Abstract: Heart failure is a global public health issue, with diagnosis depending on comprehensive assessment of cardiac dysfunction. With the advancement of biomedical technologies, utilizing biosignals for early diagnosis of heart failure has become a key strategy to improve patient survival rates and reduce treatment costs. In this context, the rapid development of deep learning technology has paved new paths for the detection of heart failure. This paper systematically reviews the latest advancements and applications of deep learning in heart failure detection. Firstly, it outlines the primary biomedical signals and public datasets involved in heart failure detection. Secondly, it provides a detailed analysis of the application and development of deep learning in the field of heart failure diagnosis, particularly emphasizing the capabilities of convolutional neural networks and long short-term memory networks in processing key biomedical signals such as electrocardiograms, heart rate variability, and heart sounds. It summarizes the advantages and limitations of these technologies, and compares the performance of various models. The article also explores the potential of hybrid models, constructed through the integration of various artificial intelligence technologies, to enhance detection accuracy and model generalizability, as well as how model interpretability can be utilized to increase the transparency of the detection process and boost physician trust. Finally, the paper summarizes the current research deficiencies and proposes future research directions, emphasizing the importance of interdisciplinary cooperation in advancing heart failure detection technology.

Key words: Heart failure, Biomedical signals, Deep learning, Convolutional neural network, Recurrent neural network, Hybrid network model

摘要： 心力衰竭是一种全球性的公共健康问题，其诊断依赖于对心脏功能障碍的综合评估。随着生物医学技术发展，利用生物信号进行心力衰竭的早期诊断已成为提高患者生存率和降低治疗成本的关键策略。在此背景下，深度学习技术的迅猛发展为心力衰竭检测开辟了新路径。文章系统综述了深度学习在心力衰竭检测中的最新进展和应用。首先，概述了心力衰竭检测所涉及的主要生物医学信号和公开数据集。其次，详细分析了深度学习在心力衰竭诊断领域的应用及其发展，特别是对卷积神经网络、长短期记忆网络在处理心电图、心率变异性、心音等关键生物医学信号的能力进行了深入分析，总结了这些技术的优势、局限，并对各类模型性能进行比较。文章还探讨了通过融合多种人工智能技术所构建的混合模型在提升检测精度和模型泛化能力方面的潜力，以及如何利用模型的可解释性来增加检测过程的透明度，提升医生的信任度。最后总结了当前研究存在的不足，并对未来研究方向提出展望，强调了跨学科合作在推动心力衰竭检测技术进步中的重要性。

关键词: 心力衰竭, 生物医学信号, 深度学习, 卷积神经网络, 循环神经网络, 混合模型