臺灣博碩士論文加值系統

心電圖的振幅會因為呼吸時胸腔體積的改變而造成心電圖的振幅變化。相關研究指出，調變的心電圖訊號與間接的呼吸量測技術，如氣動呼吸傳感器帶及呼吸感應性體積計，具有高度相關性。利用此特性從心電圖推導的呼吸訊號，簡稱EDR；基於此相關性，可以使用EDR監測睡眠呼吸狀態與診斷睡眠呼吸中止症。本研究將呈現一個自動化系統，能夠由心電圖訊號直接推導呼吸訊號以偵測睡眠呼吸中止症是否發生。本系統使用資料擷取裝置取得病人的心電圖資料，同時推導出EDR訊號；此外，本系統亦提供訊號處理的使用者介面，讓醫生可以用遠端連線的方式觀察到病患在睡眠呼吸中止症液生時的心電圖及EDR訊號。
本研究使用兩種EDR推導方法，第一種為振幅調解，另一種為使用cubic spline插補法；訊號處理驗證採用Physionet中睡眠呼吸中止症的心電圖資料與正常心電圖資料以比較此兩種方法的正確性。由結果可以發現，雖然由 plethysmograph所推導之呼吸間隔與EDR訊號間的相關性不高，但仍然存在相似性。此外，無論是由Physionet的心電圖資料或是實測的心電圖中推導的呼吸訊號，皆利用「呼吸擺動信號」來監測呼吸中止的症狀，此方法對於呼吸中止的檢測有很高的敏感度。儘管仍有改進的空間，但是本系統可以提供睡眠專家及醫生判斷睡眠呼吸中止症的必要資訊，特別是在長期的心電圖監控之下，由此系統可以增加診斷出睡眠呼吸中止症的正確率。

Electrocardiograph (ECG) amplitude is modulated through respiration due to thoracic cavity measurement variance with respect to the heart. This causes axis shifts in the ECG reference lead as the signal is being modulated. This modulating signal was found to have a correlation with certain indirect respiration techniques, such as pneumatic respiratory transducer belts and inductance plethysmograms. Termed ECG-Derived Respiration (EDR), this technique paves the way in improving sleep apnea screening and diagnosis using ECG signals because of this correlation. In this paper, a system intended for patients and sleep experts/physicians is presented where EDR signals are derived and screened for apnea using an automated technique. This system provides ECG acquisition that makes use of the National Instruments USB DAQ-6008 data acquisition device and the derivation of the EDR for the patient side, which is provided with a GUI for signal processing. A physician GUI is provided that gives remote access to the EDR and ECG signals for analysis and screening for sleep apnea.
Two EDR methods were used in this paper, one uses amplitude demodulation while the other one makes use of a cubic spline interpolation technique. These methods are tested using an ECG-apnea dataset and a normal ECG and respiration data set acquired from Physionet. It was found that although correlations of the EDR with inductance plethysmograph signal segments are low, there exist certain similarities in these segments with low correlation. EDR estimates from the Physionet records and also from our own gathered data are then screened for apnea using the derivation of a “breath swing signal”. It was found to be highly sensitive in detecting apneas when compared to reference annotations. Even though this is the case, the presented system as well as the presented EDR derivation methods is still subject to improvement, and later on will be able to provide sleep experts and physicians with necessary information in order to accurately score events as apnea, especially for long records of ECG data during sleep.

TABLE OF CONTENTS

摘要……………………………………………………………………………............................... i
ABSTRACT ………………………………………………………………………………………. ii
ACKNOWLEDGMENTS………………………………………………………………………... iii
TABLE OF CONTENTS…………………………………………………………………………. iv
LIST OF FIGURES……………………………………………………………………………...... vi
LIST OF TABLES ………………………………………………………………………………... ix

CHAPTER I INTRODUCTION…………………………………………………………………. 1
1.1 Background…………………………………………………………………………………….. 1
1.2 Motivation……………………………………………………………………………………… 2
1.3 Purpose…………………………………………………………………………………………. 3

CHAPTER II BASIC THEORY…………………………………………………………………. 4
2.1 Sleep Apnea Classification and Inductance Plethysmography………………………………… 4
2.2 Transthoracic Impedance Affects ECG Amplitude……………………………………………. 6
2.3 Noted Methods…………………………………………………………………………………. 6
2.3.1 Deriving Respiration Signal Using QRS Areas…………………………………….……. 6
2.3.2 Comparison of Methods………………………………………………………………….. 8
2.3.3 Deriving Respiration Signals during MR Examinations………………………………… 9
2.3.4 Running Average………………………………………………………………………… 12
2.4 Baseline Wander Removal……………….…………………………………………………….. 13
2.5 R-Wave Detection via So and Chan……………………………………………………………. 14
2.6 Cubic Spline Interpolation……………………….……………………………………………... 16
2.7 Breath Swing Signal Derivation for Detecting Sleep-disordered breathing…………………..... 17

CHAPTER III METHOD………………………………………………….................................... 20
3.1 Materials……………………………………………………....................................................... 20
3.1.1 Software………………………………………………….................................................. 20
3.1.2 National Instruments NI USB DAQ 6008……………………………………………….. 21
3.1.3 Acquiring Analog ECG Data…………………………………………………………….. 22
3.1.4 Thermistor for Respiration Signal……………………………………………………….. 22
3.1.5 Data Set for Evaluation – Fantasia Database and Apnea-ECG Database………………... 23
3.2 Method………………………………………………………………………………………….. 24
3.2.1 EDR Estimation…………………………………………………………………………... 24
3.2.2 NI USB DAQ and Visual C# 2008 with .NET Program Implementation……………….. 26
3.2.3 R-wave Detection………………………………………………………………………… 28
3.2.4 GUI/MySQL Database/PHP Site for Remote Access to Data…………………………… 30
3.2.4.1 Patient Side………………………………………………………………………… 31
3.2.4.2 PHP Site……………………………………………………………………………. 32
3.2.4.3 Remote GUI (Physician Side)……………………………………………………… 33
3.2.5 Breath Swing Signal Estimation for screening of Sleep Apnea………………………….. 33
3.2.6 Data Evaluation…………………………………………………………………………... 36
3.2.7 Evaluation of Method via Experiment…………………………………………………… 36

CHAPTER IV RESULTS……………............................................................................................ 38
4.1 GUI and PHP/MySQL.................................................................................................................. 38
4.1.1 Patient GUI................................................................................................................................ 38
4.1.2 Physician (with Remote Access) ........................................................................................ 40
4.1.3 MySQL Database and PHP Site.......................................................................................... 43
4.2 EDR1 and EDR2 derivation from the Fantasia Database............................................................. 46
4.3 EDR1 and EDR2 classification performance from Apnea ECG Database .................................. 48
4.4 EDR performance comparison with respiration signal from thermistor ...................................... 55

CHAPTER V DISCUSSION........................................................................................................... 64
5.1 Problems on R-wave Detection………………………………………………………………… 64
5.2 Explanation on effectiveness on EDR2 in detecting apneas …………………………………... 64
5.3 Determining the ECG lead to use ……………………………………………………………… 67
5.4 Discussion on Breath Swing Signal ……………………………………………………………. 69

CHAPTER VI CONCLUSION AND RECOMMENDATIONS …………………..................... 75
6.1 Conclusion…………………………………………………….................................................... 75
6.2 Recommendations……………………………………………………......................................... 76

REFERENCES……………………………………………………................................................. x

LIST OF FIGURES
Fig 2.1 Estimation of the direction of the mean cardiac electrical axis from measurements of QRS area ……………………………………...……………………………………………………... 7
Fig. 2.2 a) QRS area in V1 b) mean cardiac electrical axis measurement after spline interpolation c) PRT chest measurement ………………………....…………………………………….......... 7
Fig. 2.3 a) ECG after high-pass filtering b) A derived carrier signal c) signal produced by multiplication of b) by corresponding QRS peaks in a). d) A signal resulting after low pass filtering ………………………………………………………………………………...………. 9
Fig. 2.4 Felblinger and Boesch’s method in deriving respiration signal – analog method……….... 10
Fig. 2.5 Top Trace: Step function after sample and hold from ECG signal in Middle Trace. Bottom Trace: Analog demodulation after band pass filtering of top trace ................................ 11
Fig. 2.6 Block diagram of Arunachalam’s EDR Algorithm………………………………………... 12
Fig. 2.7 Two examples of obstructive sleep apnea as seen in the EDR. Duration: 3 minutes each. Apneas (indicated by arrows) in trace (a) were correctly interpreted in the blinded study, but those in (b) were missed …………………..……………………………….…………………. 18
Fig. 2.8 ECG derived respiration and its corresponding Breath Swing Signal .....………….…….. 18
Fig. 3.1 System Block Diagram…………………………………..………………………………... 20
Fig. 3.2 NI USB-6008/6009 Block Diagram……………..………………....................................... 21
Fig. 3.3. BioSenseTek ECG Signal amplifier (BeST AMP)……………………………………….. 22
Fig. 3.4 Thermistor TTF-103…………………………………………………..………………....... 23
Fig. 3.5 Thermistor Schematic applied to channel input ai0 of DAQ…….………………….......… 23
Fig. 3.6 Proposed EDR1 Signal Derivation Block Diagram ………….……………………............ 25
Fig 3.7 Proposed EDR2 Signal Derivation Block Diagram………………………..…………......... 26
Fig. 3.8 Setting up DAQ in Visual C# with .NET………………….................................................. 26
Fig. 3.9 Flow Chart for the program’s R-wave detection………………........................................... 28
Fig. 3.10 ECG record a01(blue) and its slope(green)………………………………………………. 29
Fig. 3.11 Block diagram for File Saving and Remote Access to data……………………………… 30
Fig. 3.12 Flow Chart for file saving in the Patient GUI……………………………………………. 31
Fig 3.13 PHP Site Flow Chart……………………………………………………………………… 32
Fig. 3.14 Flow Chart for opening a .txt file in the Physician Side GUI……………………………. 33
Fig. 3.15 Flow Chart for Peak Detection…………………………………………………………… 34
Fig. 3.16 Flow Chart for apnea screening using Breath Swing Signal………………………........... 35
Fig. 4.1 Patient GUI display after opening the .txt file …………………………………………….. 38
Fig. 4.2 File saved in database ……………………………………………………………………... 40
Fig. 4.3 Saved files under specified folder ………………………………………………………… 40
Fig. 4.4 Physician GUI after opening .txt file and screening for Apnea …………………………... 41
Fig. 4.5 Window indicating the AHI index, and condition ………………………………………... 42
Fig. 4.6 Saved files for the Physician GUI ………………………………………………………… 43
Fig. 4.7 MySQL Database using XAMPP server for Windows …………………………………… 44
Fig. 4.8 Table of documents (left) and members (right) table with the following format and structure ………………………………………………………………………………………... 44
Fig. 4.9 Member Login page ………………………………………………………………………. 45
Fig. 4.10 Table presented in the browser window providing file access …………………………... 46
Fig. 4.11 Respiratory Waveform (topmost) with EDR1 (0.36; middle) EDR2 (0.6; bottom) for record f1o01 ………………………………………………………………………………….... 47
Fig. 4.12 Respiratory Waveform (topmost) with EDR1 (0.23; middle) EDR2 (0.35; bottom) for record f1o07 (error in R-peak detection as clearly shown in EDR1 resulted in the large variation added for EDR2 due to spline interpolation at t = 3350-3400s) ……………………. 47
Fig. 4.13 Respiratory Waveform (top) with EDR1 (0.14; middle) EDR2 (0.13; bottom) for record f1y10 ………………………………………………………………………………………….. 48
Fig. 4.14 Record a01 EDR1(top) EDR2(center) resemblance with respiratory waveform (bottom) ………………………………………………………………………………………... 50
Fig. 4.15 Record a03 EDR1(top) and EDR2(center) similarity with Resp. waveform (bottom) ………………………………………………………………………………………... 50
Fig 4.16 Record c01 similarity between resp. (top) and EDR1 (Bottom) …………………………. 51
Fig. 4.17 a04 amplitude variance during apnea. EDR1 (top), EDR2 (middle), Resp (bottom)……………………………………………………………………………………….... 51
Fig. 4.18 Record c03 Resp (top; with added offset) EDR1 (center) and ECG peaks (noted by circles) …………………………………………………………………………………………. 52
Fig. 4.19 Record a01 correctly classified as apnea for both EDR …………………………………. 52
Fig 4.20 Record a01 incorrectly classified as normal for both EDR ………………………………. 53
Fig 4.21 Record a01 Resp waveform incorrectly classified as apnea but normal by Physionet ………………………………………………………………………………………. 54
Fig. 4.22 Record a02 Resp waveform incorrectly classified as apnea but normal by Physionet……………………………………………………………………………………….. 54
Fig 4.23 Record c01 Resp waveform incorrectly classified as apnea but normal by Physionet……………………………………………………………………………………….. 55
Fig. 4.24. R-Peaks detected on the same ECG Signal sampled at a) 1000Hz (top), b) 500Hz (middle), and c) 250Hz (bottom) in 1-min duration……………………………………………. 55
Fig. 4.25 Effect of LPF on R-waves on the ECG signal …………………………………………… 58
Fig. 4.26 Resp (top), EDR2 (middle), EDR1 (bottom) during normal breathing ………………….. 58
Fig. 4.27 Simulated Apnea EDR2 and Resp (top), EDR1 (bottom)….…………………………….. 59
Fig. 4.28 ECG and Respiratory Waveform during Normal breathing and apnea simulations …….. 59
Fig. 4.29 4-minute segment all correctly classified. Top Trace: Resp waveform from thermistor; middle: ECG; bottom: EDR1. It can be seen that the EDR waveform follows the modulating signal on the ECG during respiration ………………………………………………………….. 60
Fig. 4.30 EDR1 and EDR2 Incorrectly classified as Apnea, but should be normal. The respiratory waveform was classified as normal ………………………………………………………....... 61
Fig. 4.31 EDR1 classified as Apnea due to high peaks at normal breathing ………………………. 61
Fig. 4.32 Incorrectly classified as Apnea due to incorrect R-wave detection ……………………… 62
Fig. 4.33 4 minute segment correctly classified but at t=360-420, was classified as an apnea due to incorrect R-wave detection …………………………………………………………………. 62
Fig. 4.34 Incorrectly classified as normal due to inefficient baseline wander removal …………… 63
Fig. 5.1 Top EDR1 and its breath swing signal Bottom EDR2 and its breath swing signal ………. 65
Fig. 5.2 ECG peaks and EDR estimates for record a04 at time 2880 ……………………………… 66
Fig. 5.3 Troughs/minimums are not as prominent in the EDR2 at times t1=5113, t2=5118 and t3=5133 ………………………………………………………………………………………... 66

 
LIST OF TABLES
Table I. Average correlations and the starting time at which the correlation was recorded ……….. 46
Table II. Accuracy, Sensitivity and Specificity for EDR1 and EDR2……………………………… 49
Table III. Accuracy, Sensitivity and Specificity for Respiratory Waveforms ………………….…. 53
Table IV. R-Peak Differences using 1000Hz ECG Signal R-Peaks As Reference………………… 56
Table V. Values computed for the Student’s t-Test ……………………………. …………………. 57
Table VI. Accuracy, Sensitivity and Specificity for Respiratory waveform from thermistor, EDR1 and EDR2 …………………………………………………………………………………….... 60
 

REFERENCES
1)Arunachalam, S. P. (2009). Real-Time Estimation of the ECG-Derived Respiration (EDR) Signal using a New Algorithm for Baseline Wander Noise Removal. 31st Annual International Conference of the IEEE EMBS, Minnesota, USA, 5681-5684.
2)Babaeizadeh, S., Zhou, S. H., Pittman, S. D., White, D. P. (2011). Electrocardiogram-derived respiration in screening of sleep-disordered breathing. Journal of Electrocardiology 44, 700-706.
3)Bogdan, M. (2009). Measurement experiment, using NI USB-6008 data acquisition. University of Lucian Blaga Sibiu, Faculty of Engineering.
4)Boyle, J. Bidargaddi, N. (2009). Automatic Detection of Respiration Rate From Ambulatory Single-Lead ECG. IEEE Transactions on Information Technology in Biomedicine 13, 890-896.
5)Bragge, T., Tarvainen, M.P., Karjalainen, P.A. (2004). High-Resolution QRS Detection Algorithm for Sparsely Sampled ECG Recordings. Univ. of Kuopio, Dept. of Applied Physics Report.
6)Cretikos, M. A., Bellomo, R., Hillman, K., Chen, J., Finfer, S., Flabouris, A. (2008). Respiratory rate: The neglected vital sign. The Medical Journal of Australia 188, 657-659.
7)Cysarz, D., Zerm, R., Betterman, H., Frühwirth, M., Moser, M., Kröz, M. (2008). Comparison of Respiratory Rates Derived from Heart Rate Variability, ECG Amplitude, and Nasal/Oral Airflow. Annals of Biomedical Engineering 36, 2085-2094
8)De Chazal, Heneghan, C., Sheridan, E., Reilly, R., Nolan, P., O’Malley, M. (2003). Automated processing of the single-lead electrocardiogram for the detection of obstructive sleep apnoea. IEEE Trans. Biomed. Eng. 50, 686-696.
9)Ding, S. Zhu, X., Chen, W., Wei, D. (2004). Derivation of Respiratory Signal from Single-Channel ECGs Based on Source Statistics. International Journal of Bioelectromagnetism 6, 43-49.
10)Felblinger, J., Boesch, C. (1997). Amplitude Demodulation of the Electrocardiogram Signal (ECG) for Respiration Monitoring and Compensation during MR Examinations. Magnetic Resonance Medicine 38, 129-136.
11)Fisher, T. (1999). Butterworth/Bessel/Chebyshev Filters. University of York, Computer Science Dept. Accessed 14 Dec 2011. http://www-users.cs.york.ac.uk/~fisher/mkfilter/trad.html.
12)Fleetham MD, J. Ayas MD, N. (2006). Canadian Thoracic Society guidelines: Diagnosis and treatment of sleep disordered breathing in adults. Canadian Respiratory Journal 13, 387-392.
13)Furman, G.D., Shinar, Z., Baharav, A., Akselrod, S. (2005). Electrocardiogram Derived Respiration during Sleep. Computers in Cardiology 32, 351-354.
14)Iyengar, N. Peng, C.-K. (1996). Age-related alterations in the fractal scaling of cardiac interbeat interval dynamics. American Journal of Physiology 271, 1078-1084.
15)Kuhn, L., Rose, L. (2008). ECG Interpretation Part 1: Understanding Mean Electrical Axis. Journal of Emergency Nursing 34, 530-534.
16)Labunde, R. E. (2010). Ventricular Depolarization and the Mean Electrical Axis. Accessed 01 July 2012. Cardiovascular Physiology Concepts. Revised 7 July 2012. http://www.cvphysiology.com/Arrhythmias/A016.htm
17)Maier , C., Dickhaus, H., Laguna, P. (2005). Amplitude variability extraction from Multi-Lead Electrocardiograms for Improvement of Sleep Apnea Recognition. Computers in Cardiology 32, 355-358.
18)Mazzanti, B., Lamberti, C., de Bie, J. (2003). Validation of an ECG-Derived Respiration Monitoring Method. Computers in Cardiology 30, 613-616.
19)Mazeika, G., Swanson, R. (2007). Respiratory Inductance Plethysmography: An Introduction. Pro-Tech Services Inc. Mukiteo WA.
20)McKinley S, Levine M. Cubic Spline Interpolation. College of the Redwoods, 1998D http://online.redwoods.cc.ca.us/instruct/darnold/laproj/Fall98/SkyMeg/proj.pdf
21)McNames, J., Fraser, A. (2000). Obstructive sleep apnea classification based on spectrogram patterns in the electrocardiogram. Proc. Comput. Cardiol., 749-752.
22)Moody GB, Mark RG, Zoccola A, Mantero S. (1985). Derivation of respiratory signals from multi-lead ECGs. Computers in Cardiology 12, 113-116.
23)Moody GB, Mark RG, Bump MA, Weinstein, J., Berman, A., Mietus, J., Goldberger, A. (1986). Clinical validation of the ECG-derived respiration (EDR) technique. Computers in Cardiology 13, 507-510.
24)O’Brien, C., Heneghan, C. (2007). A Comparison of algorithms for estimation of respiratory signal from the surface electrocardiogram. Computers in Biology and Medicine 37, 305-314.
25)Pan, J., Tompkins, W. J. (1985). A Real-Time QRS Detection Algorithm. IEEE Transactions on Biomedical Engineering 32, 230-236.
26)Penzel, T. Moody, G.B., Mark, R.G., Goldberger, A.L., Peter, J.H. (2000). The Apnea-ECG Database. Computers in Cardiology 27, 255-258.
27)Penzel, T., McNames, J., de Chazal, P., Raymond, B., Murray, A., Moody, G. (2002). Systematic comparison of different algorithms for apnoea detection based on electrocardiogram recordings. Medical &; Biological Engineering &; Computing 40, 402-407.
28)Physionet (2011). CinC Challenge 2000 data sets: Data for development and evaluation of ECG-based apnea detectors. PhysioBank: Physiologic Signal Archives for Biomedical Research. Accessed 12 Dec 2011. http://www.physionet.org/physiobank/database/apnea-ecg/
29)Raymond, B., Cayton, R. (2000). Screening for obstructive sleep apnoea based on the electrocardiogram-The computers in cardiology challenge. Proc. Comput. Cardiol., 267-270.
30)Roello, M. (2005). Interpolation with Polynomials and Splines. The Code Project. Accessed 20 Dec. 2011. http://www.codeproject.com/KB/cs/SplineInterpolation.aspx
31)Schlichter, T. J. Digital Filter Design Using MATLAB. (Digital Filter Design using Matlab.pdf)
32)Singh, S., Sharma, U. (). MATLAB Based Digital IIR Filter Design. International Journal of Electronics and Computer Science Engineering 1, 74-82.
33)Simpson, D. M., de Stefano, A. (2004). Digitizing Signals – a Short Tutorial Guide. SCOPE 13, 9-16.
34)Shinar, Z. Baharav, A., Akselrod, S. (2000). Obstructive sleep apnea detection based on electrocardiogram analysis. Proc. Comput. Cardiol., 757-760.
35)Smith, I. Mackay, J., Fahrid, N., Krucheck, D. (2011). Respiratory rate measurement: a comparison of methods. British Journal of Healthcare Assistants 5, 18-23.
36)So, H. H., Chan, K. L. (1997). Development of QRS Detection Method for Real-Time Ambulatory Cardiac Monitor. Proceedings – 19th Annual International Conference IEEE EMBS, Chicago, USA, 289-292.
37)Tan K. F., Chan K. L. Choi, K. (2000). Detection of the QRS complex, P wave and T wave in electrocardiogram. Advances in medical signal and information processing 2000, 41-47.
38)Thomas, R. J., Mietus, Peng, C.K., Goldberger, A.L. (2005). An Electrocardiogram-Based Technique to Assess Cardiopulmonary Coupling During Sleep. SLEEP 28, 1151-1161.
39)Thomas, R. J., Mietus, J. E., Peng, C.K., Gilmartin, G., Daly, R.W., Goldberger, A.L., Gottlieb, D.J. (2007). Differentiating Obstructive from Central and Complex Sleep Apnea Using an Automated Electrocardiogram-Based Method. SLEEP 30, 1756-1769.
40)Travaglini, A., Lamberti, C. DeBie, J., Ferri, M. (1998). Respiratory Signal Derived from Eight-Lead ECG. Computers in Cardiology 25, 65-68.
41)Widjaja, D., Taelman, J., Vandeput, S., Braeken, M.A.K.A., Otte, R.A., Van den Bergh, B.R.H., Van Huffel, S. (2010). ECG-derived respiration: Comparison and new measures for respiratory variability. Computing in Cardiology 2010, 149-152.
42)Zhao, Y., Zhao, J., Li, Q. (2008). Derivation of Respiratory Signals from Single-lead ECG. International Seminar on Future BioMedical Information Engineering, 15-18.