Prostate ultrasound image analysis: localization of cancer lesions to assist biopsy

Abstract

Prostate cancer is the most commonly diagnosed cancer and second cause of cancer deaths in American men. Transrectal ultrasound imaging has been an acclaimed choice for systematic analysis of the internal architecture of the prostate gland and also for guiding the selection of biopsy tissue cores from suspicious lesions. The current accuracy of visual interpretation for identifying the presence of cancer from prostate ultrasound images is low. The present study addresses the statistical distribution of digital gray scale values to compare cancerous and noncancerous biopsy sites within a gland. Analyses of selected features from these data, for a limited number of cases, indicate significant differences between cancerous and non-cancerous biopsies with a retrospective biopsy classification accuracy of approximately 80%. This approach therefore shows promise for objective interpretation with improved accuracy.<>