Enhanced Deep Learning Models for Early Detection of Renal Failure from CT Imaging

Chronic kidney disease (CKD) is a progressive condition affecting over 12% of the global population and can lead to end-stage renal disease (ESRD) if not managed effectively. Early prediction of renal failure in CKD patients is crucial for timely intervention, which can slow or prevent disease progression and reduce the need for costly treatments like dialysis or transplantation. However, traditional diagnostic methods, such as serum creatinine tests and nuclear imaging, often lack precision and are invasive. This paper examines the application of optimized deep learning models, particularly those utilizing computed tomography (CT) imaging, for the early detection of renal failure. Key approaches include AI-assisted segmentation of renal-enhanced CT images and advanced models like YOLOv8, which have shown promise in accurately identifying kidney abnormalities and assessing risks. By leveraging these cutting-edge technologies, the goal is to improve early detection, enhance patient outcomes, and reduce healthcare costs, addressing the global burden of CKD. The proposed YOLOv8 model obtained a precision up to 0.986, recall up to 0.969, mAP50 up to 0.989, and mAP50-95 up to 0.972 across BOX and MASK predictions.