《生命科学》 2022, 34(8): 909-917

人工智能在无创心血管影像中的应用

杨　靖^1,* , 张英梅²

¹复旦大学附属中山医院徐汇医院心内科，上海互联网医院工程技术研究中心，上海 200031 ²复旦大学附属中山医院心内科，上海市心血管病研究所，上海 200032

摘　要：

超声心动图、心脏磁共振、多层螺旋CT、心脏核素显像等无创心血管影像检查在心血管疾病诊疗的过程中发挥着关键作用。精准医学的模式改变对临床医师提出了新的要求。然而，影像解读的时间、效率和漏诊的问题依然是应用这些方法的巨大阻碍。人工智能的研究在过去十年取得了巨大的进步，尤其是基于卷积神经网络的深度学习应用于医学图像解读之后，人工智能支持的图像和信号分析在许多应用中已经达到了专家水平。该文讨论了人工智能在上述不同的影像模态的图像分割、自动测量、诊断、指导治疗和预测结局等工作流程中的最新应用前景，介绍了相关的人工智能算法和模型构建，并阐述了这些应用辅助临床决策的潜在价值。最后，该文讨论了人工智能方法目前存在的研究局限性和在真实世界中应用的问题，以及如何克服这些局限性的方案。

通讯作者：杨　靖 , Email:ema-co@163.com

Artificial intelligence in noninvasive cardiovascular imaging

YANG Jing^1,* , ZHANG Ying-Mei²

¹Department of Cardiology, Shanghai Xuhui District Central Hospital and Zhongshan-Xuhui Hospital, Fudan University, Shanghai Internet Hospital Engineering Technology Research Center, Shanghai 200031, China ²Department of Cardiology, Zhongshan Hospital, Fudan University, Shanghai Institute of Cardiovascular Diseases, Shanghai 200032, China

Abstract:

Noninvasive cardiovascular imaging modalities play a pivotal role in the diagnostic and therapeutic decision-making of cardiovascular diseases. The changing paradigm of healthcare to precision medicine places new demands on clinicians. However, problems with the timing, efficiency, and misdiagnosis of image interpretation remain remarkable obstacles to the application of these approaches. Research on artificial intelligence (AI) has made tremendous advance over the last decade. In particular, using deep learning based on convolutional neural networks for medical image interpretation, AI-powered analysis of images and signals has reached, even surpassed expertlevel performance. In this review, the latest applications of AI in the clinical workflow, such as image segmentation, automated measurement, diagnosis support, treatment guidance, and outcome prediction for the individual imaging modalities were discussed. The proposed AI algorithms and model construction and the potential added value of these applications in clinical decision support were also introduced. Finally, we discussed the current limitations of AI research in cardiovascular imaging, barriers to translation into clinical practice, and how they can be overcome.

Communication Author：YANG Jing , Email:ema-co@163.com