Using Deep Learning Technology to Detecting Kidney Disease

Abstract

Deep learning is gaining significant importance due to data interpretation and its application to a wide range of diseases in general, especially kidney disease detection. Machine learning is widely applied in healthcare. Providing a sufficient number of samples (images) and providing the appropriate statistical software were major research challenges. Image preprocessing was then performed to convert the original medical image data into medical image data free of some unwanted distortions (noise). This was done to enhance the images, which must be of the same size and dimensions, to reveal certain image features, and transform them into high-quality medical images for use in kidney disease detection and classification. Ten statistical image metrics were used: SC, AD, MD, SSIM, RFSIM, FSIM, RMSE, LMSE, MAE, and PCC. These metrics were used to evaluate the performance of the two classification algorithms (CNN and RNN). The classification process used precision, accuracy, recall, and F1 metrics. High accuracy results were achieved, indicating that the resulting results are very good. The evaluation metrics provide the information needed to determine the performance of a classification model based on a given score (correct or incorrect). The study focused on the use of deep learning in the recognition and classification of kidney diseases. The results of the current study indicated that the CNN and RNN algorithms used in the proposed kidney disease recognition and classification system achieved acceptable results. The results achieved high classification accuracy for the proposed method for kidney diseases (90.52%, 64.64%) and for normal cases (94.05%, 77.12%) for the CNN and RNN algorithms, respectively.