【摘要】 基于心音图作出诊断是对临床听诊技术的一个有益补充，但已有心音图处理软件的功能主要集中在降噪、波形编辑和回放上，一般不作为诊断依据。HeartScreen软件的主要功能则是对心音进行自动化标注，并提取多种特征为医生提供辅助诊断信息。软件首先根据同步录制的心电信号分割出单独的心动周期，并定位主要心音起始点，再根据信号包络图定位心音终止点，最后对同类心音信号进行积累以提高信噪比。实验选取了5段心音信号，分别使用自动算法和人工方法判定心音终止时刻，两者的平均差异为2.7 ms，标准差为9.1 ms，软件生成的特征心音图也可作为临床诊断依据。该软件可满足心内科医生对心音图处理的要求，算法所标注的心音位置也与人工结果一致，因此在心内科辅助诊断领域有着较好的应用前景。更多还原

【Abstract】 The application of phonocardiogram(PCG) for diagnosis complements the traditional means of auscultation. Most commercialized PCG processing software, however, is of limited medical significance because they mainly focus on noise reduction, waveform editing and playback. The HeartScreen software is aiming at automatic heart sound annotation and feature extraction so as to provide more useful diagnostic information to doctors. Cardiac cycles are partitioned out of the simultaneously recorded ECG with the proposed software and the onset points of main heart sounds are located. The offset points of the sounds are then determined using the envelogram of the phonocardiogram. Moreover, the isolated cardiac cycles of the same class are summed up to improve signal-to-noise ratio. Five phonocardiogram sequences are used in the experiment, which are processed with both the proposed method and manual operation to locate the offset points of the constitute heart sounds. The average difference between the two approaches is 2.7 ms and the standard deviation is 9.1 ms. Besides, the typical phonocardiograms generated by the software are accepted by cardiologist as medical proofs for diagnosis. The functions of the software satisfy the needs of cardiologist. The annotations of the positions of the heart sounds made by the proposed method show conformity with manual results. As a result, the software holds the promise of being used in a wider range in computer-assisted diagnosis.更多还原