以手機實現智慧型無線心電圖儀之研究

Abstract

無線心電圖儀能夠透過無線網路即時傳輸心電圖，是遠距醫療照護應用上的關鍵技術。本研究具體實作基於JPEG-2000心電圖像壓縮後傳輸的無線心電圖儀，並於接收端伺服器開發基於二維小波轉換的身份辨識及異常心律偵測演算法，以避免其解碼程序延遲心臟病患就診治療的時間。於無線心電圖儀的傳送端，本研究移植JPEG-2000編/解碼器的開放原始碼Jsaper至嵌入式開發平台運行，以即時實現將一維心電圖訊號編碼壓縮成二維JPEG-2000影像的功能。於無線心電圖儀的解碼端，本研究利用JPEG-2000心電圖像中各子頻帶能量及原始心電圖訊號的RR區間組成的特徵向量，並分別以第k位最接近鄰居及類神經機率網路兩項技術實現身份辨識及異常心律偵測的功能。實驗結果顯示，針對多位受測者的心電圖訊號進行的身份辨識的正確率可高達90%以上，而其異常心律偵測亦可達到90%以上的正確率。

Mobile ECG system has the ability to transmit real-time Electrocardiography via the wireless network, and its value of application in telemedicine has been proved. In this research, we implemented a Mobile ECG system where the ECG is compressed according to JPEG-2000 standard before transmission. On the receiver side, efficient algorithms to realize user Identification and abnormal ECG detection are developed based on 2-D wavelet transform coefficients, such a technique allows to skip the inverse wavelet transform, and hence reduces the long delay incurred by JPEG-2000 decoding process. The proposed algorithms first apply the subband energies of JPEG-2000 encoded ECG along with RR interval of the original ECG to form feature vectors. The k-nearest neighbors method is applied to identify the user, and the probabilistic neuron network is applied to detect abnormal ECG. Experimental results show that the accuracy of user identification can achieve more than 90% and the use of probabilistic neuron network for abnormal ECG detection also performs with more than 90% of accuracy.