PI-RR系统的诊断效能及观察者间一致性:以雄激素剥夺治疗后大切片为参照

Diagnostic efficiency and inter-observer agreement among readers with variable experience of PI-RR system:using whole-mount histology after androgen deprivation therapy as reference

目的:以雄激素剥夺治疗(androgen deprivation therapy,ADT)后前列腺全组织切片为参考标准,评估前列腺磁共振成像局部复发报告(Prostate Magnetic Resonance Imaging for Local Recurrence Reporting,PI-RR)的诊断效能及观察者间的一致性。方法:119例ADT后的前列腺癌患者术前行多参数MRI(multiparametric MRI,mpMRI)检查。由3名放射科医师独立分析MRI图像,根据PI-RR对影像学表现进行评分。在每个分区的基础上评估了PI-RP检测残留癌的诊断性能。采用χ^(2)检验比较不同阅片者的癌症检出率(cancer detection rate,CDR)。通过Cohen’s Kappa分析评估总体和成对的阅片者之间在PI-RR评分和评分≥3或4分的残留癌区域的一致性。结果:病理学切片发现209个区域有癌残留。以3分为截断值时,灵敏度、特异度、阳性预测值、阴性预测值分别为74.2%~83.7%、86.4%~92.7%、51.3%~64.3%、95.4%~96.9%;以4分为截断值时,灵敏度、特异度、阳性预测值、阴性预测值分别为47.4%~56.5%、97.9%~98.6%、82.5%~85.3%、91.6%~92.9%。不同阅片者间CDR差异无统计学意义。在PI-RR评分和残留癌区检测方面,所有阅片者的总体一致性为中等,但高、中年资阅片者之间的一致性(中等至显著)高于中、低年资阅片者之间的一致性(一般至中等)。结论:PI-RR评分可以准确地评估ADT后前列腺癌的复发情况,为前列腺癌放疗后的治疗提供潜在的参考;但阅片者的经验会影响诊断结果及阅片者之间的一致性。

Objective:To evaluate the diagnostic performance and interreader agreement of Prostate Magnetic Resonance Imaging for Local Recurrence Reporting(PI-RR)using whole-mount histology of prostate after androgen deprivation therapy(ADT)as the standard of reference.Methods:Totally 119 post-ADT prostate cancer patients underwent multiparametric MRI(mpMRI)before prostatectomy.Three radiologists analyzed MRI images independently,scoring imaging findings according to PIRR.Diagnosis performances for detection of residual cancer were assessed on per-sector basis.The cancer detection rate(CDR)was compared among readers withχ^(2) test.Overall and pairwise interreader agreement in assigning PI-RR scores and residual cancer sector with score≥3 or 4 were evaluated by Cohen’s Kappa analysis.Results:Two hundred and nine sectors with residual cancer were found on histology.The sensitivity,specificity,positive predictive value and negative predictive value at cutoff of score 3 ranged from 74.2%to 83.7%,86.4%to 92.7%,51.3%to 64.3%,and 95.4%to 96.9%,respectively,and at cutoff of 4,they ranged from 47.4%to 56.5%,97.9%to 98.6%,82.5%to 85.3%,and 91.6%to 92.9%,respectively.There was no significant difference among the CDR of readers.In PI-RR scores and detection of residual cancer sectors,overall interreader agreement was moderate for all readers,but agreement was higher between senior and intermediate readers(moderate to substantial)than between intermediate and junior readers(fair to moderate).Conclusion:MRI scoring with the PI-RR assessment provides accurate evaluation of prostate cancer after ADT,but readers’experience influenced cancer diagnosis and interreader agreement.