The validity and agreement of PI-RADS v2 in the diagnosis of prostate cancer

Authors : Şehnaz TEZCAN, Ülkü BEKAR, Müge ONBAŞIOĞLU GÜRBÜZ, Giray ERGİN

Pages : 303-309

Doi:10.38053/acmj.977881

View : 24 | Download : 11

Publication Date : 2021-09-24

Article Type : Research Paper

Abstract :Aim: The purpose of this study was to evaluate accuracy of multi-parametric MRI (mpMRI) in detection of clinically significant (CS) prostate cancer (PC) and determine agreement of Prostate Imaging Reporting and Data Systems version2 (PI-RADS v2) among three readers. Material and Method: The study included 65 (32 malignancy, 33 benign) patients with clinically suspected PC who were underwent mpMRI between January 2017 and January 2020 followed by biopsy or prostatectomy. The images were evaluated by three readers who were blinded to patient data. The inter-observer agreement was analyzed with Cohen’s weighted kappa statistics. Results: 74 lesions were detected in 46 patients among 65 patients. When a PI-RADS assessment category ≥3 (K value, 0.406-0.632) was considered positive for CS PC for readers, higher sensitivity, lower specificity and lower agreement was found than PI-RADS ≥4 (K value, 0.545-0.667). The sensitivity and specificity of index lesion detection ranged from 71.8%-90.6%, 60.6%-72.7%, respectively. We found moderate to substantial agreement for index lesion detection. The agreement of PZ lesions was higher than TZ lesions. The agreement in DWI scores was higher than the agreement in T2 scores between readers. Conclusion: By using PI-RADS v2, high sensitivity but moderate specificity was found in detection of index lesion. The agreement in PI-RADS category assignment was moderate among readers. The agreement and sensitivity in threshold of PI-RADS 4 was higher than PI-RADS 3. TZ lesions showed more variability among radiologists than PZ lesions by using PI-RADS v2.
Keywords : the Prostate Imaging Reporting and Data Systems version 2, prostate cancer, inter observer agreement