Spectrum analysis for classifying and evaluating prostate tissue
In 1996, approximately 1,000,000 biopsies are performed for prostate cancer, over 300,000 new cases are detected, and over 40,000 men die from this disease in the USA. Most biopsies prove to be negative, and about 1/3 of those are false; i.e., many biopsies prove to be unnecessary or miss cancer that is present. Spectrum analysis of radio-frequency ultrasonic echo signals shows useful differences between spectral parameters of cancerous and non-cancerous prostate tissue. Our results are based on over 100 histologically classified patients, and give an ROC-curve area of 79% for cancer detection using spectrum analysis vs. 60% for conventional imaging. We can depict spectral-parameter values in 2- or 3-D images. Real-time 2-D parameter images promise to reduce false-negative biopsies by allowing better biopsy guidance and can reduce costs and risk of true-negative biopsies though improved imaging specificity. 3-D parameter images may improve volumetric disease evaluation, and thereby, treatment planning and therapy monitoring. These methods are based on an empirically validated theoretical framework that relates microscopic scatterer properties to spectral-parameter values. This research is supported in part by US DHHS NIH/NCI Grant CA53561.