Artificial Intelligence in Kidney Stone Imaging: Enhancing Classification and Detection for Improved Diagnostic Accuracy.

Full Text Links

Authors

Abstract

This study presents an advanced artificial intelligence (AI) model designed to accurately classify and detect kidney stones using medical imaging. Leveraging cloud-based computational resources, the model was trained to differentiate between stone-containing and normal kidneys while simultaneously identifying the precise localization of stones within images. The dataset consisted of 6,720 radiologic images representing clinically relevant stone cases and normal renal anatomy, with an 80-10-10 split for training, validation, and testing to ensure reliable assessment. Notably, the model achieved exceptional diagnostic performance, reflected by an average precision of 1.00 and both precision and recall reaching 99.9%. A perfect confusion matrix, demonstrating 100% correct classification of both stone and non-stone images, further underscores the robustness of the model. Model development required no physical hardware investment due to the use of cloud infrastructure, ensuring a cost-efficient and environmentally sustainable workflow. While results demonstrate strong clinical potential for automated nephrolithiasis detection, further evaluation on larger, multi-institutional datasets and across varied imaging modalities is recommended to strengthen generalizability. This work highlights the growing role of AI-enhanced diagnostic tools in urologic imaging, offering the promise of faster interpretation, improved workflow efficiency, and earlier identification of kidney stone disease.

Keywords: artificial intelligence, cloud-based modeling, convolutional neural network, deep learning, kidney stones, medical imaging, nephrolithiasis detection, object detection, ultrasound imaging, urologic diagnostics

Source

Cureus

Publication Type

Journal Article

Language

English

PubMed ID

41859640