Kidney stones are a prevalent urological disorder that, if undetected, can lead to severe complications and pain. Manual diagnostic methods using medical imaging are time-consuming and prone to human error, necessitating automated solutions for accurate and reliable diagnosis. This study proposes a novel framework for kidney stone classification utilizing deep learning for feature extraction and K-Nearest Neighbors (KNN) for classification. Using Computed Tomography (CT) images, the framework extracts features using ResNet101 and InceptionV3 models, followed by classification with KNN. The proposed method achieved a remarkable classification accuracy of 99.96%, outperforming standalone models. This work demonstrates the effectiveness of combining transfer learning with machine learning techniques to enhance diagnostic accuracy, offering a practical and efficient tool for clinical applications. Future work will explore scalability and application to additional medical conditions.