The study of phonocardiogram (PCG) in diagnosing valvular heart disease has gathered increasing attention over the past few years. Heart sound auscultation is performed at the primary care center by physician and the results are subjected to the skills and hearing ability. This has caused unnecessary referral and send home subject with potential heart disease. This issue has led to the establishment of standardized and computerized system to analyze the heart sound. This thesis investigates the optimal approach in establishing a reliable system to acquire and process heart sound to differentiate between normal and abnormal pattern. Previous studies are based on the analysis using heart sound that is recorded from single stethoscope which provides limited information regarding the heart disease. In this study, the recording based on four stethoscopes is used to record sound from four different valves with optimized analog instrumentation design. Beamforming algorithm is utilized to localize the actual source of the disease sound from all of the four recorded sound by focusing with respect to the angle of arrival of the desired disease signature. It is then followed by the implementation of Time Frequency (TF) algorithm with optimal Extended Modified B-Distribution (EMBD) kernel to suppress noises, analyze and represent the features. The experiments were conducted utilizing PCG signal that was recorded from real subject from Hospital Sultanah Aminah Johor Bahru. Each subject was screened by an echocardiogram machine. The disease was confirmed by cardiologist before the PCG recording procedure was performed. The result shows significant improvement in the quality of information that is preserved in the beamformed signal. The suggested framework is able to improve the heart murmur detection rate up to 95%. In conclusion, the localization of the exact location of the diseased sound has helped to improve the disease detection accuracy based on multi-point heart sound diagnostic system.
展开▼
机译：在过去的几年中，心电图（PCG）诊断瓣膜性心脏病的研究越来越受到关注。心脏听诊由医生在初级保健中心进行，其结果取决于技能和听力。这已导致不必要的转诊，并使潜在的心脏病患者回家。这个问题导致建立了标准化的计算机化系统来分析心音。本文研究了建立可靠的系统来获取和处理心音以区分正常模式和异常模式的最佳方法。以前的研究是基于使用单个听诊器记录的心音进行的分析，该分析提供了有关心脏病的有限信息。在这项研究中，基于四个听诊器的录音被用于通过优化的模拟仪表设计记录来自四个不同阀门的声音。通过相对于所需疾病特征的到达角度聚焦，利用波束成形算法来从所有四个记录的声音中定位疾病声音的实际来源。然后，使用带有最佳扩展修改B分布（EMBD）内核的时频（TF）算法来抑制噪声，分析和表示特征。实验是利用PCG信号进行的，该信号是从Sultanah Aminah Johor Bahru医院的真实受试者中记录的。通过超声心动图仪对每个受试者进行筛查。在进行PCG记录程序之前，心脏病已由心脏病专家确认。结果表明，波束成形信号中保留的信息质量有了显着提高。建议的框架能够将心脏杂音检测率提高到95％。总而言之，基于多点心音诊断系统，对病音确切位置的定位有助于提高疾病检测的准确性。
展开▼
