人工智能在呼吸音方面的研究进展

Research progress of artificial intelligence in breath sounds

听诊器是临床医师最常用的诊断工具,现代医学始于听诊器的发明。听诊器应用于临床已有200年历史,从最早的声学听诊器到现在的电子听诊器,其外形和传音方式经历了较大的变化,但其使用方式和临床使用价值并没有发生质的改变。随着信息技术的发展,人工智能应用于肺音的智能识别技术得到了更多的关注。本文介绍了听诊器的发展史、呼吸音的分类、传统听诊器与电子听诊器的优缺点比较和肺音智能识别技术发展的必要性如医疗资源不足、医师听诊存在主观性及电子听诊器的优势和需要改进的方向等问题;总结了电子听诊器辅助疾病诊断近年的研究进展;从肺音的标准化收集与建库、数字呼吸音的标注和定性、数据的处理、计算机肺音分析、电子肺音的可听和可视化展现等角度论述了肺音智能识别技术的现实可行性;分析了肺音领域研究目前存在的问题及改进建议;并展望了今后肺音智能识别技术的可能应用场景和使用价值。

As the most commonly used diagnostic tool for clinicians,the invention of the stethoscope is the origin of modern medicine.Stethoscopes have been used clinically for 200 years.From the earliest acoustic stethoscope to the latest electronic stethoscope,although its appearance and transmission mode have been great changes,its use-pattern and clinical value haven′t changed substantially.With the development of information technology,increasing clinical application of artificial intelligence recognition technology in lung sounds has been attention.This review introduces the history of stethoscope,the classification of breath sounds,the comparison between traditional stethoscope and electronic stethoscope,and the necessity of artificial intelligence recognition technology in lung sounds including inadequate medical resources and subjective doctor′s auscultation,as well as the advantages of electronic stethoscope and its issues requiring improvement.The research progress of electronic stethoscope assisted-disease diagnosis in recent years is summarized.The practical feasibility of lung sounds intelligent recognition technology is discussed from the perspectives of standardized collection and database construction of lung sounds,annotation and characterization of digital breath sounds,data processing,computer lung sound analysis,audibility and visualization of electronic lung sounds in this review.This review also analyzes the current problems in lung sound research and suggestions for improvement.At the end of the review,the possible application scenarios and application value of intelligent recognition technology of lung sound in the future are prospected.