以影像-光體積描記訊號評估血壓脈衝傳遞時間

Evaluate blood pressure pulse transit time with image-photoplethysmography(IPPG) signals

吳宇倫(Yu-Lun Wu)

指導教授 ： 李世光

共同指導教授 ： 吳光鐘(Kuang-Chong Wu)

國立臺灣大學/工學院/應用力學研究所/碩士(2021年)

https://doi.org/10.6342/NTU202004490

摘要

血壓量測技術發展已久，從最早期的侵入式量測隨著居家醫療的倡導而普及到民眾的日常生活；現今工商社會眾人皆忙碌，因此未來血壓量測技術的發展必須滿足即時性、便利性與舒適度。現今常見的血壓量測技術大多使用氣囊或袖帶(Cuff)對量測部位進行非侵入式加壓(例如已經普及的電子血壓計)，然而加壓部位若有傷口則可能會為病患帶來不適，且充氣與放氣過程無法滿足即時性。
為滿足上述量測條件，本研究提出以受試者臉部與手部影像-光體積描記訊號(Imaging-Photoplethysmography, IPPG)之間的脈衝傳遞時間(Pulse Transit Time, PTT)帶入血壓回歸模型，期望以無傷口、非接觸式的光學方法來評估血壓。本研究設計以工業相機配合綠光濾波片(中心波長為525nm)和變焦鏡頭，結合LabVIEW使用者介面組成IPPG影像擷取裝置，並以商用儀器(BioRadio)獲得PPG參考訊號，再以電子血壓計量取參考血壓。進行IPPG影像擷取前，透過對影像進行網格切割與頻譜積分強度計算，確定感興趣區域(Region of Interest, ROI)分別位於臉部的鼻子與雙臉頰還有手掌心部位，再藉由Dlib五官辨識套件配合OpenCV套件繪製出ROI，最後藉由移動平均濾波器與帶通濾波器等影像處理方法獲得IPPG訊號。
本研究首先對IPPG訊號進行心率分析，由於其波形並非都典型存在收縮峰特徵，因此以一階微分後的波形抓取最大斜率點之間的時間差，再換算成心率值；結果顯示，與商用儀器測得的心率相比，絕大多數資料分佈在布蘭特-奧特曼圖的一致性界限內，說明本研究團隊開發的IPPG擷取裝置與商用儀器之間有一致性。在PTT的擷取上，本研究首先以商用儀器測得的PPG參考訊號，確定受試者在呈現打招呼的姿勢下，能夠穩定獲得參考PTT之後，再進行IPPG波形間脈衝傳遞時間的計算；分析結果得知，所有受試者絕大部分的參考PTT與藉由IPPG波形得到的IPTT(Image - Pulse Transit Time)，和樣本平均數之間的差異均不超過±1.96個標準差，顯示PTT與IPTT的變異與離散程度不會太嚴重，並且大部分資料皆分佈在布蘭特-奧特曼圖的一致性界限內，顯示IPTT與PTT之間潛在一致性。
最後在血壓模型評估方面，本研究探討IPTT與血壓之間的簡單線性回歸，與再多加入心率項的多元線性回歸，結果顯示大部分受試者的多元線性回歸模型，其均方根誤差(RMSE)有顯著下降且決定係數(R2)有一定程度的提升，顯示血壓預測值及實際值的擬合得到改善，並藉由F檢定及t檢定驗證了心率項在該模型的顯著程度及其對血壓的影響皆大於IPTT；因此以心率為自變數能夠提升血壓回歸模型適配度的論述，在統計學的結果與驗證上具有高支持度。

關鍵字

影像-光體積描記法(IPPG) ； 脈衝傳遞時間(PTT) ； 心率與血壓評估 ； 多元線性回歸分析

並列摘要

In recent years, Cardiovascular Diseases (CVD) have often been among the top causes of human death. In order to prevent the occurrence of such diseases, people need to keep track of their own health conditions and monitor physiological parameters such as heart rate, body temperature, and blood pressure. The blood pressure measurement technology has been developed for a long time, from the earliest invasive measurement to the people’s daily life with the advocacy of home medical care. Nowadays, everyone in the industrial and commercial society is busy. Therefore, the development of blood pressure measurement technology must meet the immediacy, convenience and comfort.
In order to meet the above measurement conditions, this study proposes to bring the pulse transit time (PTT) between two imaging-photoplethysmography (IPPG) signals into the blood pressure regression model. It is desirable to use a non-invasive, non-contact optical method to estimate blood pressure. This research combines an industrial camera with a green light filter (with a central wavelength of 525nm), a zoom len, a LabVIEW user interface (UI) to form an IPPG image capture device, and uses the BioRadio to obtain the PPG reference signal. At last, use the sphygmomanometer to measure reference blood pressure. Before IPPG image capture, by cutting the image grid to perform spectral integration intensity calculation, it is determined that the region of interest (ROI) is located on the nose and cheeks of the face, as well as the palm of the hand. ROI is drawn by Dlib kit and OpenCV kit, and finally the IPPG signal is obtained by image processing methods such as moving average filter (MA) and band pass filter.
This research first analyzes the heart rate of the IPPG signal. Because not all of the waveforms have typical systolic peak, the time difference between the maximum slope points is captured by the waveform after the first derivative, and then converted into the heart rate value. The results show that compared with the heart rate measured by commercial instruments, most of the data are distributed within the consistency limits of the Bland-Altman Plot (B-A plot), indicating that the IPPG capture device developed by our research team is comparable to commercial instruments. In the acquisition of PTT, this study first used the PPG reference signal measured by a commercial instrument to determine that the subject can stably obtain the reference PTT in the hello posture, and then calculate the pulse transfer time between the IPPG waveforms; The result shows that the difference between the reference PTT and the IPTT (Image-Pulse Transit Time) obtained by the IPPG waveform of all subjects and the sample average value is within ±1.96 standard deviations, showing that the PTT and the the degree of variation and dispersion of IPTT is not too serious, and most of the data are distributed within the consistency limits of the B-A plot, showing the potential consistency between IPTT and PTT. Finally, in terms of blood pressure model evaluation, this study explores the simple linear regression between IPTT and blood pressure, and the multiple regression with more heart rate items. The results show that the root mean square error (RMSE) of the multiple linear regression model of most subjects has been significantly decreased and the coefficient of determination (R2) has been improved to a certain extent, showing that the fitting of the predicted value of blood pressure and the actual value has been improved. The F-test and t-test verify that the significance of the heart rate itself in the model and its influence on blood pressure are greater than IPTT; Therefore, the argument that using the independent variable of heart rate can improve the fitness of the blood pressure regression model has a high degree of support in statistical results and verification.

並列關鍵字

image-photoplethysmography (IPPG) ； pulse transit time (PTT) ； heart rate and blood pressure evaluation ； multiple regression analysis

參考文獻

[1] World Health Organization. "The top 10 causes of death." World Health Organization,. https://www.who.int/news-room/fact-sheets/detail/the-top-10-causes-of-death (accessed May 24, 2021).

[2] 台北榮總心臟內科丁予安醫師. "台灣心臟病年輕化的危機." 台北榮總心臟內科丁予安醫師. https://www1.vghtpe.gov.tw/msg/%E5%8F%B0%E7%81%A3%E5%BF%83%E8%87%9F%E7%97%85%E5%B9%B4%E8%BC%95%E5%8C%96%E7%9A%84%E5%8D%B1%E6%A9%9F920306.htm (accessed May 28, 2021).

[3] 中華民國 衛生福利部. "108年國人死因統計結果." 中華民國 衛生福利部. https://www.mohw.gov.tw/cp-16-54482-1.html (accessed May 28, 2021).

[4] Y. Yu Sun and N. Thakor, "Photoplethysmography Revisited: From Contact to Noncontact, From Point to Imaging," IEEE Transactions on Biomedical Engineering, vol. 63, no. 3, pp. 463-477, 2016, doi: 10.1109/tbme.2015.2476337.

[5] P. A. J. "The Impact of Covid-19 and Survival Strategies of the Smart Mirror Market." Princy A. J. https://www.researchdive.com/blog/the-impact-of-covid-19-and-survival-strategies-of-the-smart-mirror-market (accessed May 28, 2021).