Altunhan A. et al., 2024: Artificial intelligence in urolithiasis: a systematic review of utilization and effectiveness.

Abdullah Altunhan, Selim Soyturk, Furkan Guldibi, Atinc Tozsin, Abdullatif Aydın, Arif Aydın, Kemal Sarica, Selcuk Guven, Kamran Ahmed 
World J Urol. 2024 Oct 17;42(1):579. doi: 10.1007/s00345-024-05268-8

Abstract

Purpose: Mirroring global trends, artificial intelligence advances in medicine, notably urolithiasis. It promises accurate diagnoses, effective treatments, and forecasting epidemiological risks and stone passage. This systematic review aims to identify the types of AI models utilised in urolithiasis studies and evaluate their effectiveness.

Methods: The study was registered with PROSPERO. Pubmed, EMBASE, Google Scholar, and Cochrane Library databases were searched for relevant literature, using keywords such as 'urology,' 'artificial intelligence,' and 'machine learning.' Only original AI studies on urolithiasis were included, excluding reviews, unrelated studies, and non-English articles. PRISMA guidelines followed.

Results: Out of 4851 studies initially identified, 71 were included for comprehensive analysis in the application of AI in urolithiasis. AI showed notable proficiency in stone composition analysis in 12 studies, achieving an average precision of 88.2% (Range 0.65-1). In the domain of stone detection, the average precision remarkably reached 96.9%. AI's accuracy rate in predicting spontaneous ureteral stone passage averaged 87%, while its performance in treatment modalities such as PCNL and SWL achieved average accuracy rates of 82% and 83%, respectively. These AI models were generally superior to traditional diagnostic and treatment methods.

Conclusion: The consolidated data underscores AI's increasing significance in urolithiasis management. Across various dimensions-diagnosis, monitoring, and treatment-AI outperformed conventional methodologies. High precision and accuracy rates indicate that AI is not only effective but also poised for integration into routine clinical practice. Further research is warranted to establish AI's long-term utility and to validate its role as a standard tool in urological care.

Comment Peter Alken

A good publication and a milestone in determining one's own knowledge of AI in urolithiasis. This conclusion is expected: “Our analysis suggests a clear difference in AI performance between machine-derived data, such as CT scans, and endourological procedures that depend on human expertise. While AI excels in interpreting imaging data, its accuracy diminishes in contexts that necessitate procedural skill and adaptability. This underscores the importance of using AI as a supplementary tool in urology, enhancing decision-making while relying on the critical judgment of experienced clinicians. Therefore, the integration of AI should be approached cautiously, especially in complex surgical environments where human expertise is essential. The potential of AI in improving outcomes is undeniable, but its role should be understood as complementary, particularly in areas where human factor cannot be entirely replicated or replaced.”
In this respect, an advanced EWSL could be the first AI-guided stone removal procedure that is fully machine-controlled - quasi-robotic - from imaging to patient positioning, focussing, shockwave triggering, progress monitoring and termination, including notification of the need for a second session.

Peter Alken