Magnetic resonance imaging of clinically localized prostatic cancer.

PURPOSE: We assess the accuracy of endorectal coil magnetic resonance imaging (MRI) for detecting tumor localization, capsular penetration and seminal vesicle invasion in clinically organ confined prostate cancer. We also evaluate intra-observer and interobserver agreement in interpreting MRI studies.
MATERIALS AND METHODS: MRI studies of 51 consecutive patients a mean of 61 years old with biopsy proved prostate cancer were retrospectively read twice by 2 radiologists in random order. Both radiologists marked tumor localization, capsular penetration and seminal vesicle invasion on standard tumor maps. These findings were compared with the histopathological results of radical prostatectomy specimens.
RESULTS: The overall accuracy of detecting cancer localization was 61%. The detection rate for cancer foci less than 5 mm. was only 5% but for lesions greater than 10 mm. it was 89%. There was 91 and 80% accuracy for detecting capsular penetration and seminal vesicle invasion, respectively. Sensitivity and specificity were 60 and 63, 13 and 97, and 59 and 84% for localization, capsular penetration and seminal vesicle invasion, respectively. Intra-observer and interobserver agreement ranged from fair to good (kappa coefficient 0.240 to 0.647).
CONCLUSIONS: Endorectal MRI seems to be better than previously reported for detecting seminal vesicle invasion and tumor foci in the anterior half of the prostate. Sensitivity in detecting minor capsular penetration of the tumor was low, which can probably be improved by methodological development. MRI may be useful for locating cancer foci in patients with high prostate specific antigen values but repeatedly negative biopsy findings.