Chronic kidney disease clustering and classification using hybrid deep learning model

Abstract

Chronic Kidney Disease (CKD) is a critical global health concern, where early detection is vital for preventing severe complications and improving patient outcomes. This project addresses limitations in existing CKD classification methods, such as reliance on single data modalities and insufficient focus on early-stage detection, by developing a hybrid deep learning-based diagnostic system. The system integrates Custom Convolutional Neural Networks (CNN) and fine-tuned VGG19 models to classify CKD stages from JPG medical imaging data. The dataset, sourced from online, underwent preprocessing techniques such as resizing, normalization, and augmentation to enhance model performance and generalizability. The hybrid approach leverages the feature extraction strengths of VGG19 and the classification capabilities of Custom CNN, achieving an accuracy of 97.5% and 99.0%, respectively. Comparative analysis with existing methods demonstrated the superior scalability, reliability, and computational efficiency of the proposed system. Designed with clinical applicability in mind, the study adheres to standards such as HIPAA for data privacy and DICOM for medical imaging integration, ensuring feasibility in real- world settings. While dataset size and computational demands pose challenges, this scalable and adaptable framework lays the foundation for