Research Centre for Health and Social Informatics, Osnabrück University of Applied Sciences, Germany.
Stud Health Technol Inform. 2024 Aug 30;317:219-227. doi: 10.3233/SHTI240860.
Cardiovascular diseases are a leading cause of mortality worldwide, highlighting the urgent need for accurate and efficient diagnostic tools. Echocardiography, a non-invasive imaging technique, plays a central role in the diagnosis of heart diseases, yet the potential impact of artificial intelligence (AI) on its accuracy and speed has not yet been reviewed and summarized. This scoping review aims to address this research gap by synthesizing existing evidence on AI-assisted echocardiography's.
The study followed Arksey and O'Malley's six-stage model for scoping reviews and searched the databases PubMed, Web of Science and Livivo. Inclusion criteria encompassed studies from cardiology utilizing AI for heart diseases diagnosis in adults, published from 2018 to 2023. Data extraction focused on study characteristics, AI models employed, accuracy metrics, and diagnostic speed.
From 1059 identified studies, nine records met the inclusion criteria, categorized into view classification, left ventricular ejection fraction (LVEF) quantification, and diseases classification. Convolutional Neural Networks (CNN) were commonly used. While 44% of studies compared AI with cardiologists, those studies indicated AI's high diagnostic accuracy, with mean accuracy ranging from 87% to 92%. Three studies assessed AI's speed, demonstrating significant time savings.
The review highlights AI's potential in enhancing diagnostic accuracy and efficiency in echocardiography, particularly in regions with limited access to specialized cardiologists. However, further research is needed to assess AI's specific added value compared to cardiologists, optimize training data quality, and enable real-time image processing.
心血管疾病是全球范围内导致死亡的主要原因之一，这突显了对准确且高效的诊断工具的迫切需求。超声心动图作为一种非侵入性的成像技术，在心脏病的诊断中发挥着核心作用，但人工智能（AI）对其准确性和速度的潜在影响尚未得到审查和总结。本范围性综述旨在通过综合现有的关于 AI 辅助超声心动图的证据来解决这一研究空白。
该研究遵循 Arksey 和 O'Malley 的六阶段模型进行范围性综述，并在 PubMed、Web of Science 和 Livivo 数据库中进行了搜索。纳入标准包括利用 AI 进行成人心脏病诊断的心脏病学研究，发表时间为 2018 年至 2023 年。数据提取重点关注研究特征、使用的 AI 模型、准确性指标和诊断速度。
从 1059 项已识别的研究中，有 9 项记录符合纳入标准，分为视图分类、左心室射血分数（LVEF）定量和疾病分类。卷积神经网络（CNN）被广泛使用。虽然 44%的研究将 AI 与心脏病专家进行了比较，但这些研究表明 AI 具有很高的诊断准确性，平均准确性范围从 87%到 92%。有三项研究评估了 AI 的速度，表明其显著节省了时间。
该综述强调了 AI 在增强超声心动图诊断准确性和效率方面的潜力，特别是在那些获得专业心脏病专家服务有限的地区。然而，需要进一步研究来评估 AI 与心脏病专家相比的具体附加值，优化训练数据质量，并实现实时图像处理。
