早期发现结构性心脏病对于改善预后至关重要，但广泛筛查仍然受限于超声心动图等影像工具的成本 和可及性。近期，人工智能用于早期发现结构性心脏病的研究成果登上《自然》杂志。AI独立分析 心电图（ECG）数据的检测准确率甚至高于AI辅助下的人类心脏病专家。 这项最新的研究开发了一款名为EchoNext的人工智能筛查工具，通过分析常规心电图（ECG）数 据，能够从中准确识别出有结构性心脏病的患者。这将为下一步确定哪些患者需要进一步接受超声心 动图诊断进行有效筛选，避免了医疗检查资源的浪费。 虽然心电图用途广泛，但也存在局限性——心电图无法反映出心脏的结构，仅能为结构性心脏病提供 间接诊断依据。超声心动图利用超声波获取心脏图像，可用于明确诊断心脏瓣膜病、心肌病、肺动脉 高压以及其他需要药物或手术治疗的心脏结构问题。 目前结构性心脏病都可以通过超声心动图明确诊断，但这种诊断方法成本高、需要专业知识以及合适 的患者选择，限制了其全面应用。因此，迫切需要更好地对患者进行风险分层，并确定哪些患者应转 诊进行超声心动图检查，从而提高结构性心脏病的诊断率和早期治疗率。 在这项最新的研究中，人工智能模型接受了23万名患者超过120万对 ...

Core Viewpoint - The article discusses the increasing prevalence of Structural Heart Disease (SHD) and highlights the development of an AI screening tool, EchoNext, which can accurately identify patients with SHD from standard electrocardiograms (ECGs) [1][4][10]. Group 1: Overview of Structural Heart Disease - Structural Heart Disease (SHD) includes conditions affecting heart valves, walls, or chambers, impacting millions globally [1]. - Early detection of SHD can reduce mortality, treatment costs, and improve quality of life, but many patients are diagnosed late due to the lack of affordable screening tests [2]. Group 2: Development of EchoNext - Researchers from Columbia University and NewYork-Presbyterian Hospital developed EchoNext, an AI tool that analyzes ECG data to identify SHD patients with higher accuracy than human experts [3][4]. - EchoNext aims to provide a cost-effective method to determine which patients require further expensive echocardiogram examinations [7]. Group 3: Performance and Validation of EchoNext - The AI model was trained on over 1.2 million ECG-echocardiogram pairs from 230,000 patients to detect various forms of SHD [10]. - In a validation study across four healthcare systems, EchoNext demonstrated high accuracy in identifying SHD, including conditions like heart failure and valvular disease [12]. Group 4: Clinical Application and Results - In a study involving nearly 85,000 patients who had not undergone echocardiograms, EchoNext identified over 7,500 individuals (9%) at high risk for undiagnosed SHD [13]. - Among those identified as high-risk, 55% underwent their first echocardiogram, with nearly 75% diagnosed with SHD, indicating a positive rate twice that of those without AI assistance [14]. Group 5: Comparison with Human Experts - A comparison of EchoNext with 13 cardiologists showed that while AI assistance improved the accuracy of human assessments, EchoNext outperformed human experts with an accuracy of 77.3%, sensitivity of 72.6%, and specificity of 80.7% [16][17].