Localisation of Zones of Cancer Detection in Prostate Gland Using Ratio Matrix and Radial Scanning of 2D Trans-rectal Ultrasound Images;

Abstract

Researchers have continued to proffer various solutions to the challenge of delineating from Trans-rectal<br /> ultrasound (TRUS) 2D-images of the prostate the regions of desired property. This paper presents an algorithm<br /> that categorises the detected regions suspected to be cancerous, hyper-echoic pixels, in the prostate gland from<br /> a 2D Trans-rectal Ultrasound images into three zones. The developed algorithm uses radial scanning of the pixels<br /> of the prostate gland image from common seed point both to detect and delineate the suspected cancerous<br /> pixels into zones, namely peripheral, transition and central, by applying ratios of the anatomical zones of the<br /> prostate gland. Expert knowledge, intensity and gradient features were implemented to delineate regions of<br /> interest. MATLAB programming tool was used for creating the codes that implemented the algorithms. Samples<br /> of TRUS 2D-images of the prostate for patients with raised PSA values (>10 ng/ml) used in a previous work by<br /> Award (2007) were used for testing the algorithm. The test results showed that the algorithm could detect zones<br /> of the prostate boundary exhibit image properties for cancer cells and also the percentage of malignancy<br /> detected in zones agreed with existing research findings. Comparison of detection results with that of an expert<br /> radiologist yielded the following performance parameters; accuracy of 88.55% and sensitivity of 71.65%.