An enhanced ultrasound kidney image classification for health care. R Vinoth 1 , K Bommannaraja 2* 1 Department of Electronics and Communication Engineering, Mahendra Engineering College, Tamil Nadu, India 2 Department of Electronics and Communication Engineering, KPR Institute of Engineering and Technology, Tamil Nadu, India Abstract An enhanced medical decision support system for classification of Ultrasound Kidney images is developed for health care and presented in this paper. The image enhancement was done by removing the speckle, salt and pepper noises using fuzzy c means filtering and the Gray Level Coocurrence Matrix was obtained for feature extraction. Gabor wavelets and Histogram equalization were used for the selection of texture features. The classification is done using SVM, ANN, K-NN and Hybrid classifiers and the accuracy of classification was found to be 99.6% for the SVM- ANN hybrid classifier. The developed system is expected to provide support for the medical practitioners for decision making to provide an enhanced health care. Keywords: Gray level coocurrence matrix, SVM classifier, ANN Classifier, KNN Classifier, Ultrasound kidney image, Health care. Accepted on April 14, 2016 Introduction In the past few decades, medical imaging and associated systems are found to play vital role in developing an accurate computer assisted Medical Decision Support System for the clinical practitioners for better health care. The development in the soft computing techniques further motivated the research to use them for developing more decision support systems. It has been witnessed that the Ultrasound (US) image of kidney is preferred by most of the physicians for diagnosis of kidney. It was also reported that identification of the kidney disease from the US image is considered to be the challenging task due to inherent limitations. With the development in the image processing tools, the classification of US kidney has become accurate and preferred. Many researchers have reported various techniques to overcome these limitations and provided variety of solutions for classification of abnormalities in the kidneys. Leavline and Singh [1] reported the impulse noise removal using the standard median filtering technique and its variants were analyzed. Simulations were carried out on a set of standard gray scale images and the state of the art median filter variants were compared in terms of the well known image quality assessment metrics namely Mean Square Error (MSE) and Peak Signal to Noise Ratio (PSNR). Kenny and Nor [2] also proposed a novel method for the removal of noise from images. The Cluster based Adaptive Fuzzy Switching Median (CAFSM) filter was found to have the capability in handling realistic impulse noise model for real world applications and easy to implement the filtering phase replaces the detected noise pixels in the image. Shruthi et al. [3] proposed a despeckling filter scheme based on Gabor wavelet in order to enhance the quality of image by reducing the speckle noises. It was showed that higher the PSNR values lower the MSE value which indicates that more noise is removed. Joseph et al. [4] also proposed a new weighted linear filtering approach using Local Binary Patterns (LBP) for reducing the speckle noise in ultrasound images. It was reported that improved results can be achieved by reducing the noise without affecting the image content. It was found that the developed weighted linear filter performs better then filters in terms of quantitative analysis and edge preservation. Onder and Karacali [5] developed to perform the automated texture classification of histology slides using grayscale images and manifold learning method. Texture feature vectors were obtained using local gray scale co-occurrence matrices and the dimension of the feature vector space was lowered using Isomap dimension reduction. In a lower dimension feature space, k-means clustering operation was performed in order to provide separate texture clusters. Raja et al. [6] reported the classification of US kidney images using dominant Gabor wavelet. The Gabor wavelet was determined by maximizing the similarity between original preprocessed image and reconstructed Gabor image. It was observed that, the dominant Gabor wavelet improves the classification efficiency appreciably. The possibility of implementing a computer aided diagnosis system exclusively for US kidney images were explored. ISSN 0970-938X www.biomedres.info Biomed Res- India 2016 Special Issue S46 Biomedical Research 2016; Special Issue: S46-S52 Special Section:Computational Life Sciences and Smarter Technological Advancement