Objective: To evaluate the potential benefit of semiautomated localization of prostate cancer using clustering analysis on three-dimensional (3-D) contrast-enhanced power Doppler images.
Methods: Thirty patients with biopsy-proven prostate cancer and scheduled for radical prostatectomy underwent a 3-D contrast-enhanced power Doppler scan prior to surgery. A 3-D ellipsoid model was manually fitted around the prostate. The model automatically divided the prostate into 12 zones. After calculation of a so-called clustering map, the clustering values of each zone were calculated. They were compared with whole-mount section histopathology. Region-of-interest (ROI) analysis was performed with bootstrapping to evaluate overall performance.
Results: The ROI analysis yielded area under the curve (AUC) values of 0.65 with a corresponding standard error of 0.03.
Conclusion: Semiautomatic localization based on clustering analysis of blood flow aids in localization of prostate tumors. A clustering map is an easy-to-interpret extension to standard power Doppler images.