Objective:

To compare the diagnostic efficacy of radiologists of different seniority in assessing clinically significant prostate cance

r (csPCa) based on the Prostate Imaging Reporting and Data System version 2.1 (PI-RADS v2.1) and to test the consistency of the scores.

Methods:

Three radiologists with different seniority (working years are 8

6

3 years respectively) retrospectively analyzed images of 583 patients who underwent multiparametric magnetic resonance imaging (mpMRI) before prostate biopsy. Receiver operating characteristic (ROC) curve was drawn from the obtained diagnostic results and pathological results to determine the best diagnostic performance of the three. Using the Kappa test to assess the consistency of diagnostic scores among three radiologists.

Results:

The area under the ROC curve (AUC) of the diagnostic efficacy of three radiologists with different seniority were 0.85

0.86

0.80. There was no significant difference in AUC between senior and middle-aged physicians (

P

=0.43). There was a statistically significant difference in AUC between senior and middle-aged radiologists and junior senior radiologists (

P

＜0.05). The sensitivity

specificity

positive predictive value and negative predictive value of high

middle

andlow-qualified physicians were 87%

82%

81%

88%

and 89%

82%

81%

90%

and 79%

82%

79%

82%. The consistency results of the comparison of diagnostic scores between senior and middle-qualified physicians was moderate (Kappa=0.70)

between middle-qualified and junior physicians was good (Kappa=0.76)

and between senior and junior-qualified physicians was moderate (Kappa=0.64).

Conclusion:

The diagnostic efficacy of PI-RADS v2.1 for csPCa is high among physicians of different levels of experience and its recommended for clinical application. The diagnostic efficacy of PI-RADS v2.1 differed among different years of experience of physicians

with senior and middle-qualified physicians outperforming junior physicians

but the consistency of their diagnoses was average

and PI-RADS v2.1 still needs further improvement.