PNN-based analysis system to classify renal pathologies in Kidney Ultrasound Images

Abstract

In this paper, a computer assistive tool is proposed to Process and analyse ultrasound Kidney Images for the classification of Renal Pathologies. The Ultrasound Kidney Images are classified into four classes: Normal, Cyst, Calculi and Tumor. Scanned Kidney Ultra-Sound (US) Images are obtained and Knowledge pertaining to common Pathologies from an Urologist Perspective is utilized as inputs to carry out the classification. The Images are preprocessed for the removal of Speckle noises by applying Median and Gaussian filter. Optimal thresholding segmentation algorithm is used to obtain the region of Interest. A set of first order statistical features are extracted. These features are given as inputs for training and testing the probabilistic neural network classifier. Hold out method is adopted where in 50% images are used for training and remaining 50% images are used for testing. The efficiency of the classifier is finally evaluated. A classification rate of 93.5% is obtained. The results achieved, are based on performance metrics calculations and are highly satisfactory.