| Abstract: |
Transrectal biopsies of the prostate are guided using B-mode ultrasound. About 30% of initial biopsies in men suspected of having prostate cancer are positive. A slightly lower incidence of positive biopsies occurs in repeat examinations of men having negative initial biopsies. These rates of detection imply that the sensitivity of conventional ultrasound-guided procedures is between 40% and 50%. With about 330,000 new cases of prostate cancer detected annually in the US, these figures suggest that more than 600,000 actually positive glands are biopsied at least once, and that even with repeat biopsies, almost 300,000 cancers go undetected; i.e., produce false-negative results. The figures also suggest that well over 1,000,000 biopsy examinations are performed annually, which indicates that the biopsies of about 400,000 men produce true-negative results and hence, in hindsight, were unnecessary. We have acquired ultrasonic and biopsy data from more than 250 patients. We have compared the classification efficacy of spectrum analysis of the radiofrequency (RF) echo signals (present, but not utilized, in conventional ultrasound instruments) with B-mode imaging currently used for biopsy guidance. Our latest comparison of spectrum-analysis results with B-mode image-interpretation data from 96 patients and 448 biopsies produced relative operating characteristic (ROC) curve areas of 0.75 for spectrum analysis and 0.58 for B-mode biopsy guidance; the gold standard for these comparisons was the histology findings of the 448 biopsies. Our results indicate that biopsy guidance based on spectrum analysis can improve sensitivity and cancer detection by approximately 50%; i.e., increasing the number of cancers detected annually in the US to more than 450,000. Similarly, spectrum analysis potentially can aid in evaluating and monitoring detected cancers by helping to assess the number, location, and volume of lesions and by providing a quantitative means of determining the nature and rate of change in lesion properties. Obvious applications are guiding therapy (brachytherapy, cryotherapy, etc.), determining whether surgery is warranted, assisting in watchful waiting, titrating treatment, and assessing responses to neoadjuvant therapy. Spectral methods can be incorporated into clinically useful images. Such images are digitally generated from RF signals by computing spectra at each pixel location. Pixel values in the resulting image may depict spectral parameters, tissue properties (such as scatterer sizes or concentrations), or on the basis of comparisons with database values, either most likely tissue type or level of suspicion (LOS). Either two- or three-dimensional presentations can be generated; 3D renderings can be formed from sets of component 2D images. In 2D, any of these formats, but particularly the LOS format, could be used for more effective biopsy guidance; in two- or three- dimensions, they could be used for lesion assessment. |
