超越外科医生肉眼:人工智能用于游离血管化皮瓣术后监测的叙述性综述
该叙述性综述检索2000年至2026年2月的文献,纳入9项研究,评估人工智能用于游离皮瓣术后监测的方法与临床转化。AI系统在检测动脉缺血和静脉淤血方面诊断准确率为82%至98.4%,但多数研究为回顾性,缺乏前瞻性验证和外部验证,存在数据集有限和方法异质性等障碍。
要点
- AI辅助游离皮瓣监测诊断准确率82%至98.4%,多用于检测动脉缺血和静脉淤血。
- 纳入9项研究中仅1项非回顾性,无前瞻性人体临床试验。
- 主要障碍包括数据集有限、缺乏外部验证和方法异质性。
核心结论
- AI辅助游离皮瓣术后监测是有前景的辅助手段,但方法异质性、数据稀缺和缺乏前瞻性验证阻碍广泛临床实施。
研究设计与样本
- 叙述性综述与系统检索混合方法,检索PubMed、Embase、Web of Science和Google Scholar从2000年1月至2026年2月2日。
- 纳入9项报告皮瓣活力、缺血检测、静脉淤血、血栓预防或挽救率等临床相关结局的研究。
- 两名独立评审者进行文献筛选和数据提取,分歧通过共识解决。
主要发现
- 多数研究使用监督机器学习模型,主要基于可见光图像、红外成像、光电容积描记、高光谱成像或多模态传感器数据。
- 除1项研究外均为回顾性,未发现前瞻性人体临床试验。
- AI系统在检测动脉缺血和静脉淤血方面诊断准确率为82%至98.4%。
临床怎么用
- AI系统可作为传统监测的辅助手段,用于早期缺血检测和减少工作量。
- 临床实施前需解决方法异质性、数据稀缺和缺乏前瞻性验证等问题。
局限性
- 数据集有限,缺乏外部验证,方法异质性大,无标准化结局指标。
- 对多样肤色和临床环境的泛化性存在担忧。
- 缺乏前瞻性人体临床试验验证。
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
BACKGROUND: Postoperative monitoring of free vascularized flaps (FVFs) is critical for early detection of ischemia and timely salvage. Although traditional monitoring relies heavily on clinical examination and adjunctive devices, these approaches are labor‑intensive, subjective, and resource‑dependent. Recent advances in artificial intelligence (AI), particularly machine learning and deep learning, have enabled automated, data‑driven approaches to postoperative free‑flap monitoring. However, the clinical readiness, methodological robustness, and translational challenges of these AI‑based systems remain incompletely understood. OBJECTIVE: This review examines recent studies on AI‑assisted postoperative free‑flap monitoring, focusing on technical methodologies, clinical applications, translational relevance, current limitations, and future research directions. METHODS: A systematic-narrative hybrid review was conducted by searching PubMed, Embase, Web of Science, and Google Scholar from January 2000 to February 2, 2026. Eligible studies reported clinically relevant outcomes, including flap viability, ischemia detection, venous congestion, thrombosis prevention, and salvage rates. Studies focused on preoperative prediction models, non‑AI monitoring techniques, editorials, and opinion articles were excluded. Two independent reviewers performed study selection and data extraction, with disagreements resolved by consensus. RESULTS: Nine studies met the inclusion criteria. Most studies used supervised machine‑learning models, primarily using visible‑light images, infrared imaging, photoplethysmography, hyperspectral imaging, or multimodal sensor data for flap surveillance. All but 1 study were retrospective, and no prospective human clinical trials were identified. AI‑based systems demonstrated promising performance in detecting arterial ischemia and venous congestion, often achieving diagnostic accuracies comparable to or exceeding those of traditional monitoring methods. Diagnostic accuracy across studies ranged from 82% to 98.4%. Key challenges included limited data sets, lack of external validation, heterogeneous methodologies, absence of standardized outcome metrics, and concerns about generalizability across diverse skin tones and clinical settings. CONCLUSIONS: AI‑assisted postoperative free‑flap monitoring is a rapidly evolving and promising adjunct to conventional surveillance. Although current evidence suggests benefits for early ischemia detection and workload reduction, significant barriers-particularly methodological heterogeneity, data scarcity, and limited prospective validation-must be addressed before widespread clinical implementation. Future progress will depend on standardized data sets, transparent model development, multicenter prospective studies, and ethically sound, interoperable AI frameworks that enable equitable and scalable precision microsurgical care.
引用
- Ebrahimian M, Kamali R. Seeing Beyond the Surgeon's Eye: Artificial Intelligence for Postoperative Monitoring of Free Vascularized Flaps: A Narrative Review. Ann Plast Surg. 2026. PMID: 42554452. doi:10.1097/SAP.0000000000004848
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。