摘要
心音采集过程中混入的干扰噪声影响着心音诊断,目前多通过手动方式选择干扰较少的信号段做后续分析。为从采集信号中筛选出干扰最少、稳定性最强的最佳心音信号,提出一种最佳心音信号的自动选择方法。对采集的25例正常和119例患先天性心脏病儿童的心音信号,基于离散小波变换与哈达玛积相结合定位心动周期。根据心动周期信号的周期稳定性及功率谱密度相似性计算质量因子,将质量因子最大的连续3个心动周期信号作为最佳心音信号。由心脏病专家通过音频回放对信号选择的成功率和有效性进行评估。结果表明,最佳心音信号自动选择的成功率为95.83%,选择成功信号均包含对应疾病的典型听诊特点。该方法选择性能良好且自动执行,为心音信号的全自动分析提供参考。
Heart sound(HS)diagnosis is affected by interference noise mixed in the process of HS recording.HS signal segments with less interference are often selected manually for subsequent analysis at present.A novel method for automatically selecting the optimum HS signal with minimum interference and maximum stability from the collected HS signal is presented.The cardiac cycles of 25 healthy children and 119 children with congenital heart disease are located using discrete wavelet transform combined with Hadamard product.Quality index is calculated according to the cycle stability and power spectral density similarity of cardiac cycle signal,and the 3 consecutive cardiac cycle signals with maximum quality index are taken as the optimum HS signal.The success rate and effectiveness of signal selection are evaluated by the cardiologist through audio playback.The results show that the success rate of optimum HS signal selection is 95.83%,and that the selected signal contains the typical auscultation characteristics of the corresponding diseases.In conclusion,the proposed method has good performance and executes the HS signal selection automatically,providing a reference for the fully automatic analysis of HS signal.
作者
王佳明
张安琪
郝宏燕
王春云
陈思
WANG Jiaming;ZHANG Anqi;HAO Hongyan;WANG Chunyun;CHEN Si(Procurement Center,Nanjing Drum Tower Hospital,the Affiliated Hospital of Nanjing University Medical School,Nanjing 210008,China;National Research Center of Pumps and Pumping System Engineering Technology,Jiangsu University,Zhenjiang 212013,China)
出处
《中国医学物理学杂志》
CSCD
2022年第11期1401-1406,共6页
Chinese Journal of Medical Physics
基金
国家自然科学基金(51805218)。
关键词
最佳心音信号
心音自动选择
质量因子
小波变换
功率谱密度相似性
optimum heart sound signal
automated heart sound selection
quality index
wavelet transform
power spectral density similarity