Purpose: To prospectively investigate the local staging accuracy of 3-T endorectal magnetic resonance (MR) imaging for prostate cancer by using whole-mount-section histopathologic analysis as the standard of reference.
Materials and methods: This study was approved by the institutional review board, and informed consent was obtained from all patients. In 35 consecutive patients (median age, 62.3 years) with biopsy-proved prostate cancer, 3-T endorectal MR imaging was performed. High-spatial-resolution endorectal T2-weighted fast spin-echo images of the prostate were obtained in three planes. MR images were prospectively evaluated by two experienced radiologists and a third radiologist who was less experienced with regard to local disease extent by using five established extracapsular criteria. Whole-mount-section histopathologic analysis was the standard of reference. Evaluation was performed according to octant and patient. Sensitivity, specificity, positive and negative predictive values, overall accuracy, and interobserver agreement were calculated.
Results: Thirty-two patients who underwent radical prostatectomy were enrolled in this study. Accuracy, sensitivity, and specificity of local staging were 94% (30 of 32), 88% (seven of eight), and 96% (23 of 24), respectively, for both experienced radiologists, and these values were 81% (26 of 32), 50% (four of eight), and 92% (22 of 24), respectively, for the less experienced radiologist. There was substantial agreement between both experienced readers (kappa = 0.42-0.79) and moderate agreement between the less experienced reader and the experienced readers with respect to all extracapsular criteria. In regard to the three cases of minimal capsular invasion, two were detected by both experienced radiologists.
Conclusion: In this study, high accuracy for staging of prostate cancer at 3-T endorectal MR imaging, with moderate to substantial observer agreement, was demonstrated. In addition, minimal capsular invasion could be detected.
RSNA, 2005.