Feature Extraction and Classification of Digital Kidney Ultrasound Images: A Hybrid Approach

Abstract

Ultrasound image is widely used medical imaging modality for the diagnosis of diseases. In this work, a novel hybrid approach for classification of diseased kidney medical ultrasound images is proposed. The segmented images are passed through the stage of feature extraction. Different features, namely, Haralick, shape, wavelet, Tamura, and histogram oriented gradient features are used for classification. The method focuses on recognizing the most dominant feature set that characterizes the texture of a stone, cyst or normal kidney ultrasound image. Extracted features are individually used for classification by three different classifiers namely k-nearest neighborhood, fuzzy k-nearest neighborhood, and support vector machine. The hybrid approach for feature extraction containing the combination of wavelet and shape features shows the optimal classification accuracy. The efficiency of the method is tested with performance parameters such as accuracy, sensitivity, and specificity. The results obtained show the efficiency of the proposed method.