IDX-DR检测糖尿病视网膜病变的诊断准确性:一项系统综述与Meta分析
该系统综述与Meta分析检索PubMed、Embase、Scopus和Web of Science截至2024年10月5日的文献,纳入13项研究共13,233例成人1型或2型糖尿病患者,评估IDX-DR检测糖尿病视网膜病变的合并灵敏度为0.95(95%CI:0.82-0.99),合并特异度为0.91(95%CI:0.84-0.95),SROC曲线AUC值为0.95。
要点
- IDX-DR检测糖尿病视网膜病变的合并灵敏度为0.95(95%CI:0.82-0.99)。
- IDX-DR检测糖尿病视网膜病变的合并特异度为0.91(95%CI:0.84-0.95)。
- SROC曲线AUC值为0.95,提示IDX-DR在不同临床环境中具有高诊断准确性。
核心结论
- IDX-DR是糖尿病视网膜病变筛查的高效诊断工具,具有稳健的灵敏度和良好的特异度。
- 将其整合到临床实践,尤其在资源有限环境中,可能改善早期检测并减少视力丧失。
- 需要谨慎实施以应对过度诊断等挑战,并确保该工具补充临床判断。
- 未来研究应探索AI筛查的长期影响并解决其使用中的伦理问题。
研究设计与样本
- 系统综述与Meta分析,检索PubMed、Embase、Scopus和Web of Science截至2024年10月5日发表的文献。
- 纳入涉及成人1型或2型糖尿病且报告灵敏度、特异度等诊断指标的研究。
- 采用双变量随机效应模型进行Meta分析,生成SROC曲线评估诊断性能,使用MetaDisc软件2.0版进行统计分析。
- 共纳入13项研究,涉及13,233名参与者。
主要发现
- IDX-DR的合并灵敏度为0.95(95%CI:0.82-0.99)。
- IDX-DR的合并特异度为0.91(95%CI:0.84-0.95)。
- SROC曲线确认IDX-DR在不同临床环境中检测糖尿病视网膜病变具有高诊断准确性。
- AUC值为0.95,表明高灵敏度和特异度,提示IDX-DR检测糖尿病视网膜病变具有稳健的诊断性能。
关键数值与推荐
- IDX-DR为FDA批准的自主人工智能系统,用于改善糖尿病视网膜病变筛查的可及性。
- 纳入标准为涉及成人1型或2型糖尿病且报告灵敏度、特异度等诊断指标的研究。
- 主要结局为IDX-DR的合并灵敏度和特异度。
临床怎么用
- IDX-DR是糖尿病视网膜病变筛查的高效诊断工具,具有稳健的灵敏度和良好的特异度。
- 将其整合到临床实践,尤其在资源有限环境中,可能改善早期检测并减少视力丧失。
- 需要谨慎实施以应对过度诊断等挑战,并确保该工具补充临床判断。
- 未来研究应探索AI筛查的长期影响并解决其使用中的伦理问题。
局限性
- 需要谨慎实施以应对过度诊断等挑战,并确保该工具补充临床判断。
- 未来研究应探索AI筛查的长期影响并解决其使用中的伦理问题。
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
PURPOSE: Diabetic retinopathy (DR) is a leading cause of vision loss worldwide, making early detection critical to prevent blindness. IDX-DR, an FDA-approved autonomous artificial intelligence (AI) system, has emerged as an innovative solution to improve access to DR screening. This systematic review and meta-analysis aimed to evaluate the diagnostic accuracy of IDX-DR in detecting diabetic retinopathy. DESIGN: Systematic review and meta-analysis. METHODS: A comprehensive literature search was conducted across PubMed, Embase, Scopus and Web of Science, identifying studies published through October 5, 2024. Studies involving adult patients with Type 1 or Type 2 diabetes and reporting diagnostic metrics such as sensitivity and specificity were included. The primary outcomes were pooled sensitivity and specificity of IDX-DR. A bivariate random-effects model was used for meta-analysis, and summary receiver operating characteristic (SROC) curves were generated to assess diagnostic performance. Statistical analyses were performed using MetaDisc software version 2.0. RESULTS: Thirteen studies involving 13,233 participants met the inclusion criteria. IDX-DR's pooled sensitivity was 0.95 (95% CI: 0.82-0.99), and its pooled specificity was 0.91 (95% CI: 0.84-0.95). The SROC curve confirmed IDX-DR's high diagnostic accuracy in detecting diabetic retinopathy across various clinical environments. The AUC value of 0.95 demonstrated high sensitivity and specificity, indicating a robust diagnostic performance for IDX-DR in detecting diabetic retinopathy. CONCLUSION: IDX-DR is a highly effective diagnostic tool for diabetic retinopathy screening, with robust sensitivity and good specificity. Its integration into clinical practice, especially in resource-limited settings, can potentially improve early detection and reduce vision loss. However, careful implementation is needed to address challenges such as over-diagnosis and ensure the tool complements clinical judgment. Future studies should explore the long-term impacts of AI-based screening and address ethical considerations surrounding its use.
引用
- Khan Z, Gaidhane AM, Singh M, et al. Diagnostic Accuracy of IDX-DR for Detecting Diabetic Retinopathy: A Systematic Review and Meta-Analysis. Am J Ophthalmol. 2025. PMID: 39986640. doi:10.1016/j.ajo.2025.02.022
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。