Abnormality Detection in the Renal Ultrasound Images using Ensemble MSVM Model

The renal or kidney disease is emerging as a common chronic disease similar to hypertension, diabetes, and cardiovascular disease. It is a silent killer disease which is to be monitored regularly to avoid serious complications. Therefore, this paper deals with the automatic detection of renal abnormalities using ensemble multiple-support vector machine (MSVM) classification model. The pre-trained convolutional neural networks (CNN) are used for automatic feature extraction from renal ultrasound (US) images. The extracted features are given to the ensemble MSVM model. This classifier model is designed by combining MSVM to get a quality classifier. The renal US images dataset is divided into normal and abnormal class. The proposed technique is validated with the total of 2085 images. On the evaluation of classifier based on 4-fold cross validation, average accuracy (ACC), sensitivity and specificity obtained are 89.53%, 90.81%, 88.02%, respectively, which is the highest ACC in comparison to the existing method.