DETECTION AND CLASSIFICATION OF KIDNEY DISEASE BY USING A DEEP LEARNING ALGORITHM

Asma Munawer; Humayun Salahuddin; Prof. Dr. Muhammad Rafiq; Dr. Hafsa Munawar; Muhammad Wajid Maqbool

Authors

Asma Munawer Department of Computer Science, Times University Multan
Humayun Salahuddin Department of Computer Science, Times University Multan.
Prof. Dr. Muhammad Rafiq Department of Computer Science, Times University Multan.
Dr. Hafsa Munawar Nephrology department, Pakistan institute of medical sciences, Islamabad
Muhammad Wajid Maqbool Department of Information Sciences, University of Education Lahore, Multan Campus.

Abstract

Kidney disease is a critical global health issue that requires accurate and timely diagnosis to improve patient outcomes. This research presents an automated kidney disease classification system that uses advanced deep learning techniques on medical imaging data. A benchmark dataset consisting of four classes—Normal, Cyst, Stone, and Tumor—was collected from the Kaggle repository and systematically organized for analysis. Comprehensive preprocessing techniques, including image resizing, normalization, removal of corrupted data, and data augmentation, were applied to enhance data quality and model performance. The dataset was divided into training, validation, and testing sets to ensure reliable and unbiased evaluation. Three deep learning models were developed and compared: a custom Convolutional Neural Network (CNN), a transfer learning–based VGG-16 model, and the Swin Transformer model. The custom CNN was designed to extract hierarchical features directly from input images, while the VGG-16 model leveraged pre-trained weights to improve learning efficiency. The Swin Transformer model was utilized to capture both local and global contextual information through its hierarchical attention mechanism. Model performance was evaluated using standard metrics, including accuracy, loss, precision, recall, and F1-score. Experimental results demonstrate that all models achieved high classification performance; however, the Swin Transformer model outperformed the others, achieving a training accuracy of 99.90% and testing accuracy of 99.75%, along with minimal loss values. These findings highlight the effectiveness of transformer-based architectures in medical image classification and propose their potential for supporting clinical decision-making in kidney disease diagnosis.

Keywords: Kidney Disease, Medical Image Classification, Deep Learning, Convolutional Neural Network (CNN), VGG-16, Swin Transformer, Transfer Learning, Kidney Cyst, Kidney Stone, Kidney Tumor, Medical Imaging, Image Preprocessing, Disease Detection, Computer-Aided Diagnosis