Transfer Learning Empowered Multi-Class Classification of Kidney Diseases: A Deep Learning Approach

The growing rate of chronic kidney disease and a shortage of renal specialists make renal dysfunction a major worldwide health concern. This study, which makes use of technological breakthroughs, suggests an AI-driven system for the autonomous diagnosis of kidney abnormalities, with an emphasis on chronic kidney disease, which is characterized by a steady reduction in function. Many conditions, such as cysts, stones, and tumors, can cause renal failure, these conditions frequently demonstrate no symptoms at first. In an attempt to overcome the lack of medical professionals, this research uses deep learning and a modified fine-tuned InceptionV3 pre-trained model to examine 12,446 Computed Tomography (CT) full abdomen and urogram images that were obtained from several hospital databases. The dataset includes both normal and common kidney disease images such as cysts, tumors, and stones. The proposed model exhibits improved accuracy through careful annotation and model training, it achieves remarkable performance metrics such as 96.52% accuracy rate, along with average precision of 97.76%, 98.28% recall and 0.95 F1 score. With its adaptable clinical diagnosis solution, this cutting-edge technology makes kidney disorders easily identifiable, allowing for immediate action in the treatment of renal disorders.