Deep Learning-Based Multiclass Classification Of Kidney Disorders Using Ct Images

Abstract

The early and accurate diagnosis of kidney disorders is critical for effective treatment and improved patient outcomes. This research presents a deep learning–based approach utilizing Convolutional Neural Networks (CNNs) to classify kidney conditions namely Cyst, Stone, Tumor, and Normal using computed tomography (CT) images. A dataset comprising 12,446 CT images, sourced from Kaggle, was methodically split into training, validation, and test sets using a reproducible strategy to ensure robust evaluation. The CNN model was developed and trained using Python, with TensorFlow and Keras as the primary frameworks, supported by Google Colab for accelerated training on GPU resources. The proposed model achieved an exceptional classification accuracy of 99.89% and an Area under the Curve (AUC) score of 100%, underscoring its high precision and reliability in medical image classification tasks. The results demonstrate that deep learning offers a powerful, scalable solution for assisting radiologists in the automated detection of kidney disorders, potentially transforming diagnostic workflows in clinical settings.