文献速递心脑血管急症

用于检测ST段抬高型心肌梗死的临床验证人工智能模型的开发

Development of Clinically Validated Artificial Intelligence Model for Detecting ST-segment Elevation Myocardial Infarction.
PubMed 文献 · Ann Emerg Med · 2024-07-26
一句话结论基于韩国前瞻性PCI注册库心电图开发的深度集成AI模型诊断STEMI准确率92.1%、敏感度95.4%、特异度91.8%,但作者认为尚需前瞻性验证其真实世界临床获益。

本研究利用韩国前瞻性经皮冠状动脉介入治疗注册库的心电图波形数据,由两名独立委员会认证的心脏科医师依据对应冠状动脉造影数据建立STEMI/非STEMI标准,开发了组合5个卷积神经网络的深度集成模型,并用18,697份心电图(其中1,745份即9.3%为STEMI)建模。模型准确率92.1%、敏感度95.4%、特异度91.8%,在基于症状的心电图数据集临床验证、与临床医师比较及外部验证中表现均衡且出色。作者结论认为该深度集成AI模型表现均衡出色、基于梯度加权类激活映射具有合理可解释性,但认为尚需在真实世界环境中开展前瞻性验证以评估临床获益。

要点

  • 建模数据为韩国前瞻性PCI注册库心电图18,697份,其中1,745份(9.3%)为STEMI,金标准由两名独立委员会认证心脏科医师依据冠脉造影确定。
  • AI模型为组合5个卷积神经网络的深度集成模型,诊断STEMI准确率92.1%、敏感度95.4%、特异度91.8%。
  • 模型在基于症状的心电图数据集临床验证、与临床医师比较及外部验证中表现均衡且出色。
  • 以梯度加权类激活映射可视化,模型具有合理的可解释性。
  • 作者认为应开展前瞻性验证,评估该模型在真实世界环境中的临床获益。

适用人群与场景

  • 研究使用韩国前瞻性经皮冠状动脉介入治疗注册库的心电图波形数据,包括建模数据集18,697份心电图、基于症状的心电图数据集及外部验证数据。

诊断标准与关键阈值

  • 由两名独立委员会认证的心脏科医师依据对应的冠状动脉造影数据,为每份心电图建立STEMI或非STEMI的标准作为参照标准。

研究设计与关键结果

  • 建模数据集共18,697份心电图,其中1,745份(9.3%)为STEMI。
  • AI模型诊断STEMI的准确率92.1%、敏感度95.4%、特异度91.8%。
  • 模型在临床验证、与临床医师比较及外部验证中表现均衡且出色。
  • 以梯度加权类激活映射可视化,模型具有合理的可解释性。

证据类型、等级与局限

  • 证据类型:基于前瞻性经皮冠状动脉介入治疗注册库心电图的AI模型开发与验证研究。
  • 局限:作者认为应开展前瞻性验证,评估该模型在真实世界环境中的临床获益。

对床旁实践的改变

  • 作者认为,该深度集成AI模型表现均衡且出色,具有合理的可解释性;尚需在真实世界环境中开展前瞻性验证以评估其临床获益。

提炼说明

  • 本条依据论文摘要整理;摘要未涉及「处置流程与药物剂量」「红旗征与不能漏的情况」「指南或共识出处」,故未列出这些栏目;全文请点「阅读原文」。
英文原文(摘要)· 点击展开

STUDY OBJECTIVE: Although the importance of primary percutaneous coronary intervention has been emphasized for ST-segment elevation myocardial infarction (STEMI), the appropriateness of the cardiac catheterization laboratory activation remains suboptimal. This study aimed to develop a precise artificial intelligence (AI) model for the diagnosis of STEMI and accurate cardiac catheterization laboratory activation.

METHODS: We used electrocardiography (ECG) waveform data from a prospective percutaneous coronary intervention registry in Korea in this study. Two independent board-certified cardiologists established a criterion standard (STEMI or Not STEMI) for each ECG based on corresponding coronary angiography data. We developed a deep ensemble model by combining 5 convolutional neural networks. In addition, we performed clinical validation based on a symptom-based ECG data set, comparisons with clinical physicians, and external validation.

RESULTS: We used 18,697 ECGs for the model development data set, and 1,745 (9.3%) were STEMI. The AI model achieved an accuracy of 92.1%, sensitivity of 95.4%, and specificity of 91.8 %. The performances of the AI model were well balanced and outstanding in the clinical validation, comparison with clinical physicians, and the external validation.

CONCLUSION: The deep ensemble AI model showed a well-balanced and outstanding performance. As visualized with gradient-weighted class activation mapping, the AI model has a reasonable explainability. Further studies with prospective validation regarding clinical benefit in a real-world setting should be warranted.

来源:Ann Emerg Med,查看原文。版权归原作者所有。

参考来源

  1. Lee SH, Jeon KL, Lee YJ, et al. Development of Clinically Validated Artificial Intelligence Model for Detecting ST-segment Elevation Myocardial Infarction. Ann Emerg Med. 2024. PMID: 39066765. doi:10.1016/j.annemergmed.2024.06.004

供执业医师抢救与诊疗参考,不替代临床判断;剂量请结合体重、肝肾功能和本院药品规格复核,以指南原文为准。