基於血管彈力波及脈衝傳遞時間之多變參數建立血 壓估算法 Development of a Blood Pressure Estimation Based on Femoral Artery Elasticity Wave and Pulse Transit Time Parameters Adviser: Huang Ji-Jer Presenter:Syu Hao-Yi Date:2014/03/12 OUTLINE 1. Introduction 1. Brief introduction 2. ECG, PPG ,FET background and principles 3. Research on motivation and purpose 2. Paper Review 3. Materials and Methods 1. 2. 3. 4. 5. 6. 7. 8. Sensor application Analog circuits Signal processing Digital Circuits GUI Parameter definition Software flowchart Regression analysis and estimation of blood pressure OUTLINE 4. Results and Discussion 1. 2. 3. 4. 5. 6. 7. Analog circuit analysis and measurement Digital filtering assessment and actual measurements Digital filtering results built on firmware Verification system NI/LabWindows GUI Auto gain control setting Experiments 5. Conclusions and Future Work 1.INTRODUCTION INTRODUCTION • Brief introduction – Health promotion and fertility decline in mainly reason for the problem of an aging population – An international reach 7% of the population aged 65 is an aging , 14% called for aged, 20% called for super-aged 出處:行政院經濟建設委員會人力規劃處101年7月23日 1.INTRODUCTION INTRODUCTION • Brief introduction – Cardiovascular defects and disease caused by diet on life 出處:行政院衛生署 1.INTRODUCTION INTRODUCTION • Brief introduction – Growing population, chronic diseases and functional disorders are increasing, except for medical services ,the home care services have been our needs Photo Source : http://www.2cm.com.tw/zoomin_content.asp?sn=0912020016 1.INTRODUCTION INTRODUCTION • ECG, PPG ,FET background and principles – ECG records the heart in nerve conduction arising from potential changes graphics – Potential change of the heart begins with the rhythm generator of sinoatrial node (SA node) Photo Source : http://www.bostonscientific.com/lifebeat-online/heart-smart/electrical-system.html 1.INTRODUCTION INTRODUCTION • ECG, PPG ,FET background and principles – The electrocardiogram (ECG) or elektrokardiogramm (EKG) is a medical standard for testing the human heart for defects and diseases – The typical amplitude of the R wave of the ECG signal is approximately 1 mV 1.INTRODUCTION INTRODUCTION • ECG, PPG ,FET background and principles – A photoplethysmograph (PPG) the use of light source to measure the blood volume change of the blood vessel – Each ECG signal cycle corresponds to a PPG signal Photo Source : INVESTIGATION OF NEW ELECTRO - OPTICAL TECHNIQUES FOR MONITORING PATIENTS WITH COMPROMISED PERIPHERAL PERFUSION IN ANAESTHESIA 1.INTRODUCTION INTRODUCTION • ECG, PPG ,FET background and principles – PPG signal amplitude with the blood out of the tissue will be directly proportional to the change Photo Source : INVESTIGATION OF NEW ELECTRO - OPTICAL TECHNIQUES FOR MONITORING PATIENTS WITH COMPROMISED PERIPHERAL PERFUSION IN ANAESTHESIA 1.INTRODUCTION • ECG, PPG ,FET background and principles – FET(Femoral artery elasticity wave) 1.INTRODUCTION • Research on motivation and purpose – Capture ECG 、PPG 、 FET signal characteristics(PAT、HR、TDB、FET) estimate blood pressure •ECG •PPG •FET The Circuits of Amp. Gain Offset and ADC NI/LabWindows GUI and AGC setting 、data save Regression analysis and estimation of blood pressure 2.PAPER REVIEW • Continuous and non-invasive blood pressure measurement methods – Tonometric Method – Vascular Unloading Method – Pulse Transit Time 2.PAPER REVIEW • Comparison of continuous blood pressure measurement 壓張式血 血管無 脈 衝 到 脈 衝 到 達 時 間 恆 壓 血壓之動 壓量測 負載法 達時間 間歇性校正 低壓 態低壓量 測 動 作 雜 ● ◎ ◎ ● ● ● 收縮壓 ● ● ● ● ● ● 舒張壓 ● ● ● ● ● 平均壓 ● ● ● ● ● 訊 動 態 連 ● ● ● ● 續 操 作 方 ● 便 ●:符合 ◎:部分符合 ● 2.PAPER REVIEW • Long-term monitoring topics for various applications about unconstrained or noncontact measurement of physiological signals – – – – Mobile Bed Chair Toilet 2.PAPER REVIEW • A Preliminary Study for Unconstrained Pulse Arrival Time (PAT) Measurement on a Chair – Proceedings of the 2005 IEEEEngineering in Medicine and Biology 27th Annual Conference Shanghai, China, September 1-4, 2005 – Ko Keun Kim1, Youngjoon Chee1, Juwan Park1, Jungsoo Kim1, Yong Kyu Lim1, and Kwang Suk Park2 2.PAPER REVIEW • In this paper, the correlation between the typical PAT and the unconstrained PATs (APATs) are compared • A-PATs: A-PATs were measured from the pressure pulse of the air cushion (APP) 2.PAPER REVIEW • To validate the correlation with typical PAT, the three feature points was used for the A-PAT as follows: 1) the first minimum point of the APP, 2) the zero crossing point of the APP, and 3) the steepest descent point of the APP 2.PAPER REVIEW • A Smart Health Monitoring Chair for Nonintrusive Measurement of Biological Signals – Hyun Jae Baek, Gih Sung Chung, Ko Keun Kim, and Kwang Suk Park, Senior Member, IEEE – IEEE TRANSACTIONS ON INFORMATION TECHNOLOGY IN BIOMEDICINE, VOL. 16, NO. 1, JANUARY 2012 2.PAPER REVIEW • PAT was substituted by RJ interval(RJI), defined as the time interval from ECG R peak to BCG J peak 2.PAPER REVIEW PAPER REVIEW1 • Effect of confounding factors on blood pressure estimation using pulse arrival time – This article has been downloaded from IOPscience. Please scroll down to see the full text article.2008 Physiol. Meas. 29 615 – (http://iopscience.iop.org/0967-3334/29/5/007) – Jung Soo Kim, Ko Keun Kim, Hyun Jae Baek and Kwang Suk Park 2.PAPER REVIEW 2.PAPER REVIEW • Enhancing the estimation of blood pressure using pulse arrival time and two confounding factors – This article has been downloaded from IOPscience. Please scroll down to see the full text article.2010 Physiol. Meas. 31 145 – (http://iopscience.iop.org/0967-3334/31/2/002) – Hyun Jae Baek, Ko Keun Kim, Jung Soo Kim, Boreom Lee and Kwang Suk Park 2.PAPER REVIEW • Methods – Clinical data from dental anesthesia – Data from unconstrained biological signal a) Application on a toilet seat measurement systems b) Application in a vehicle c) Application at a computer 2.PAPER REVIEW REVIEW2 • Results() – BP=a+b*PAT – BP=a+b*PAT+c*HR*d*TDB Performance enhancement for BP estimation by inclusion of confounding factors in clinical data under dental anesthesia. Performance enhancement for BP estimation by inclusion of confounding factors in the unconstrained data. 3.Materials and Methods • PPG sensor – Reflective Optical Sensor with Transistor Output – The CNY70 is a reflective sensor that includes an infrared emitter and phototransistor in a leaded package which blocks visible light •Detector type: phototransistor • Dimensions (L x W x H in mm): 7 x 7 x 6 • Peak operating distance: < 0.5 mm •Typical output current under test: IC = 1 mA • Emitter wavelength: 950 nm • Daylight blocking filter 3.Materials and Methods • ECG&FET Sensor – SCC Series pressure sensors – – – – 0...5 to 0...100 psi Low cost sensor element Internal temperature compensation Differential, gage and absolute – Disposable ECG electrodes pads 3.Materials and Methods • Analog circuits – ECG,PPG and FET block diagram of the measurement circuit ECG 3.Materials and Methods LP Filter(100Hz) HP Filter(0.5Hz) AGC Notch(60Hz) PPG ECG-RA FET ECG-RL LP Filter(22Hz) AGC Notch(60Hz) FET LP Filter(100Hz) HP Filter(0.5Hz) Notch(60Hz) AGC Microprocessor PPG&ECG -LA 3.Materials and Methods • Analog circuit – ECG、FET second-order high-pass circuit H f (S ) S 2 Gs 2 S 2 1 ch 2 R3 * R 4 * C1* C 2 3.Materials and Methods • Analog circuit – ECG、PPG、FET second-order low-pass circuit H f (S ) S 2 Gs 2 S 2 1 cl 2 R3 * R 4 * C1* C 2 3.Materials and Methods • Analog circuit – ECG、PPG、FET notch filter circuit H f (S ) S 2 Gs 2 S 2 1 cn 2 R3 * R 4 * C1* C 2 3.Materials and Methods • Analog circuits – Auto gain control block diagram Latch 8bit MSP430 4bit Offset Analog Signal Gain Gain Voffset 4051 R25 0 12 3 4 5 6 7 4051 0 12 3 4 5 6 7 ………… G6 G5 G4 Rg G7 Offset ………… G2 G1 G0 RmG3 4bit R51 R48 R50 OP R24 Vf OP R47 Vin R60 OP R49 R61 Differential Amplifier inverting amplifier OP Output inverting amplifier 3.Materials and Methods • Analog circuits – Offset circuits Rm R 47 R 48 ) * (5)) ( * (5)) 式1 R 48 Rm R 47 R 48 Rm R 47 Rm 150 100 Vf (( ) * (5)) (( ) * (5)) 式2 100 Rm 60 100 Rm 60 Vf (( R 61 R51 (1 ) 式3 R 49 R 61 R50 R51 Vo2 Voff 式4 R50 R 61 R50 R51 R51 Vo Vo1 Vo2 Vin * Voff 式5 R 49 R 61 R50 R50 R 61 R51 R51 Vin Voff (Vin Vf ) 式6 R50 R50 R50 Vo1 Vin 3.Materials and Methods • Analog circuits – Offset circuits Rm/ ohm 準位調整 24k 0.4347 30k 0.2631 36k 0.1020 40k 0.0000 47k -0.1690 51k -0.2606 56k -0.3703 62k -0.4954 3.Materials and Methods • Analog circuits – Gain circuits R 25 Rg ) * ( ) R 24 R 60 5k Rg Vgain Voffset * ( ) * ( ) 10k 1k Vgain Voffset * ( Rg/ohm 1k 3k 6k 10k 15k 20k 51k 100k 倍率 0.1 0.3 0.6 1 1.5 2 5.1 10 3.Materials and Methods • Signal processing – Type Cascade form and Direct form Butterworth design a five order low-pass filter five order highpass filter and notch filter – High-pass cut-off frequency (fc) of 0.5Hz to eliminate baseline drift – 60Hz stop-band filter (Notch) to filter out noise in utility power 3.Materials and Methods • Signal processing – Filter system transfer function M H ( z) m b z m m 0 N a z n 0 n b0 b1 z 1 bM z M 1 a1 z 1 a N z N n M N k 0 k 1 y (n) bk x[n k ] a k y[n k ] 3.Materials and Methods • Signal processing – Direct form structure diagram 3.Materials and Methods • Signal processing – Cascade form y1[n] y 2 [ n] x[n ] H1[ z] H 2[ z] ..... y[n] H Ns [z ] – Signal processing sub-process calculates its output into the relationship wk [n] yk 1[n] a1k wk [n 1] a2k wk [n 2] yk [n] b0k wk [n] b1k wk [n 1] b2k wk [n 2] 3.Materials and Methods • Signal processing – Cascade form structure diagram yk 1[n] 1 wk [n] a1k Z-1 a2 k Z-1 b0 k b1k b2 k yk [n] 3.Materials and Methods • Signal processing – High pass and low pass 5 order、notch in Matlab filter before in Matlab filter after in Matlab 3.Materials and Methods • Signal processing – Multi-scale Mathematical Morphology (3M filter), based on set operations, provides a way to analyze signals using nonlinear signal processing operators that incorporate the geometry information of the signal 3.Materials and Methods • Signal processing – The shape information of the signal is extracted by using a structure element to operate on the signal, such operators serve two purposes, i.e. extracting the useful signal and removing the artifacts 3.Materials and Methods • Signal processing – 3M Filter related equations Dilation : f g (n) (max i ) (( f ( n i ) g (i )) Erosion : fg (n) (min i ) (( f ( n i ) g (i )) Opening : f g (n) f g (g )( n) Clo sin g : f * g (n) fg ( g )( n) 3.Materials and Methods • Signal processing – In this stage, the 3M filter used to eliminate baseline drift and reduce interference noises (motion artifacts, muscle contraction) 3.Materials and Methods • Signal processing – PPG signal, PPG 1st derivative, PPG 2nd derivative, and the process of peak detection 3.Materials and Methods • Signal processing – Find out 1st derivative local maxima – To find zero-crossing point PPG st PPG 1 derivative local maxima nd PPG 2 derivative local maxima ,crossing zero 3.Materials and Methods • Digital Circuits – BP estimates System Architecture Physiological signal AGC AGC MSP430F5438A A/D 12bit DSP(Filter 、AGC) DSP UART NI/LabWindows GUI Matlab GUI& Analysis,SPSS 3.Materials and Methods • Digital Circuits – Firmware Flowchart Stat Intialization a.Port b.timer c.ADC AGC Notch HP 5 order LP 5 order Interface END 3.Materials and Methods • GUI – National Instruments/LabWindow CVI 9.0 訊號顯示 RS232設定 資料儲存 增益與準位調整 3.Materials and Methods • Blood pressure parameter definition – – – – PAT:Pulse arrive time HR:RRinterval TDB:Arterial stiffness index FET: femoral artery elastic wave time parameters HR PAT TDB FET 3.Materials and Methods • Software flowchart – Matlab GUI Stat Detrend Filter(LP、 HP、 Notch) Normalize Peak detector • Regression analysis and estimation of blood pressure – SPSS – About the correlation between the legs – BP = a + b * PAT + c * HR + d * TDB + e * FET END 4. Results and Discussion • Analog circuit analysis and measurement – ECG High pass filter – PSpice circuit simulation frequency response fch 1 0.53Hz 2 10k *10k * 30 * 30 4. Results and Discussion • Analog circuit analysis and measurement – ECG Actual measuring high-pass circuit Bode fch:0.58Hz 4. Results and Discussion • Analog circuit analysis and measurement – ECG Low pass filter – PSpice circuit simulation frequency response fcl 2 1 104.07 Hz 510 * 510 * 3 * 3 4. Results and Discussion • Analog circuit analysis and measurement – ECG Actual measuring low-pass circuit Bode fcl:86.7Hz 4. Results and Discussion • Analog circuit analysis and measurement – ECG Notch filter – PSpice circuit simulation frequency response 1 fcn 61.21Hz 2 * 26k * 0.1 4. Results and Discussion • Analog circuit analysis and measurement – ECG Actual measuring notch circuit Bode 4. Results and Discussion • Analog circuit analysis and measurement – PPG Low pass filter – PSpice circuit simulation frequency response fcl 2 1 22.96 Hz 2.1k * 2.1k * 3.3 * 3.3 4. Results and Discussion • Analog circuit analysis and measurement – PPG Actual measuring low-pass circuit Bode fcl:27.1Hz 4. Results and Discussion • Analog circuit analysis and measurement – PPGG Notch filter – PSpice circuit simulation frequency response 1 fcn 61.21Hz 2 * 26k * 0.1 4. Results and Discussion • Analog circuit analysis and measurement – PPG Actual measuring notch circuit Bode 4. Results and Discussion • Analog circuit analysis and measurement – The actual circuit diagram ECG FET(R) FET(L) PPG Main board 4. Results and Discussion • Digital filtering assessment and actual measurements – In the rear of the system analyzed using 3M filter to eliminate baseline measurements obtained with ECG and PPG signals Determine structural elements 4. Results and Discussion • Signal filtering results with R-wave detection 式5 4. Results and Discussion • Digital filtering assessment and actual measurements – In the better than part of 3M filter eliminates baseline IIR filter P[1, 2] 2 Ponesided ( )d,0 1 2 1 0Hz 0~0.5Hz 0.5~45Hz (watts/he (watts/hertz) (watts/her rtz) tz) IIR 0.00073 0.1660000 0.690046 Filter 23 6 3M 0.30950 Filter 00 0.0480000 0.052909 3 4. Results and Discussion • Digital filtering results built on firmware 1ms 4. Results and Discussion • Digital filtering results built on firmware Filter before in Matlab Filter after in Matlab Filter after in MCU 4. Results and Discussion • Verification system – Contact measurement ECG compared with commercially available instruments 4. Results and Discussion • NI/LabWindows GUI 4. Results and Discussion • Auto gain control setting – Gain and offset adjustment 0.1x 0.3x 0.6x 1.0x 4. Results and Discussion • Summary – Sensors application and analog circuit design – Filter circuit – Offset and Gain – AGC circuit – Signal processing and firmware build – IIR Filter – 3M Filter – NI/LabWindows GUI – Waveform display – AGC Adjustment – Data storage 4. Results and Discussion • Experiments – Four waveform measurement system – Signal feature extraction 4. Results and Discussion • Experiments – BP estimation parameters analysis 4. Results and Discussion • Experiments(FET correlation between the legs) – About the correlation between the legs – FET accounted for HR time proportion System Initial 系統量測(R-peak、FET(R)-peak、 FET(L)-peak) Over 3min HR FET 4. Results and Discussion • Experiments(FET correlation between the legs) – Total : ECG 、FET(R)、FET(L) peak number – Se : success rate – Corr : FET correlation between left and right legs Time Total(bea True(b ts) eats) Se Corr FET(R FET(L) )/RRI /RRI A 3min 198 194 0.979798 0.802293 0.440273 0.443683 B 3min 332 322 0.96988 0.911003 0.467319 0.451698 C 3min 225 223 0.991111 0.843615 0.424077 0.411738 D 3min 263 258 0.980989 0.814634 0.482105 0.460597 E 3min 252 249 0.988095 0.863384 0.437047 0.456417 4. Results and Discussion • Experiments(estimation blood pressure) – – – – – System Initialization Cuff pressure measuring Contact system and finomerter simultaneous measurement Cuff pressure measuring Over System Initial Cuff 量測 Rest 5min 2min 取標準差 取平均值 系統量測/動態式 血壓儀器量測 3min T:17min Rest Cuff 量測 Over 2min 5min t 4. Results and Discussion • Experiments(estimation blood pressure) – Finometer is a noninvasive hemodynamic cardiovascular monitor based on the measurement of finger arterial pressure 4. Results and Discussion • Experiments(estimation blood pressure) – The actual figure Cuff pressure Contact system Finometer 4. Results and Discussion • Experiments(estimation blood pressure) – Blue: Finometer pressure signal – Black: ECG Signal – Red: PPG Signal – Green: Right leg femoral artery pressure – Yellow: Left femoral artery pressure 4. Results and Discussion • Experiments(estimation blood pressure) – Single- and multiple-regression analysis – BP=a+b*PAT+c*HR+d*TDB+e*FAP 4. Results and Discussion 4. Results and Discussion • Experiments(estimation blood pressure) – Performance of cuffless BP estimation with regard to combinations of parameters used in regression analysis Variable PAT PAT+HR PAT+TDB PAT+FET PAT+HR+TDB PAT+HR+FET PAT+HR+TDB+FET R 0.93544 0.93619 0.93717 0.93687 0.93796 0.93756 0.93951 SBP R^2 0.87537 0.87678 0.87857 0.87803 0.88005 0.87932 0.88292 Adj.R^2 0.87478 0.87561 0.87741 0.87686 0.87832 0.87758 0.88066 R 0.59539 0.75456 0.62561 0.60788 0.77345 0.76168 0.77999 DBP R^2 0.37019 0.57162 0.40257 0.38567 0.60076 0.58208 0.61077 Adj.R^2 0.36708 0.56751 0.39664 0.37956 0.59499 0.57602 0.6032 4. Results and Discussion • Experiments(estimation blood pressure) – SBP Scatter plot 5mmHg 5mmHg 4. Results and Discussion • Experiments(estimation blood pressure) – DBP Scatter plot 5mmHg 5mmHg 4. Results and Discussion • Experiments(estimation blood pressure) – Boxplots 4. Results and Discussion • Experiments(estimation blood pressure) – Reproducibility of multiple regression analysis for BP estimation SBP 526.465 (1.3813 * PAT ) (0.0012 * HR ) (0.0158 * TDB ) (0.0241* FET ) DBP 265.741 (0.5681* PAT ) (0.0299 * HR ) (0.0358 * TDB ) (0.00267 * FET ) 1 0.922729043 0.9 0.8 0.847635012 0.746011445 0.680468163 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 SBP(Training) DBP(Training) SBP(Test) DBP(Test) 4. Results and Discussion • Experiments(estimation blood pressure) – Reproducibility of multiple regression analysis for BP estimation System Initial Cuff 量測 Rest 5min 2min 取標準差 取平均值 系統量測/動態式 血壓儀器量測 3min T:17min Rest Cuff 量測 Over 2min 5min t 0.804 0.795 0.710 4. Results and Discussion • Experiments(estimation blood pressure) Training – Reproducibility of multiple regression analysis for BP estimation Test SBP DBP 1 0.9 0.8 0.944 0.795 0.804 0.710 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 Training Test SBP DBP 4. Conclusions and Future Work • Conclusions – 3M Filter remove baseline drift better than IIR filter – FET high correlation between the legs average of 0.844 – Multiple regression analysis for SBP and DBP estimation ,r and r^2 of SBP was 0.939 0.882,r and r^2 of DBP was 0.779 0.610 – The estimated BP was compared with the measured BP simultaneously ,the performance results of this test set decrease a little 4. Conclusions and Future Work • Future work – PPG to be more stable – Build different filters in MCU – Analyze more subjects – FET possibility to replace the TDB parameters – Unconstrained system for BP estimation and application to home care system or firmware development References • • • • • • • • 中華民國2012年至2060年人口推計,行政院經濟建設委員會人力規劃處, 2012 http://www.bostonscientific.com/lifebeat-online/heart-smart/electrical-system.html P. Kyriacou, “Investigation of new clectro-optical techniques for monitoring patients with compromised peripheral perfusion in anaesthesia,” 2001 http://zh.scribd.com/doc/73084909/Differential-Auscultatory-Technique 徐大川,連續血壓系統的改良與驗證,國立中央大學電機工程學系碩士論 文,2008 游生益,非接觸式心搏率量測系統應用於心跳變益率監測,南台科技大學電 機工程研究所碩士論文,2012 J. S. Kim , K. K. Kim , H. J. Baek and K. S. Park "Effect of confounding factors on blood pressure estimation using pulse arrival time", Physiol. Meas., vol. 29, pp.615 -624 2008 H. J. Baek , K. K. Kim , J. S. Kim , B. Lee and K. S. Park "Enhancing the estimation of blood pressure using pulse arrival time and two confounding factors", Physiol. Meas., vol. 31, pp.145 -157 2010 References • • • • • Κ. L. Kim , Y. Chee , J. Park , J. Kim , Y. Κ. Lim and Κ. S. Park "A preliminary study for unconstrained pulse arrival time (PAT) measurements on a chair", 27th Annual International Conference of IEEE Engineering in Medicine and Biology Society, 2005 Hyun, J.B.; Gih, S.C.; Ko, K.K.; Kwang, S.P. A Smart health monitoring chair for nonintrusive measurement of biological signals. IEEE Trans. Inform. Technol. Biomed. 2012, 16, 150–158. 101年死因統計結果分析,行政院衛生署,2012 http://www.2cm.com.tw/zoomin_content.asp?sn=0912020016 Abe, H., Chen, W., Togawa, T., Maeda, T., & Arai, R. (2007, November). Development of a mobile phone based beat-by-beat sphygmomanometer. InInformation Technology Applications in Biomedicine, 2007. ITAB 2007. 6th International Special Topic Conference on (pp. 285-287). IEEE. References • Lee, J. S., Chung, G. S., Beak, H. J., Lim, Y. G., Lee, J. S., Jeong, D. U., & Park, K. S. (2009, November). A new approach of unconstrained sleep monitoring and pulse arrival time extraction using PPG pillow and CC-ECG electrode system. In Information Technology and Applications in Biomedicine, 2009. ITAB 2009. 9th International Conference on (pp. 1-3). IEEE. Thank you for your attention