大部分的呼吸機是以氣流、氣壓等物理參數做控制,但在臨床上還有許多不足的地方,本研究以生理訊號的觀點從肌電訊號上找出自主與非自主呼吸的差異,並期望以此來提供有用的資訊改善呼吸機的治療。 本系統以MSP430為核心發展一套呼吸肌電圖訊號擷取平台,擷取、紀錄訊號去除橫膈膜肌電訊號中的心電雜訊,並從訊號中計算出相關的生理資訊,如呼吸頻率、心率與心率變異等資訊,且透過USB將資料傳至電腦端並於電腦端作分析進一步之處理、分析(如肌電訊號的過零率及頻譜功率分析)。在肌電訊號的特徵值分析中,自主與非自主呼吸在三個頻段的頻譜功率皆呈現顯著差異(p<0.05),而過零率上則有明顯的自體差異。在特徵值的應用上,由於其判斷呼吸的閥值選定困難,所以本研究以半自動的方式做評估,而其正確率皆高於83%。在平台演算法的驗證上,本研究以呼吸氣流溫度比對演算法其正確性高達98.9%。 此外本研究以MIT-BIH資料庫的多重睡眠電圖驗證本研究的特徵值方法,不但都能反應出受測者的呼吸動作(特別是在呼吸窒息的狀況),在難以觀測的肌電訊號中也能看出其呼吸動作的特徵值變化。
Most existing ventilators are controlled by physical parameters such as the volume of air flow and air pressure. However, these parameters have their intrinsic limitations when applied to clinical applications such as the physiological adaptation control. Therefore, this thesis was aimed at identifying the differentiation between spontaneous breathing and compulsive breathing using electromyogram (EMG) signals to improve the schedule of lung ventilators. The respiratory EMG acquisition platform was integrated using a microprocessor (MSP430) to acquire, save and process the EMG signal. The processes of EMG signal were including the cancellation of ECG interference from diaphragmatic EMG and the calculation for physiological related information such as respiratory rate, heart rate and heart rate variability (HRV) etc. After all processes had been done on platform, the data could be conveyed to computer by USB for further analyses (such as the zero-cross rate and the spectrum of signal) and others application. There were significant difference (p<0.05) between spontaneous and compulsive breathing on the power spectrum. However, this thesis using zero-crossing rate for self pattern recognition. Furthermore, for breath detection using EMG signals, the accuracies were above 83% using semi-automatic method that was to set the threshold manually. A thermistor air flow of respiratory-detection was devised for the reference of the EMG respiratory-detection. The accuracies of described method was above 98.9%. Furthermore, the EMG-respiratory-detection method was checking against the downloaded data from HTUMIT-BIH Polysomnographic DatabaseUTH. Results showed that the method was sensitive to the muscular movement of respiration (particularly to the dyspnea data).