摘要
目的采用前列腺影像报告与数据系统(PI-RADS v2)评估不同年资医师间评分一致性的主要影响因素。方法回顾性分析60例前列腺癌(PCa)患者MRI图像,其中Gleason评分3+3、3+4、4+3、4+4、4+5、5+5每组各10例;5名影像诊断经验分别为21(R1)、10(R2)、5(R3)、3(R4)、1(R5)年的医师作为观察者。根据PI-RADS v2标准,培训5名医师并独立评分,R1独立重复一次评分并结合病理,将其作为评分参考标准;5名医师独立评估熟悉程度与是否遵从双盲原则自我评分;采用Kappa检验两医师间一致性;采用Kruskal-Wallis检验5名医师整体是否存在差异,采用Mann-Whitney两两比较5名医师的差异;采用受试者工作特征曲线(ROC)分析不同年资医师的诊断效能,计算曲线下面积(AUC)及95%可信区间(95%CI)。结果 60例MRI资料共计1500幅有效图像,涉及责任病灶共计431幅。5名观察者中熟悉程度评分均较高,双盲依从性较好;5名医师与最后诊断结果的Kappa一致性检验显示,R1、R2、R3的Kappa系数〉0.75,R4(K=0.607)与R5(K=0.584)的诊断一致性良好;R1前后两次评分无统计学差异(P=0.759);医师间评分存在差异(H=2.643,P=0.027),医师年资越高,评分一致性越好;低年资医师评分(R4与R5)差异不明显,但与R1评分存在显著差异;不同医师主观评分间的AUC随着医师的年资降低逐渐降低,R5的AUC最低。结论观察者间评分差异的来源是多样的,工作年资是影响主观评分准确性的主要因素。
Objective To evaluate the repeatability of inter-observers using Prostate Imaging Reporting and Data System Vision 2( PI-RADS v2) score system. Methods Sixty cases were included in this retrospective study. All of the subjects underwent prostate imaging at 3 T MRI scanner before radical surgery. The inter-agreement of observers using PIRADS score system was then analysed. Standardized "Likert"scales from 1 to 5 were used to evaluate the imaging,the scores and details of the prostate had been prospectively interpreted by 5 radiologists. The experience of 5 readers range from1 to 21 years; scores of PI-RADS in readers were applied using Kruskal-Wallis test to perform statistical analysis. MannWhitney and Kappa test were calculated to detect the significant difference of inter-observers; ROC curve and AUC were performed to evaluate the diagnostic performance of inter-readers. Results Of the 60 cases with prostate cancer; 1500 high quality images were analyzed. Five readers had good agreement( H = 2. 643,P = 0. 027). There was insignificant inter-group difference between pre-and post-scoring in the reader with 21 years ' experience( P = 0. 759). Insignificant inter-group difference for both R4 and R5( P = 0. 127),and Kappa test was observed between R4( K = 0. 607),and R5( K =0. 584),respectively. The diagnostic performance of readers was observed with the PI-RADS score and became excellent( AUC 0. 882). The ROC curve show that R1、R2、R3、R4 and R5 have high value( AUC 0. 882),R5 have lower value( AUC = 0. 882,0. 784-0. 980% 95 CI). Conclusion Good agreement of inter-observers and systematic study of the prostate enable a reliable persuasion of scientific paper.
作者
邓明
徐海波
胡金香
张在鹏
徐国斌
梅列军
龙清云
李欢
廖如芳
周舟
DENG Ming;XUHaibo;HUJinxiang(Department of Radiology,Zhongnan Hospital of Wuhan University,Wuhan,430071,P.R.China)
出处
《临床放射学杂志》
CSCD
北大核心
2018年第8期1402-1405,共4页
Journal of Clinical Radiology
基金
国家自然科学基金资助项目(编号:81372369)
关键词
观察者
一致性
前列腺
对策
磁共振成像
Observers
Consistency
Prostate
Strategy
Magnetic resonance imaging