基于心音特征参数的婴幼儿先心病建模研究

In the developing countries，the incidence and mortality of congenital heart disease (CHD) last high. If missing the best treatment period at the age of 1～5 years old, CHD will cause seriously harm to children, even lead to death. Therefore,the timely diagnosis and suitable treatment of CHD become a social hot issue. The heart sound analysis is low-cost and noninvasive. And synchronous heart sounds from different auscultation areas can reflect the heart mechanical function state comprehensively and is helpful to provide the effective reference to exact CHD diagnosis. Considering that the CHD patients under 5 years old are always thin, timid and active, this project will design a four-channel heart sound acquisition system aiming to children under 5 years old firstly. And then the multidimensional features of clinical CHD heart sounds will be extracted combining with pathological information. Finally, on the basis of heart sound features’ distribution, the statistics method will be applied to build mathematical analysis models which are independent of maching learning methods, like support vector machine, for three typical CHDs with single defect, i.e. ventricular septal defect, atrial septal defect and patent ductus arteriousus, and also for tetralogy of fallot which is a common complex CHD case. According to the research on CHD modeling based on heart sound feature parameters, this project expects to distinguish the normal heart sound from the four typical CHD heart sounds and give a preliminary judgment for the size of the congenital defect within a range of statistical error, which will imporve the accuracy and reliability of the CHD diagnosis. In addition, the analysis modelbase will be built and shared via the internet for the experts in the heart sound research area.It will not only provide auxiliary analysis reference for CHD screening, early diagnosis and postoperative follow-up detection and promote the reasonable configuration of social medical resources, but also provide a communication platform and promote the practical process of achievement on heart sound analysis.

先天性心脏病（先心病）在发展中国家的发病率和死亡率居高不下，若不能在5岁之前确诊并妥善治疗，将对婴幼儿健康造成严重危害，甚至危及生命。因此，本项目拟针对5岁以下先心病患儿身体瘦小、胆小、好动的特点，设计婴幼儿先心病多听诊区心音同步采集方案；应用现代信号处理方法，提取先心病患儿心音的多维特征参数；采用数理统计方法，重点对室间隔缺损、房间隔缺损、动脉导管未闭以及法乐氏四联症四种典型先心病，建立脱离机器学习方法的数学分析模型。本项目预期在准确分析正常心音与四种典型先心病心音基础上，再对心脏的先天性缺损孔径大小进行一定统计误差范围内的预判断，进一步提高先心病分析准确率和可靠性。本项目拟建立基于心音特征参数的婴幼儿先心病分析模型库，为心音研究提供成果共享交流平台，促进心音研究成果的实用化进程，同时为我国先心病患儿的筛查、初期诊断及术后随访提供辅助性的分析参考信息, 推进有限医疗资源的科学化配置。

本项目以先天性心脏病（先心病）为关注点，以先心病筛查和心脏日常管理的社会需求为研究背景，通过心音信号对先心病的病理进行深入分析，为医生诊断提供有效辅助手段。.项目针对主要先心病患者群体，5岁以下患儿身体瘦小、胆小、好动的特点，采用舒适柔软的面料设计听诊服，针对患儿身量特征，研发幼儿听诊器，建立先心病幼儿听诊系统，可以实现婴幼儿先心病多听诊区心音的同步采集及分析；同时，为满足家庭日常监护的需要，还研发了单导无线心音听诊系统，可以方便地采集心音信号了解心脏功能状态。.项目从数字信号处理的各个角度，深入进行心音特征提取算法的研究。从时域、频域、时频域、能量、高维度等多角度提取先心病患儿心音的多维特征参数。从先心病病理特征入手，对特征参数的物理意义进行提取，从心音信号这一简单但信息量丰富的生理信号入手，对先心病病情进行解读。对多维特征参数进行特征分析和筛选，针对病情特点，提取有效特征参数组合。.项目重点对室间隔缺损、房间隔缺损、动脉导管未闭三种典型单一型先心病以及法乐氏四联症这一典型复合型先心病，以提取的有效心音特征参数组合为基础建立并优化数学分析模型。同时，对心脏的先天性缺损孔径大小及复杂先心病中各关联病症病理信息进行细化研究，可以对临床先心病病情跟踪随访提供有效参考。.本项目完成计划的研究内容，发表相关论文18篇，其中SCI收录5篇。获得国家发明专利授权一项，实用新型专利授权一项。和合作单位联合申请并获得军队科技进步二等奖一项。.本项目成果为基于心音信号的先心病分析研究提供了坚实的理论研究基础，为幼儿先心病筛查及心脏病这一社会型慢性病预防提供有效参考，为推动社会大健康进程贡献一份力量。