在本論文中，利用信號處理技術分析血液透析病患動靜脈廔管的血流聲信號，並且提供一個偵測廔管狹窄與否的演算法，統計分析病患得到一結論。血流信號的組成包括了一個1~1.67Hz之間的基頻信號，以及一個頻寬隨血管的阻塞程度而定的載波，因此利用基頻資訊，在觀察其中載波的成分，可以得知血管狹窄的程度。而本演算法困難點在於信號與脈動有關的低頻成份(1~1.67Hz)以及與代表血流聲的高頻成份(300Hz以上)，其能量大小相差約100dB，高頻是判斷的依據，但其能量相對太小造成的正確判讀的困難性。在一開始，我們先以短時傅立葉轉換對信號進行分析，藉由觀察信號時頻域的能量分佈，探討廔管狹窄對血流聲音信號的影響。接著以設計的定位平均濾波器處理信號中的雜訊問題，再用數位濾波器來選取待測信號的特定頻寬成份。
我們的演算法判斷結果中，Miss Fire的機率在5 %以內，但是False Alarm的機率大約23%是需要以後再改進的部份。

In this thesis, we analyzed the acoustic signal from the hemodialysis patient’s arteriovenous graft, and developed an algorithm for stenosis detection. Based on our analysis, we found that their blood flow generates a baseband signal which is between 1 to 1.67Hz over a carrier whose frequency range is highly correlated to the stenosis level in the arteriovenous graft. The carrier signal is much weaker(less than -100dB) than the baseband signal, as a result it is very difficult to observe the carrier directly and come up with a reliable detection result. By analyzing the signal in the time-frequency domain, we design an align-and-average scheme to suppress possible noises, and based on our analysis and observation, an array of cascaded bandpass filters are applied for detection purpose.
Using our new technique, the results have been proved to within 5% of miss fire rate. The false alarm rate is about 23%, unfortunately. This leaves room for future improvement, however.

參考文獻

[1] 葉偉成/葉守正編譯，醫用物理學，合記圖書出版社

[2] Shung, K. K., Principles of Medical Imaging , Academic Press, San
Diego, 1992.

[3] National Kidney Foundation, “K/DOQI Clinical practice guidelines for vascular access,” Am. J. Kidney Dis., 37,pp.137~181,2001.

[4] Krivitski, N. M., D. MacGibbon, R. D. Gleed, and A. Dobson,
“Accuracy of Dilution Techniques for Access Flow Measurement During Hemodialysis, ” Am. J. Kidney Dis., 31(3), pp. 502~508, 1998.

[5] Tobin, R., “Wall Pressure Spectra Scaling Downstream of Stenosis
in Steady Pipe Flow,” J. Biomech., 9, pp. 633~640,1976.

[6] Fredberg, J., “ Origin and Character of Vascular Murmers:
Model Studies ,” J. Acoust. Soc. Am., 61, pp. 1077~1085,1977.

[7] Sato, T., Tsuji, K., Kawashima, N., Agishi, T., Toma, H., “ Evaluation of Blood Access Dysfunction Based on a Wavelet Transform Analysis of Shunt Murmurs, ” Journal of Artificial Organs, 9 (2), pp. 97~104, 2006.

[8] Jane, R., Rix, H., Caminal, P., Laguna, P., “Alignment Methods for
Averaging of High-Resolution Cardiac Signals_ A Comparative Study of Performance, ” IEEE Transactions on Biomedical Engineering , 38 (6), pp. 571~579,1991.

[9] Bataillou, E., Thierry, E., Rix, H., “Weighted Averaging Using Adaptive Estimation of The Weights, ” Computers in Cardiology, pp. 37~40,1992.

[10] Kotas, M. ,“Projective Filtering of Time-Aligned ECG Beats, ” IEEE Transactions on Biomedical Engineering ,51 (7), pp. 1129~1139,2004.

[11] L. R. Rabiner and R. W. Schafer, “ Digital Processing of Speech Recognition Signals, ” Prentice-Hall Co. Ltd, 1978.

[12] N.Erdol, C.Castelluccia and A.Zilouchian, “ Recovery of Missing Speech Packets Using the Short-Time Energy and Zero Crossing Measurements ”, IEEE Trans.Speech and Audio Processing,1(3), July 1993.

[13] Douglas O'Shaughnessy, “ Speech Communications :Human and Machine, 2nd edition ”, The institute of electrical and electronics Engineers, Inc, New York, 2000.

[14] 陳松琳，“以類神經網路為架構之語音辨識系統 ”，2002

[15] Navakatikyan, M.A., Barrett, C.J., Head, G.A., Ricketts, J.H., Malpas, S.C., “ A Real-Time Algorithm for The Quantification of Blood Pressure Waveforms ” IEEE Transactions on Biomedical Engineering, 49 (7), pp. 662~670,2002.

[16] A.V. Oppenheim, R.W. Schafer and J.R. Buck, “ Discrete-Time Signal Processing , 2nd edition ,” Prentice-Hall, 1999.

[17] Advantech Co., “ User Manual of USB-4711A,” Advantech Co.,
Nov.2007.

[18] 葉偉成/葉守正，醫用物理學，合記圖書出版社。
[19] 王小川，語音信號處理，二版” ，全華圖書股份有限公司，2007。
[20] 張智星，MATLAB程式設計及應用，清蔚科技，2004。

[21] 羅華強，信號處理：MATLAB的應用，全華圖書股份有限公司，民國92 年。
[22] 翁宗輝，「高雜訊人體QRS波偵測系統之研製」，國立台灣科技大學電機工程系所碩士論文，2003。
[23] 謝文豪，「利用頻寬偵測血液透析病患動靜脈廔管栓塞之技 術研究」，國立台灣科技大學電機工程系所碩士論文，2008。