以脈波參數實現無臂帶式血壓測量

Using Parameters of Pulse Waveform to Realize the Cuffless Blood Pressure Measurement

賴邵恒(SHAO-HENG LAI)

指導教授 ： 劉省宏

朝陽科技大學/資訊學院/資訊工程系/碩士(2019年)

摘要

由於穿戴技術的進步，許多生理監測儀器逐漸轉變為可穿戴式裝置。但是，血壓監測的儀器在目前消費市場中，依然採用臂帶的測量方式。現在，基於脈搏傳遞時間(PTT)的理論，開發了臂帶血壓測量技術，但其準確性並不好。根據心臟血流動力學理論，血壓與動脈特徵有關。因此，本研究的目的是利用光電容積描繪法與體阻抗容積描繪法，測量脈搏波，並擷取其特徵，利用多維度回歸模型與多層類神經網路估測血壓。經過評估後發現多維度回歸模型的測量誤差較小。有20名受試者參與實驗，並且透過運動來改變受試者的血壓。光電容積描繪法結果顯示，以最好的參數組合透過多維度回歸模型估測的收縮壓和舒張壓，其平均的均方根誤差分別為 5.877mmHg和4.334 mmHg，均優於單一參數脈波傳遞時間，7.796 mmHg和6.039 mmHg。體阻抗積描記法結果顯示，其平均的均方根誤差分別為 5.387mmHg和5.528 mmHg，均優於單一參數脈波傳遞時間，6.324mmHg和6.480 mmHg。透過多層類神經網路，其最好光電容積描繪法的參數組所估測的收縮壓與舒張壓的值為 8.080mmHg和4.620mmHg，均優於單一參數PTT， 8.214mmHg和4.648mmHg。

關鍵字

光電容積描繪法 ； 體阻抗容積描繪法 ； 多維度回歸模型 ； 多層類神經網路 ； 血壓測量

並列摘要

According to the advancement of wearable technology, many physiological monitoring instruments are gradually converted into wearable devices.However, the blood pressure monitor still is a cuff-type device in the consumer market, which also can not perform the beat-by-beat continuous blood pressure measurement. Recently, the cuffless blood pressure measurement has been developed based on the pulse transit time (PTT) but its accuracy is not better. According to the cardiac hemodynamic theorem, the blood pressure relate to the arterial characteristics. Therefore, the purpose of this study was to use the characteristics of the pulse wave measured by the photoplethysmography (PPG) and impedance-plethysmography (IPG) to estimate the blood pressure with a multi-dimension regression model and deep neural network. The contour of pulse wave has some characteristics of the artery. There were 20 subjects participating the experiment, and the blood pressure of the subject was changed by the exercise. PPGresults showed that the root mean square errors using multi-dimension regression model of the estimated systolic and diastolic pressures with the multi-parameters were 5.877mmHg and 4.334 mmHg and were better than those only using the PTTparameter, 7.796 mmHg and 6.039 mmHg. IPG results show that the multi-dimension regression model root mean square error of the estimated systolic and diastolic pressures with the multi-parameters were 5.387mmHg and 5.528 mmHg and were better than those only using the PTTparameter, PTT, 6.324 mmHg and 6.480 mmHg. PPG results showed that the best estimated error of the systolic and diastolic blood pressures with the deep neural network were 8.08 mmHg and 4.62 mmHg. Such results were better than those by using single-parameter PTT 8.21 mmHg and 4.64 mmHg.

並列關鍵字

photoplethysmography ； impedance-plethysmography ； multi-dimension regression ； deep neural network ； blood pressure measurement

參考文獻

[1] S. Majumder, T, Mondal, M.J. Deen, “Wearable sensors for remote health monitoring,” Sensors, vol.17, pp.130-174, 2017.

[2]K.K. Tremper, S.J. Barker, Pulse oximetry. Anesthesiology, vol.70, pp.98-108, 1989.

[3]S-H. Liu, J-J. Wang, C-H. Su, T-H. Tan, “Development of a patch-type electrocardiographic monitor for real time heartbeat detection and heart rate variability analysis,” J. Med. Bio. Eng., vol. 38, pp. 411-423, 2018

[4]S-H. Liu, Y-J. Chang, “Using accelerometers for physical actions recognition by a neural fuzzy network,” Tel. e-Heal., vol. 15, pp. 867-876, 2009,

[5]J.J. Carr, J.M. Brown, Introduction to biomedical equipment technology, 4th ed., Prentice Hall, 2001.