The University of British Columbia

In this paper we describe a new approach to tissue characterization for detection of prostate cancer. We propose that if a specific location in the prostate tissue undergoes continuous interactions with ultrasound, the time series of RF echo signals from that location would carry "tissue characterizing" information. This phenomenon is due to different microstructures of normal and cancerous tissues. We use Higuchi's methodology to compute the fractal dimension of RF echo time series as a measure of the complexity. Averaged fractal dimension over a region of interest of the prostate tissue is utilized as the sole tissue characterizing feature and applied along with a Bayesian classifier. The results are validated based on detailed histopathologic maps of malignancy. The area under ROC curve is 0.894 and accuracies of up to 86% are acquired, indicating the effectiveness of our tissue characterization approach based on the fractal analysis of RF time series