Automatic Detection and Classification of Chronic Kidney Diseases Using CNN Architecture

Abstract

In medical imaging field, ultrasound imaging technique provides a tremendous service on disease recognition. It assists as non-catheterization, hazardous emission-free diagnosis at less expenses of cost. The main aim of this investigation is to detect and recognize the chronic kidney diseases CKD makes initiation and home for dysfunction of several organs of the homo sapiens. It may induce the heart disease, ischemic attack, cardiomyopathy, and cardiac disease through the hypertension. The early prediction of the kidney deformation would save the life from dreadful diseases. Renal sonography is the basic imaging technology used for kidney disease diagnosis. In this work, automatic detection and classification of kidney diseases using deep convolutional architecture have been proposed. For localizing kidney diseases, histogram-based feature along with AdaBoost algorithm is used. Deep convolutional neural network is used to recognize of kidney diseases, and TensorFlow batch prediction method is computed for recognition of diseases categories. The performance accuracy for detection of kidney disease is given as 89.79%. The performance of classification of chronic kidney diseases using CNN achieves an accuracy rate of 86.67%.