Recent Advances in Deep Learning-based Chronic Kidney Disease Detection

Chronic Kidney Disease (CKD) is a progressive disorder that affects millions worldwide, leading to severe complications if not detected early. Traditional diagnostic methods, including blood tests and biopsies, often fail to identify CKD in its initial stages, necessitating the adoption of advanced techniques for improved accuracy. Deep learning (DL) has emerged as a powerful tool in medical imaging, significantly enhancing CKD detection using Computed Tomography (CT) scans. DL models, such as Convolutional Neural Networks (CNNs), and hybrid architectures, have been extensively applied to extract features, classify abnormalities, and segment kidney structures. This review provides a comprehensive analysis of DL-based methodologies for CKD detection, covering pre-processing techniques, feature extraction methods, and classification approaches. A comparative evaluation of existing models is conducted, highlighting their performance, advantages, and limitations. While DL offers remarkable accuracy and automation, challenges like annotated datasets, high computational costs, and limited interpretability remain significant concerns. Emerging techniques, including transfer learning (TL), attention mechanisms, and multi-modal imaging methods, have exposed promising results in improving CKD diagnosis. The integration of Artificial intelligence (AI) in CKD detection has minimized human errors, accelerated diagnosis, and enabled real-time monitoring, contributing to personalized treatment strategies. However, clinical acceptance requires further advancements in model optimization, data availability, and explainable AI techniques to ensure reliability and transparency. This review supports as a valuable source for researchers and healthcare professionals aiming to advance CKD diagnosis through DL methodologies.