A Review of the Intraoperative Use of Artificial Intelligence in Urologic Surgery

Abstract

Introduction: Future evolutions of artificial intelligence (AI) will support autonomous surgery, conducted without the need for human decision making and implementation, but we have not yet achieved this level of technology. Presently, the predominant applications of AI in urological surgery are achieved using the tool of computer vision. This review aims to summarise potential intra-operative AI tools for urologists. Method: A systematic search was conducted through Scopus, PubMed, Embase, and Medline by two independent reviewers, with a third to resolve any conflicts. As a rule, only original articles describing the use or potential use of artificial intelligence intra-operatively in urologic surgery were included. A total of 60 articles were reviewed. Key content and findings: There is significant research investigating the ability to diagnose bladder tumours using AI assistance at the time of cystoscopy, with studies showing the ability to also grade tumour based on appearance and differentiate between carcinoma in situ and indeterminate lesions. With the aid of AI, kidney stones can accurately be identified and diagnosed morphologically intra-operatively. Various studies show the ability to overlay 2D and 3D anatomical models on a surgeon’s screen, as well as correctly identify important anatomical landmarks and surgical instruments, with AI support. All types of intra-operative data can be analysed with AI to assess surgeon performance, predict post-operative outcomes such as continence post prostatectomy, and recognise complications such as bleeding and ischemia. Conclusions: AI holds great potential for urologists during surgery to improve safety, diagnostic accuracy, identification of anatomical structures and surgical instruments, assessment of the surgeon for self-evaluation, and prediction of post-operative outcomes. Before the use of AI as an aid during surgery becomes standard practice, more prospective studies are needed to evaluate its real-world application, feasibility, and costs.

Ähnliche Arbeiten