Explainable deep learning for early diagnosis of chronic kidney disease from CT images in Bangladeshi patients

Kidney failure, or end-stage renal disease (ESRD), represents the final stage of chronic kidney disease (CKD) and poses a life-threatening risk if not addressed promptly. Early detection of CKD is critical for preventing progression to ESRD, yet current diagnostic methods remain time-consuming and often reliant on manual interpretation. This study introduces an integrated framework for automated CKD diagnosis, specifically designed for the Bangladeshi population, which combines segmentation, classification, and explainable artificial intelligence (XAI). Using the CT Kidney Dataset, a modified U-Net model was developed for kidney region segmentation, achieving an accuracy of 98%, a Dice coefficient of 98%, and an Intersection over Union (IoU) of 97%. For the classification task, a novel lightweight Kid-Net model, based on EfficientNetB3, was proposed, achieving 99.30% accuracy in cross-validation for distinguishing between normal, cyst, stone, and tumor categories. To enhance model transparency, Grad-CAM was applied for visualizing the regions of interest, thus improving interpretability. Furthermore, the KidVision framework was introduced to outline the clinical deployment pipeline, offering a scalable and efficient solution for real-world nephrology applications. The results demonstrate that the proposed framework not only delivers high accuracy but also facilitates early and automated detection of kidney-related disorders, contributing to improved clinical decision-making and patient outcomes.