 in the low and medium GS group(Gleason score≤8)and 18 in the high GS group(Gleason score≥9)in PCa.Basing on the texture analysis of FireVoxel software,the aera of under the ROC curve of Mean,Signal Stddev,Inhomogenity,Histogram Skewness,Histogram Kurtosis,Histogram Entropy indicators for diagnosis of PCa were 0.711(P=0.011),0.850(P=0.000),0.405(P=0.255),0.515(P=0.862),0.640(P=0.093),0.569(P=0.407).To further simplify the diagnostic performance of Signal Stddev indicators,when the Signal Stddev is less than 38,the AUC for PCa diagnosis is 0.846(P=0.000).However,the value of each indicator in predicting GS group is not ideal.The AUC is 0.483(P=0.107873),0.581(P=0.447),0.560(P=0.575),0.675(P=0.101),0.517(P=0.873),0.577(P=0.471)respectively.Conclusion Without increasing scan time and machine configuration,Mean and Signal Stddev indicators basing on FireVoxel software MR-T_(2)WI texture analysis can quickly,simply and effectively improve the diagnostic efficiency of T_(2)WI for PCa detection.It is worthy for clinical promotion,especially for most primary units.