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 / 中华放射学杂志

ABSTRACT

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.