心音訊號之無線監測與分析系統

Abstract

本研究之目的為設計一套心音訊號之無線監測系統，以電子式聽診器為基礎，量測心音聲波之連續訊號，配合以ZigBee通訊規範所設計之無線傳輸模組來建構系統無線傳輸的平台，再搭配個人電腦為伺服主機，利用其運算儲存能力對資料做後續的處理。整體系統可細分為類比訊號處理模組、數位訊號取樣模組、無線傳輸模組、伺服端即時顯示軟體等。在類比訊號處理模組中，使用自製的放大、濾波電路來建構，而數位訊號取樣及無線傳輸模組，使用MICAz無線收發模組與MIB510界面板來建置。在硬體的建置上，成功地將自製電子聽診器與無線傳輸模組整合，達到模組化與體積縮減之目的。本研究在伺服端的部份，利用C++ Builder視窗程式開發工具，撰寫心音監測系統的應用軟體，提供心音聲波訊號的顯示、放音、紀錄、分析與資料庫建立等功能。由於心音訊號的時域分析，第一心音與第二心音的標定為重要的工作，所以本研究建立心音訊號分離演算法則，能夠在多變的心音訊號中，以92.5%的正確率標示出第一心音與第二心音之主要心音成份於時間軸的位置，再以此對應位置的資訊為基礎，進一步能夠求出心音訊號型態上相關的特徵參數，且透過二次心音量測的實驗，計算心音型態特徵參數的實驗誤差。其後針對網際網路所提供的心音資料庫之心音檔案，進行系統程式運算，繼以統計與分析所求得的心音型態特徵參數，使系統主要能夠辨別pulmonic insufficiency與aortic insufficiency兩種異常心音。

The purpose of this research is to develop a wireless monitoring system of heart sound signals based on an electronic stethoscope. The wireless module that uses the ZigBee communication protocol serves as the wireless platform that transmits the heart sound signals to a desktop computer for further signal processing and analyses, taking the advantage of its computation and memory storage capability. The components of this system include an analog signal processing module, a digital signal sampling module, a wireless module, and a real-time display software. The analog signal processing module comprises amplifying and filtering circuits while the digital signal sampling module and wireless module are the wireless module of MICAz and the MIB510 interface board. The electronic stethoscope was successfully integrated with the wireless module to achieve a hand-held and modular design. An algorithm as well as a software was developed using C++ Builder which provides the functionality of heart sound analysis for the integrated wireless monitoring system. Since the first and second heart sound are the two most important features for heart sound analyses in time domain, we focused on developing an algorithm that automatically detects the positions of first and second heart sounds in continuous heart sound signals. The algorithm was tested with many heart sound signals and the accuracy of detection was 92.5%. Several simple heart sound features were then derived from a heart sound signal based on the first and second heart sound detected. Two sets of experiments were carried out to acquire heart sounds from various subjects. These heart sound samples together with the samples from internet databases were analyzed and compared using the developed software. We found that abnormal heart sounds such as the pulmonic insufficiency and aortic insufficiency can be classified with the algorithm.