儿科放射学人工智能研究现状与未来建议
本研究对2005—2024年Medline、Embase、Web of Science及Cochrane Library进行范围综述,共纳入789篇儿科放射学AI研究。结果显示研究主要来自中国和美国,最常见影像模态为X线、MRI和超声,最常见亚专业为骨肌、神经和胸部。多数研究聚焦图像判读与诊断,提示在沟通、教育、政策及利益相关者视角方面存在不平衡。
要点
- 儿科放射学AI研究以骨肌、神经、胸部亚专业为主,阅片时可优先关注这些领域的新工具。
- X线、MRI、超声是儿科AI研究最常用模态,相关算法落地场景相对集中。
- 图像判读与诊断占绝大多数研究,图像质量增强与运动伪影校正相关AI应用仍较少。
核心结论
- 2005—2024年儿科放射学AI研究以图像判读与诊断为主,中国和美国贡献最多,需加强全球合作并改善数据集代表性以降低算法偏倚。
研究设计与样本
- 研究类型为范围综述,检索Medline、Embase、Web of Science和Cochrane Library,时间范围为2005—2024年。
- 纳入标准为原始研究、关注<18岁儿科人群、且与临床放射学和AI有直接应用相关的文献,最终纳入789篇。
- 采用描述性频率进行汇总分析,并使用卡方检验比较不同类别之间的差异。
诊断标准与关键数值
- 最常见影像模态为X线298篇(37.8%)、MRI 260篇(33.0%)、超声114篇(14.4%),组间差异P<0.001。
- 最常见亚专业为骨肌260篇(33.0%)、神经227篇(28.8%)、胸部130篇(16.5%),组间差异P<0.001。
- 图像分析任务以图像判读/诊断为最多(719篇,91.1%),其次为伪影与运动减少/图像质量增强(44篇,5.6%),P<0.001。
- 发文最多的国家为中国220篇(27.9%)、美国200篇(25.4%)、加拿大51篇(6.5%),P<0.001。
临床应用与阅片要点
- 日常关注儿科放射学AI进展时,可优先检索骨肌、神经和胸部领域,X线、MRI和超声场景更具临床转化潜力。
- 目前在沟通、教育、政策和利益相关者视角方面的AI研究明显不足,采购或评估相关工具时应注意证据缺口。
- 阅片工作中引入儿科AI辅助工具时,需关注训练集人群代表性不足可能带来的偏倚。
局限性
- 研究归属为范围综述,主要揭示宏观趋势,未对单篇文献进行偏倚风险或方法学质量评价。
- 纳入研究集中在少数国家和常见亚专业,其他地区和亚专科的AI证据覆盖可能不足。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
英文原文(摘要)· 点击展开
BACKGROUND: Most artificial intelligence (AI) research in radiology has focused on adults. Understanding macro-level trends in pediatric radiology AI can help guide, streamline, and bolster future research.
OBJECTIVE: To detail the current landscape of published AI research in pediatric radiology, filling a key research gap, as most radiology AI research has focused on adults.
MATERIALS AND METHODS: We conducted a scoping review, with a comprehensive literature search of Medline, Embase, Web of Science, and Cochrane Library from 2005 to 2024. Literature included for review were (1) original articles, (2) investigations that focused on pediatric populations (<18 years of age), and (3) articles with direct applications to clinical radiology and AI. We extracted each article's study information, clinical application of focus, imaging modality, and the use of AI. We used descriptive frequencies to analyze summary statistics, and Chi-square testing to determine differences between categories.
RESULTS: In total, we found 4,376 articles and included 789 articles in the review. The top three countries most active in scholarship related to AI in pediatric radiology were China (220, 27.9%), the USA (200, 25.4%), and Canada (51, 6.5%) (P<0.001). The most common imaging modalities were radiography (298, 37.8%), MRI (260, 33.0%), and ultrasonography (114, 14.4%) (P<0.001). The most common subspecialties represented were musculoskeletal (260, 33.0%), neurological (227, 28.8%), and chest imaging (130, 16.5%) (P<0.001). The top two image analysis tasks discussed were image interpretation/diagnosis (719, 91.1%), and artifact and motion reduction/enhancing image quality (44, 5.6%) (P<0.001).
CONCLUSION: Most pediatric radiology AI research originated from China and the USA, and focused on image interpretation/diagnosis. Thematic imbalances, particularly underrepresentation in research on communication, education, policy, and stakeholder perspectives, offer a guide for pediatric radiology AI development. There is a need for improved global collaboration and improved patient representativeness in datasets for pediatric radiology AI research to reduce bias with AI algorithms. The results from this scoping review offer a practical roadmap to inform future research and funding priorities in pediatric radiology AI.
引用
- Kamran, Widjaja, Sy, et al. The current state of artificial intelligence research in pediatric radiology and recommendations for the future: a scoping review. Pediatr Radiol. 2026. PMID: 41579167. doi:10.1007/s00247-025-06462-5
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。