Scholarlink

<p>Artificial intelligence (AI) in cardiology has evolved from rule-based expert systems to data-driven, learning models that can support diagnostic and therapeutic decision-making in acute coronary syndrome (ACS). At their core, these are statistical prediction models that infer conditional probabilities for clinical events from data and are therefore, in principle, ideally suited for risk-based decision making. This review article focuses on risk stratification in ACS, with particular emphasis on the detection of type&#xa0;1 or occlusive myocardial infarction (MI). Models based on clinical variables and high-sensitivity troponin measurements use continuous troponin values in combination with patient characteristics to estimate individual probabilities of MI instead of relying on fixed troponin thresholds. In retrospective studies, they reduce the “observe zone”, enable faster and safer rule-out decisions, and have the potential to relieve pressure on emergency departments. However, they remain dependent on laboratory assays and currently lack prospective evidence of improvements in patient relevant outcomes. Electrocardiogram (ECG)-based approaches promise greater clinical impact: classical machine learning (ML) on preselected ECG features already yields interpretable models that quantify ischemic patterns and can detect occlusive MI even in the absence of ST-segment elevation. Deep learning models applied directly to raw ECG signals additionally exploit subtle spatiotemporal changes and have outperformed conventional ECG interpretation in early studies. Prospective data on the detection of ST segment elevation myocardial infarction (STEMI) indicate that this can shorten the time to reperfusion.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Künstliche Intelligenz in der Risikostratifikation des akuten Koronarsyndroms

  • 期刊论文

  • 发表日期：

    2026年1月14日

查看全文
  • A. Büscher,
  • L. Plagwitz,
  • J. Blaschke,
  • D. Heider,
  • L. Eckardt

摘要

Artificial intelligence (AI) in cardiology has evolved from rule-based expert systems to data-driven, learning models that can support diagnostic and therapeutic decision-making in acute coronary syndrome (ACS). At their core, these are statistical prediction models that infer conditional probabilities for clinical events from data and are therefore, in principle, ideally suited for risk-based decision making. This review article focuses on risk stratification in ACS, with particular emphasis on the detection of type 1 or occlusive myocardial infarction (MI). Models based on clinical variables and high-sensitivity troponin measurements use continuous troponin values in combination with patient characteristics to estimate individual probabilities of MI instead of relying on fixed troponin thresholds. In retrospective studies, they reduce the “observe zone”, enable faster and safer rule-out decisions, and have the potential to relieve pressure on emergency departments. However, they remain dependent on laboratory assays and currently lack prospective evidence of improvements in patient relevant outcomes. Electrocardiogram (ECG)-based approaches promise greater clinical impact: classical machine learning (ML) on preselected ECG features already yields interpretable models that quantify ischemic patterns and can detect occlusive MI even in the absence of ST-segment elevation. Deep learning models applied directly to raw ECG signals additionally exploit subtle spatiotemporal changes and have outperformed conventional ECG interpretation in early studies. Prospective data on the detection of ST segment elevation myocardial infarction (STEMI) indicate that this can shorten the time to reperfusion.