# Artificial Intelligence in Urolithiasis Imaging and Intervention: A Narrative Review of Current Applications, Barriers, and Future Directions

## TL;DR

This review explores how artificial intelligence is transforming kidney stone diagnosis and treatment, improving imaging accuracy and procedural predictions.

## Contribution

A comprehensive overview of AI applications in urolithiasis, highlighting current uses, challenges, and future research directions.

## Key findings

- AI improves CT and US accuracy for stone detection and differentiation.
- AI models predict success rates of procedures like ESWL and PCNL better than traditional systems.
- Radiomics and algorithms show promise in preoperative stone composition analysis.

## Abstract

The treatment of urolithiasis is changing quickly by moving away from conventional diagnostic techniques and toward more complex, data-driven strategies. A major part in this change is being played by artificial intelligence (AI) through providing the clinicians with invaluable assistance. This study examines the state of AI applications in urolithiasis today and how they affect everything from treatment planning to initial imaging. AI models are improving the accuracy of computed tomography (CT) and ultrasonography (US) in diagnoses. These techniques provide automatic stone detection throughout the diagnostic process and a precise stone burden calculation, and even assistance in differentiating difficult mimics, such as ureteral stones, from phleboliths. Additionally, sophisticated algorithms and radiomics are demonstrating great promise in determining the composition of stones preoperatively from imaging data or even digital photos. AI has also changed and improved the intervention for kidney stones, which is highlighted by models now capable of predicting the success of procedures like extracorporeal shock wave lithotripsy (ESWL) and percutaneous nephrolithotomy (PCNL), in some cases outperforming traditional scoring systems. Despite this progress, significant hurdles remain, particularly the need for large datasets and ensuring models are reliable and generalizable across different clinical settings. Successfully integrating these powerful tools into daily urological practice will require a concerted effort toward developing best-practice guidelines, robust training programs, and strong interdisciplinary collaboration. This review aims to summarize current AI applications in imaging and intervention for urolithiasis, identify limitations, and outline future research directions.