基于深度学习的智能听诊技术在先天性心脏病诊治中的研究进展  

Advances in research on deep learning-based intelligent auscultation technology in the diagnosis and treatment of congenital heart disease

作　　者：赵宇航 王洋 陈润森 笪敏[1] 王智琪[1] 莫绪明[1] 戚继荣[1] Zhao Yuhang;Wang Yang;Chen Runsen;Da Min;Wang Zhiqi;Mo Xuming;Qi Jirong(Department of Cardiothoracic Surgery,Children’s Hospital of Nanjing Medical University,Nanjing 210000,China)

机构地区：[1]南京医科大学附属儿童医院心胸外科,南京210000

出　　处：《临床小儿外科杂志》2024年第7期633-640,共8页Journal of Clinical Pediatric Surgery

基　　金：江苏省卫生健康委科研项目(K2023036)。

摘　　要：先天性心脏病(congenital heart disease, CHD)是常见的出生缺陷之一, 随着人们对CHD认识的加深和筛查手段的不断进步, CHD患儿已能够在早期得到诊治。CHD早期筛查主要通过双指标检测法, 即心音听诊和脉氧测定。脉氧测定目前已有较成熟的商用设备, 但心音听诊受个人经验和外界因素的影响较大, 易出现误诊、漏诊。近年来随着人工智能(artificial intelligence, AI)的不断发展, 心音信号的数字化采集、存储、分析技术日趋成熟, 进而促使智能化心血管疾病听诊辅助诊断技术成为可能。目前基于深度学习(deep learning, DL)的AI算法在CHD心音听诊辅助诊断方面已有较多研究, 但大部分仍处于算法研究阶段, 且基于特定数据集实现, 尚未在临床得到验证。本文就目前AI听诊技术的研究现状进行综述, 总结近年来基于DL的智能听诊技术在CHD领域的发展, 并提出心音听诊在临床应用中亟待解决的问题。Congenital heart disease(CHD)is one of the most common congenital birth defects.With the increasing understanding of CHD and advancements in screening methods,children with CHD can be diagnosed and treated early,thereby improving survival rates and quality of life.Early screening for CHD primarily uses a dual-indicator method,namely heart sound auscultation and pulse oximetry.While there are mature commercial devices for pulse oximetry,heart sound auscultation is highly influenced by individual experience and external factors,making misdiagnosis and missed diagnosis more likely.In recent years,the continuous development of artificial intelligence(AI)has led to the maturation of digital acquisition,storage,and analysis of technologies for heart sound signals,thus making intelligent auscultation-assisted diagnosis for cardiovascular diseases possible.Currently,many studies focus on AI algorithms based on deep learning(DL)for assisting in heart sound auscultation diagnosis of CHD,though most remain at the algorithm research stage and are based on specific datasets,yet to be validated in clinical settings.This article reviews the current state of AI auscultation technology research,summarizing the current developments in DL-based intelligent auscultation technology in the field of CHD,and highlighting the issues that need to be addressed for clinical application of heart sound auscultation.

关 键 词：先天性心脏病 人工智能 深度学习 外科手术 儿童 

分 类 号：R54[医药卫生—心血管疾病]