Purpose: To retrospectively assess the use of the European Society of Urogenital Radiology (ESUR) Prostate Imaging Reporting and Data System (PI-RADS) criteria and 3-T multiparametric magnetic resonance (MR) imaging for detection of extraprostatic extension (EPE) of prostate cancer.
Materials and methods: The institutional review board approval requirement was waived. Consecutive patients with prostate cancer (n = 133) underwent 3-T multiparametric MR imaging before prostatectomy. Lesions were assessed by using ESUR/PI-RADS criteria for T2-weighted, diffusion-weighted, and dynamic contrast material-enhanced imaging, and by using the sum of these scores. Zonal dominant parameters corresponding to the score of diffusion-weighted imaging for peripheral zone lesions and to T2-weighted imaging scores for transitional zone lesions were calculated. In addition, the presence of EPE in each patient was evaluated on the basis of subjective multiparametric MR imaging features. Histopathologic examination of whole-mount radical prostatectomy specimens was used as the reference standard. Sensitivity, specificity, positive predictive, and negative predictive values; likelihood ratios; and areas under the receiver operating characteristic curve were calculated for each parameter on the basis of its usefulness for prediction of EPE.
Results: EPE was found in 60 of 133 (45%) patients. Receiver operating characteristic curve analysis for the prediction of EPE revealed an area under the curve of 0.72 for T2-weighted, 0.67 for diffusion-weighted, and 0.64 for dynamic contrast-enhanced imaging; 0.74 for the dominant parameter; and 0.74 for the sum of the PI-RADS scores, and a score of 5 was defined as the best threshold for the individual parameters, with a score greater than or equal to 13 as the threshold for the sum of the PI-RADS scores. By applying these thresholds, sensitivity, negative predictive value, and negative likelihood ratio (ruling out EPE) were 77%, 77%, and 0.36, respectively, and specificity, positive predictive value, and positive likelihood ratio (ruling in EPE) were 64%, 64%, and 2.15, respectively, for the dominant parameter. Feature analysis showed an area under the curve of 0.72; sensitivity, negative predictive value, and negative likelihood ratio of 63%, 72%, and 0.56, respectively, and specificity, positive predictive value, and positive likelihood ratio of 78%, 70%, and 3.77, respectively.
Conclusion: ESUR/PI-RADS criteria showed moderate overall accuracy for use in the prediction of EPE, and these results were similar to those of multiparametric MR imaging assessment of features in this study sample.
(©) RSNA, 2015