Chronic kidney disease (CKD) poses a significant global health concern, the nature of the disease requires sophisticated diagnostic approaches. Glomerulonephritis, a leading cause of CKD, requires early detection for effective intervention. This paper introduces an innovative solution that integrates lightweight machine learning models within a fog-cloud collaborative platform to optimize kidney disease diagnosis. Due to the complexity of glomerulonephritis diagnosis, the platform aims to bridge radiologists, pathologists, clinicians, nephrologists, and patients, using lightweight, adaptive, and robust algorithms tailored for fog-based deployment. Conventional diagnostic methods face challenges of subjectivity and labor intensiveness, particularly in diagnosing subtypes like IgA nephropathy and PLA2R-associated membranous nephropathy. To overcome these challenges, we propose the integration of AI diagnosis without imposing heavy processing loads on the Picture Archiving and Communication Systems (PACS). Hence, we introduce the concept of a Lightweight Nephritis Pathological Diagnosis Cloud-Edge-Collaborative Platform. This platform entails: 1) establishing a comprehensive kidney pathological image database; 2) developing an automatic renal pathological image processing system; 3) creating machine learning algorithms for swift and precise diagnosis; and 4) deploying an intelligent diagnosis cloud service platform with a robust clinical decision-support rule induction system by federated learning. In order to fulfill these functions, we introduce three novel innovations. Firstly, a robust and lightweight rule-induction algorithm is proposed for clinical test result classification, the predictive model can be updated quickly. Secondly, an adaptive new deep learning architecture is introduced for analyzing IgA Nephritis histological images, with fast model refreshing. Finally, a novel fusion scheme is proposed to enable collaborative decision-making online among distributed users, utilizing fine-grained federated learning principles. These innovations collectively enhance the privacy and accuracy of glomerulonephritis diagnosis, improving kidney disease management.