Rough Wavelet Hybrid Image Classification Scheme

This paper introduces a new computer-aided classification system for detection of prostate cancer in Transrectal Ultrasound images (TRUS). To increase the efficiency of the computer aided classification process, an intensity adjustment process is applied first, based on the Pulse Coupled Neural Network (PCNN) with a median filter. This is followed by applying a PCNN-based segmentation algorithm to detect the boundary of the prostate image. Combining the adjustment and segmentation enable to eliminate PCNN sensitivity to the setting of the various PCNN parameters whose optimal selection can be difficult and can vary even for the same problem. Then, wavelet based features have been extracted and normalized, followed by application of a rough set analysis to discover the dependency between the attributes and to generate a set of reduct that contains a minimal number of attributes. Finally, a rough confusion matrix is designed that contain information about actual and predicted classifications done by a classification system. Experimental results show that the introduced system is very successful and has high detection accuracy.