前列腺癌局部复发影像报告系统对前列腺癌根治术后局部复发的诊断效能及观察者间评分一致性研究

Diagnostic performance and inter-observer consistency of prostate imaging recurrence reporting system in the detection of local recurrence after radical prostatectomy in patients with prostate cancer

目的评估前列腺癌局部复发影像报告系统(PI-RR)对前列腺癌根治性切除术(RP)后局部复发的诊断效能及不同年资医师间评分的一致性。方法本研究为横断面研究。回顾性收集2015年7月到2021年10月苏州大学附属第一医院RP后发生生化复发(BCR)而接受多参数MRI(mpMRI)的前列腺癌患者176例。6名不同年资放射科医师(2名低年资、2名高年资及2名专家医师)按照PI-RR评分标准分析mpMRI图像并对主病灶进行评分。基于穿刺病理结果、随访影像学检查及前列腺特异性抗原水平判定术后结局,分为局部复发组(54例)和未局部复发组(122例)。使用组内相关系数(ICC)和Kappa检验评估不同年资医师PI-RR评分的一致性。采用受试者操作特征(ROC)曲线分析不同年资医师PI-RR评分对前列腺癌RP术后局部复发的诊断效能,使用DeLong检验比较不同年资医师PI-RR评分的ROC曲线下面积(AUC),并对结果进行Benjamini-Hochberg法FDR校正。计算以PI-RR评分≥3或4分为阈值时的灵敏度、特异度。结果不同年资医师PI-RR评分ICC(95%CI)为0.70(0.64~0.76),专家与高年资医师PI-RR评分间Kappa值为0.528,专家与低年资医师间Kappa值为0.325,高年资与低年资医师间Kappa值为0.370。低年资、高年资及专家医师PI-RR评分诊断前列腺癌RP术后局部复发的AUC(95%CI)分别为0.73(0.65~0.81)、0.81(0.74~0.88)及0.86(0.80~0.93)。专家医师PI-RR评分的AUC高于高年资及低年资医师的AUC(Z=2.22、3.21,FDR校正P=0.039、0.003),高年资医师PI-RR评分AUC高于低年资医师(Z=2.22,FDR校正P=0.026)。以PI-RR评分≥3分为局部复发时,低年资医师、高年资医师和专家医师PI-RR评分的灵敏度分别为0.59、0.65和0.78,特异度分别为0.82、0.93和0.95;以PI-RR评分≥4分为局部复发时,低年资医师、高年资医师和专家医师PI-RR评分的灵敏度分别为0.50、0.54和0.69,特异度分别为0.88、0.96和0.97。结论PI-RR评分能够准确地诊断前列腺癌RP术后局部复发,不同年资医师PI-RR评分间一致性中等,诊断效能受阅片者经验的影响。

Objective To evaluate the diagnostic efficacy of prostate imaging recurrence reporting(PI-RR)system for detecting local recurrence after radical prostatectomy(RP)in prostate cancer(PCa)and to assess the consistency of the PI-RR scores assigned by different seniority radiologists.Methods This study was a cross-sectional study.A total of 176 PCa patients who underwent multi-parametric MRI(mpMRI)for biochemical recurrence(BCR)after RP from July 2015 to October 2021 at the First Affiliated Hospital of Soochow University were retrospectively collected.The mpMRI images were reviewed and the PI-RR scores of the main lesions were assigned independently by six different seniority radiologists(2 junior,2 senior and 2 expert radiologists).Following the reference standard determined by biopsy pathologic results,follow-up imaging,or prostate specific antigen levels,the patients were divided into two groups:54 patients with local recurrence and 122 patients without local recurrence.The intraclass correlation coefficient(ICC)and Kappa test were used to evaluate the consistency of the PI-RR scores by different seniority radiologists.The receiver operating characteristic(ROC)curve analysis was employed to evaluate the diagnostic efficacy of the PI-RR scores assessed by different seniority radiologists for detecting local recurrence of PCa after RP.The DeLong test was utilized to compare the areas under the ROC curve(AUC)of different seniority radiologist PI-RR scores and a false discovery rate(FDR)was applied to correct results using the Benjamini and Hochberg method.Sensitivity and specificity were calculated according to the cutoff value of PI-RR score≥3 or 4.Results The ICC(95%CI)of all different seniority radiologists was 0.70(0.64-0.76).The Kappa value was 0.528,0.325 and 0.370 respectively between expert and senior radiologists,expert and junior radiologists,senior and junior radiologists.The AUC(95%CI)of junior,senior,and expert radiologists were separately 0.73(0.65-0.81),0.81(0.74-0.88),and 0.86(0.80-0.93).The AUC of the expert radiologist PI-RR score was higher than those of senior and junior radiologist PI-RR scores(Z=2.22,3.21,FDR P=0.039,0.003).The PI-RR score of senior radiologist had higher AUC than that of junior radiologist(Z=2.22,FDR P=0.026).With the PI-RR score of 3 or greater as a cutoff value,the sensitivity of junior,senior and expert radiologists were respectively 0.59,0.65,and 0.78 and the specificity were 0.82,0.93,and 0.95.With the PI-RR score of 4 or greater as a cutoff value,the sensitivity of junior,senior and expert radiologists were respectively 0.50,0.54,and 0.69 and the specificity were 0.88,0.96 and 0.97.Conclusion PI-RR score can accurately diagnose local recurrence of PCa after RP.PI-RR score has a moderate inter-reader consistency across different seniority radiologists.And the diagnostic performance is influenced by the experience of radiologists.