• Dae Chul Jung, Hak Jong Lee, Seung Hyup Kim, Ghee Young Choe, and Sang Eun Lee.
• Department of Radiology, Research Institute and Hospital, National Cancer Center, Goyang, Korea.
• J Magn Reson Imaging. 2008 Jul 1; 28 (1): 144-50.

PurposeTo develop an image-based classification system for seminal vesicle lesions (SVL) to detect seminal vesicle invasion (SVI) in prostate cancer and to evaluate whether pattern analysis of SVL with MR imaging could improve the accuracy of evaluating SVI.Materials And MethodsThe MR images of 217 patients who had undergone retropubic radical prostatectomy (RRP) due to prostate cancer were retrospectively analyzed by two uroradiologists, focusing on SVL. The SVL on T2-weighted sequences was classified as five classes. In each class the results were correlated with histopathologic findings.ResultsFourteen (6.5%) of 217 patients had evidence of SVI at histopathologic evaluation after RRP. Pattern analysis of SVL showed 71.4% sensitivity, 96.6% specificity for predicting SVI. ROC curves for the subjective scoring of SVI showed that reader 1 had an area under the curve (AUC) of 0.69 and reader 2 had an AUC of 0.81. The overall accuracy of pattern analysis was superior to both serum prostate-specific antigen (PSA) level and subjective scoring (P < 0.01, McNemar test).ConclusionThe classification system of SVL on the basis of their imaged morphologic features can provide an objective standard and simplify the abnormal findings to help predict SVI in prostate cancer. Pattern analysis of SVL with MR imaging improves the accuracy of detecting SVI.(c) 2008 Wiley-Liss, Inc.

13 keywords...

hide…