人工智能在主动脉瘤影像分析中的应用:范围综述
本范围综述通过系统检索PubMed共识别661条记录,最终纳入59项将人工智能用于主动脉瘤医学影像的原创研究。深度学习尤其是U-Net架构最常用,主要用于自动分割与定量分析,在血管分割、动脉瘤检测、内漏识别和风险预测方面技术表现较高,但外部验证、多模态数据整合及前瞻性临床评估仍不常见。
要点
- 深度学习(尤其U-Net)是主动脉瘤影像AI中最常用的方法
- AI主要用于自动分割与定量分析,并涉及动脉瘤检测、内漏识别与风险预测
- 外部验证、报告透明度和临床整合是当前主要局限
核心结论
- AI在主动脉瘤评估中展现出通过准确影像分析、自动定量和流程支持提升诊疗的较大潜力。
- 当前局限包括缺乏外部验证、报告透明度不足以及临床整合不充分。
研究设计与样本
- 范围综述,系统检索PubMed识别661条记录,去重及多阶段筛选后纳入59项研究。
- 纳入标准为将人工智能应用于主动脉瘤诊疗医学影像的原创研究,提取研究设计、影像模态、AI方法、验证与数据集特征等信息。
主要发现
- 深度学习尤其U-Net架构是最常应用的方法,AI主要用于主动脉病理的自动分割与定量分析。
- AI模型在血管分割、动脉瘤检测、内漏识别和风险预测等应用中取得较高技术性能,并报告效率与诊断准确性改善。
- 外部验证、多模态数据整合与前瞻性临床评估仍不常见。
临床怎么用
- AI可辅助主动脉瘤影像的自动分割与定量分析,并用于动脉瘤检测、内漏识别与风险预测。
- 临床落地前需关注外部验证与报告透明度不足的问题。
局限性
- 缺乏外部验证。
- 报告透明度不足。
- 临床整合不充分,前瞻性临床评估少见。
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)。
原文
OBJECTIVE: Artificial intelligence (AI) has emerged as a promising approach to enhance automated image analysis, quantitative assessment, and clinical decision-making in aortic aneurysm care. This scoping review aims to summarize the current cumulative body of evidence. METHODS: A systematic PubMed search identified 661 records, with 59 studies ultimately included after de-duplication and multi-stage screening based on predefined eligibility criteria, focusing on original research applying artificial intelligence to medical imaging in aortic aneurysm care. Data were extracted on study design, imaging modality, AI methodology, validation, and dataset characteristics. RESULTS: Fifty-nine studies were included. Deep learning approaches, particularly U-Net-based architectures, were the most frequently applied methods, with AI primarily used for automated segmentation and quantitative analysis of aortic pathology. AI models achieved high technical performance across applications, including vessel segmentation, aneurysm detection, endoleak identification, and risk prediction, with reported improvements in efficiency and diagnostic accuracy. However, external validation, multimodal data integration, and prospective clinical evaluation remained uncommon. CONCLUSION: AI demonstrates substantial potential to enhance aortic aneurysm disease assessment through accurate image analysis, automated quantification, and improved procedural support. However, current limitations include lack of external validation, reporting transparency, and clinical integration.
引用
- Hofmann A. A Scoping Review on Artificial Intelligence for Image-Based Analysis in Aortic Aneurysm Care. Ann Vasc Surg. 2026. PMID: 42951185. doi:10.1016/j.avsg.2026.10.042
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。