Efficient Deep Transfer-Based Ensemble System for Robust Kidney Stone Identification

Chronic kidney disease is a major health problem across the world, and kidney stones are a prevalent illness that makes the kidneys work less well. Because kidney stone disorders don't show any symptoms, it's very important to find them early and accurately to avoid serious problems. This work uses datasets, Kidney data and CT Kidney Stone data, and innovative methods including Relief for feature selection and KNN with k-Fold validation for classification to improve the accuracy of diagnoses. For both classification and detection tasks, a group of deep neural networks and designs based on transfer learning have been used. The suggested method is quite accurate, as shown by the classification results, which reveal that Xception got 100% accuracy on the CT Kidney Stone data and 98% on the Kidney data. YOLOv5s6 has a mean average Precision (MAP) of 87% across both datasets, which shows that it works well for detection. These findings show that combining inductive transfer-based ensemble DL with optimal feature selection is a good way to find kidney stones quickly and reliably. This might help improve early diagnosis and patient outcomes.