Khondker Adree, Kwong Jethro C C, Rickard Mandy, Erdman Lauren, Gabrielson Andrew T, Nguyen David-Dan, Kim Jin Kyu, Abbas Tariq, Fernandez Nicolas, Fischer Katherine, 't Hoen Lisette A, Keefe Daniel T, Nelson Caleb P, Viteri Bernarda, Wang Hsin-Hsiao Scott, Weaver John, Yadav Priyank, Lorenzo Armando J
Division of Urology, The Hospital for Sick Children, Toronto, ON, Canada; Division of Urology, Department of Surgery, University of Toronto, Toronto, ON, Canada.
Division of Urology, Department of Surgery, University of Toronto, Toronto, ON, Canada; Temerty Centre for AI Research and Education in Medicine, University of Toronto, Toronto, ON, Canada.
J Pediatr Urol. 2025 Apr;21(2):532-538. doi: 10.1016/j.jpurol.2024.10.003. Epub 2024 Oct 5.
Artificial intelligence (AI) and machine learning (ML) methods are increasingly being applied in pediatric urology across a growing number of settings, with more extensive databases and wider interest for use in clinical practice. More than 30 ML models have been published in the pediatric urology literature, but many lack items required by contemporary reporting frameworks to be high quality. For example, most studies lack multi-institution validation, validation over time, and validation within the clinical environment, resulting in a large discrepancy between the number of models developed versus the number of models deployed in a clinical setting, a phenomenon known as the AI chasm. Furthermore, pediatric urology is a unique subspecialty of urology with low frequency conditions and complex phenotypes where clinical management can rely on a lower quality of evidence.
To establish the AI in PEDiatric UROlogy (AI-PEDURO) collaborative, which will carry out a living scoping review and create an online repository (www.aipeduro.com) for models in the field and facilitate an evidence synthesis of AI models in pediatric urology.
The scoping review will follow PRISMA-ScR guidelines. We will include ML models identified through standardized search methods of four databases, hand-search papers, and user-submitted models. Retrieved records will be included if they involve ML algorithms for prediction, classification, or risk stratification for pediatric urology conditions. The results will be tabulated and assessed for trends within the literature. Based on data availability, models will be divided into clinical disease sections (e.g. hydronephrosis, hypospadias, vesicoureteral reflux). A risk assessment will be performed using the APPRAISE-AI tool. The retrieved model cards (brief summary model characteristics in table form) will be uploaded to the online repository for open access to clinicians, patients, and data scientists, and will be linked to the Digital Object Identifier (DOI) for each article.
We hope this living scoping review and online repository will offer a valuable reference for pediatric urologists to assess disease-specific ML models' scope, validity, and credibility to encourage opportunities for collaboration, external validation, clinical testing, and responsible deployment. In addition, the repository may aid in identifying areas in need of further research.
人工智能（AI）和机器学习（ML）方法在越来越多的儿科泌尿外科环境中得到越来越广泛的应用，拥有更广泛的数据库，并且在临床实践中的应用兴趣也越来越大。儿科泌尿外科文献中已发表了30多种ML模型，但许多模型缺乏当代报告框架要求的高质量要素。例如，大多数研究缺乏多机构验证、长期验证以及临床环境中的验证，导致开发的模型数量与临床环境中部署的模型数量之间存在很大差异，这种现象被称为人工智能鸿沟。此外，儿科泌尿外科是泌尿外科的一个独特亚专业，疾病发生率低且表型复杂，临床管理可能依赖较低质量的证据。
建立儿科泌尿外科人工智能（AI-PEDURO）协作组织，该组织将进行实时范围综述，并为该领域的模型创建一个在线知识库（www.aipeduro.com），并促进儿科泌尿外科AI模型的证据综合。
范围综述将遵循PRISMA-ScR指南。我们将纳入通过四个数据库的标准化搜索方法、手工检索论文以及用户提交的模型识别出的ML模型。如果检索到的记录涉及用于儿科泌尿外科疾病预测、分类或风险分层的ML算法，则将其纳入。结果将制成表格并评估文献中的趋势。根据数据可用性，模型将分为临床疾病类别（如肾积水、尿道下裂、膀胱输尿管反流）。将使用APPRAISE-AI工具进行风险评估。检索到的模型卡片（以表格形式呈现的简要模型特征）将上传到在线知识库，供临床医生、患者和数据科学家公开访问，并将链接到每篇文章的数字对象标识符（DOI）。
我们希望这个实时范围综述和在线知识库能为儿科泌尿外科医生评估特定疾病的ML模型的范围、有效性和可信度提供有价值的参考，以鼓励合作、外部验证、临床试验和负责任部署的机会。此外，该知识库可能有助于确定需要进一步研究的领域。
