Analysis of three-dimensional Doppler ultrasonographic quantitative measures for the discrimination of prostate cancer

Objective. The purpose of this study was to determine whether several quant itative ultrasonographic measures have potential to discriminate prostate c ancer from normal prostate and to determine the best combination of these m easures. The true spatial distributions of cancer within the prostates stud ied were obtained histologically after radical prostatectomy The relationsh ip between Doppler ultrasonography and microvessel count was also investiga ted. Methods. Three-dimensional Doppler ultrasonographic data were acquired from 39 patients before radical prostatectomy The removed prostate was sec tioned, and whole-mount hematoxylineosin-stained slides were used to identi fy all regions of cancer within each prostate. These histologic and ultraso nographic data were spatially registered. Doppler ultrasonographic measures were calculated within uniformly sized three-dimensional regions that were either entirely cancerous or noncancerous, and receiver operating characte ristic analysis was performed on the results. Microvessel counts were made within each contiguous cancerous region and correlated with ultrasonographi c measures. Results. Color pixel density was the best simple measure for di scriminating prostate cancer (accuracy, 80%). The mean power mode value (no rmalized mean power in color pixels) was inversely related to cancer with a n accuracy of 1 - normalized mean power in color pixels = 65% (low mean pow er is more cancerous). When color pixel density was combined with the norma lized mean power in color pixels, its accuracy improved slightly to 84%. Th e peak microvessel count had a negative correlation with color pixel densit y as well as with cancer stage. Conclusion. Doppler ultrasonography does pr ovide discriminatory information for prostate cancer, with color pixel dens ity being the most promising measure.