- 學位論文
使用集合經驗模態分解方法計算從單一導程心電圖與口鼻氣壓器訊號之呼吸頻率
Respiratory Rate Estimation from Single Lead ECG and Oral-nasal pressure signals using Ensemble Empirical Mode Decomposition
摘要
本文研究使用一組非侵入性生物醫學訊號監測呼吸的可行性。研究的訊號是心電圖（ECG）和口鼻氣壓（oral-nasal airway pressure）訊號。研究的目的是分別從這兩種訊號計算呼吸速率，其中ECG訊號來源裝置對於人工智能物聯網 AIoT 世代 ( Artificial Intelligent Internet of Thing)來說，在非臥床持續監測呼吸(Ambulatory continuous monitoring respiration)訊號是普遍採用的裝置。 這項研究的動機是希望利用非侵入性設備來持續監測呼吸活動，從而成為當前呼吸監測技術的替代方法，傳統監測可能會干擾自然呼吸，並且在某些應用（例如壓力測試或睡眠研究）中較難處理。另一方面，現行心電圖裝置已經在臨床常規中使用，從它們獲得的計算呼吸信息能產生附加值，該附加值能提供患者更完整的生理訊息。 在本研究中，採用了集成經驗模態分解(Ensemble Empirical Mode Decomposition)與巴特沃斯濾波器(Butterworth filter)和主成份分析 (principal component analysis) 、零交叉 (zero-crossing)與資料融合(data fusion)等技術相結合，來計算呼吸速率並相互必較。呼吸速率是嚴重疾病患者惡化的關鍵參數和睡眠呼吸暫停診斷的重要參數。根據實驗結果，在EDR (ECG Derived Respiration)方面，先利用 Vortal dataset 測試與訓練演算法獲得了每分鐘呼吸頻率的(平均絕對誤差, 均方根誤差)為（2.20, 2.92）BPM (Breath Per Minute)，從既有文獻報告得知本研究結果優於其他多數現有技術，接著再將完成的演算法測試 BIDMC dataset的加護病房病人資料也獲致(1.46, 2.61)的結果，證明本演算法的可靠性。 另一方面，針對口鼻壓力計本文也提出了一種自動計算呼吸速率的方法，可以應用在睡眠呼吸中止分析與在學術研究中作為與計算的呼吸速率比較之參考呼吸速率 (reference respiratory rate)。
並列摘要
This thesis investigates the feasibility of using a set of non-invasive biomedical signals to monitor respiration. The signals of interest are the electrocardiogram (ECG), and oral-nasal airway pressure (Paw) signals. The aim of study is to estimate breathing rates from these two signals respectively, which are the candidates of respiratory ambulatory monitoring devices and homecare in the AIoT (Artificial Intelligent Internet of Things) world. This study is motivated by the desirability of monitoring respiratory activity from non-invasive devices to be an alternative of current respiration-monitoring in clinical setting which may interfere with natural breathing, and are unmanageable in some applications such as stress test or sleep studies. Furthermore, if these noninvasive devices are those already used in the clinical routine, the respiratory information obtained from them represents an added value of which creates a more completed overview of the patient status. In this thesis, it used the Ensemble Empirical Mode Decomposition (EEMD) together with other techniques of digital filtering Butterworth filter, principal component analysis (PCA), zero-crossing and data fusion to estimate respiratory rate which is critical parameter of deterioration for patients of serious illness and diagnosis of sleep apnea. According to the results of experiments of using Vortal dataset for training the algorithm, it obtained (2.20, 2.92) breath per minute for the respiratory rate as their mean absolute error and root mean square error, outperforming other existing state-of-the-art methods. To verify the efficacy and reliability of the algorithm, it was tested again using BIDMC dataset and came up with the result of (1.46, 2.61) to further prove the performance of algorithm. Besides, in this thesis, it also proposed an automatic estimator to obtain accurate respiration rate from oral-nasal pressure, which is one of important factors in sleep apnea analysis. The respiratory rate measured by oral-nasal is usually as a reference compared to the estimated respiratory rate in the academic research.
參考文獻
延伸閱讀
- 陳奕瑋、莊育瑋、胡威志（2011）。建構呼吸肌電圖擷取平台及其在自主與非自主呼吸判斷之應用。先進工程學刊，6(2)，131-139。https://doi.org/10.29948/JAE.201104.0009
- 楊惠喬、劉士嘉、蔡耀德、翁紹仁、黃世哲、詹明澄（2021）。資訊系統應用於呼吸器脫離評估效益之探討。呼吸治療，20(2)，1-9。https://doi.org/10.6269/JRT.202107_20(2).0001
- Chiu, Y. F. (2013). 應用於即時人體呼吸特徵萃取之低複雜度超寬頻雷達訊號處理系統 [master's thesis, National Tsing Hua University]. Airiti Library. https://www.airitilibrary.com/Article/Detail?DocID=U0016-2511201311375756
- 屠翔鈺（2020）。結合靜態特徵與動態資料於長短期記憶之呼吸器脫離預測方法〔碩士論文，逢甲大學〕。華藝線上圖書館。https://doi.org/10.6341/fcu.M0707381
- Chiang, L. L., Yu, C. T., Liu, C. Y., Lo, Y. L., Kuo, H. P., & Lin, H. C. (2006). Six-month Nocturnal Nasal Positive Pressure Ventilation Improves Respiratory Muscle Capacity and Exercise Endurance in Patients with Chronic Hypercapnic Respiratory Failure. Journal of the Formosan Medical Association, 105(6), 459-467. https://doi.org/10.29828/JFMA.200606.0003