論文顯示區塊
論文名稱:應用人工智慧技術於肺音診斷系統
Application of artificial intelligent technology in diagnostics the pulmonary sounds
研究生:周承漢 Cheng-Han Chou
指導教授:陳美勇 Mei-Yung Chen
    學位類別:碩士(Master)
    學校名稱:國立臺灣師範大學
    記錄編號:GN0697730155
系所名稱:機電科技研究所
畢業學年度:98
      語文別:中文
關鍵字:肺音聽診 Pulmonary sounds analysis
類神經網路 Artificial neural network
小波轉換 Wavelet transform
接受器操作特性曲線 Receiver operating characteristic curve
全文說明:(本論文20150830公開)
電子全文

    論文頁數:94
摘要:摘要
　　胸腔聽診為診斷肺部病症的主要方法，醫生藉由聽診器聽取肺部聲音，憑藉其專業認知與經驗來判斷不同的肺音所代表的病症。在120 Hz以下的生理訊號是由心音與肺音組成，而人耳對於低頻的靈敏度不高，故造成醫生在聽診判斷上的困難。為解決此一問題，本研究的目的為建構多種肺音辨識系統，用來辨識肺泡音(vesicular breath sounds)，支氣管音(bronchial breath sounds)， 氣管音(tracheal breath sounds)，爆裂音(crackle)，哮喘音(wheeze)，喘鳴音(stridor)等六種常見肺音。
首先使用壓電麥克風與資料擷取卡NI-PXI 4472B擷取人體肺音訊號，並作訊號預處理。接著以小波轉換作為特徵擷取之方法，透過圖形監控軟體LabVIEW 設計小波轉換之架構，訊號分解後之六個頻段做標準差與平均值運算，以得十七個特徵值。在分類器方面，本研究以倒傳遞與學習向量量化類神經網路作為系統分類器之子系統，用以模擬網路之可行性與內部參數，再經由LabVIEW建構類神經網路，分別測試其網路分類率，最後整合各子系統並建構二階段式類神經網路，以提升系統之可靠度。由實驗結果顯示，相較於傳統聽診方式，本研究成功建構出一套多種肺音診斷系統，可正確地分類出六種常見肺部聲音，彌補人耳對於低頻靈敏度不高的缺點，並由圖形監控軟體LabVIEW建構人機介面，顯示肺音之頻譜、登記病歷資料等，可供醫生作為診斷肺部疾病病患之輔具。其結果顯示，本研究所建構之系統其辨識率可達95%。
關鍵詞：肺音聽診、類神經網路、小波轉換、接受器操作特性曲線。
  Chest auscultation is a main and efficient way to diagnose lung disease, it is a subjective process that depending on the physician’s experience and ability to differentiate between different sound patterns. Because physiological signals composed of heart sound and pulmonary sound are above 120HZ and the in sensitive of the human ear to the lower frequency, it is not easy to make diagnostic classification successful. In order to solve this problem, this study aims to construct a variety of pulmonary sound (PS) recognition system for classification of six different PS classes: Vesicular breath sounds, bronchial breath sounds, tracheal breath sounds, crackles, wheezes, stridor sounds.
  First, we use the piezoelectric microphone and data acquisition card NI-PXI 4472B to acquire PS signals, and signals preprocessing. The wavelet transform as feature extraction method, the PS signals were decomposed into the frequency subbands. Through statistical method we get the seventeen feature vectors which are used the neural network's input vector. This research used back-propagation (BP) neural network and learning vector quantization (LVQ) neural network to be  subsystem, and the two neural networks are integrated together as a two stage system that can increase the reliability. The neural networks' performance is verified by the receiver operating characteristic (ROC) curve. Comparing with traditional auscultation method, this study successfully construct a variety of pulmonary sound diagnostic system can correctly classify the six common pulmonary sounds. In this study, can be improved that human ear’s insensitive to the lower frequency, and show its pulmonary sounds wave, characteristic value and spectral analysis chart are shown by the human-machine interface design. By the research of this paper, the recognition rate of system is up to 95%.
    論文目次:目  錄
摘要      Ⅰ
ABSTRACT      Ⅱ
致謝      Ⅲ
目錄      IV
圖目錄      VIII
表目錄      XII
第一章  緒論      1
1.1  前言      1
1.2  研究動機與目的    3
1.3  文獻回顧    4
1.3.1電腦化肺音診斷系統    4
1.3.2 訊號處理    6
1.3.3 肺音辨識    6
1.4  本論文之貢獻    9
1.5  論文架構    9
第二章  理論基礎    11
2.1  呼吸系統    11
2.2  肺音的發生機制    12
2.3  肺音種類區分    12
2.3.1支氣管音    13
2.3.2氣管音    14
2.3.3肺泡音    14
2.3.4哮喘音    15
2.3.5爆裂音    16
2.3.6喘鳴音    16
2.4  呼吸音與病理關係    17
2.5  小波轉換    21
2.5.1小波轉換的演進    21
2.5.2連續小波轉換    23
2.5.3尺度縮放與時間平移    24
2.5.4離散小波轉換    24
2.6  類神經網路    26
2.6.1學習策略    29
2.6.2網路架構    30
2.6.3類神經網路的特性    31
2.6.4學習效能    32
2.7  數位訊號處理及取樣原理    32
2.8  接受器操作特性曲線 (ROC Curve )    35
第三章  系統架構    38
3.1  系統整體架構    38
3.2  系統硬體    38
3.2.1系統之感測器    39
3.2.2麥克風性能參數    40
3.3  資料擷取卡    42
3.4  軟體架構    43
3.4.1特徵擷取單元    43
3.4.2分類器單元    44
第四章  系統架構設計與配置    46
4.1  肺音濾波與放大電路    46
4.2  母小波函數    47
4.3  多尺度分解    51
4.4  特徵擷取    52
4.5  類神經網路架構    53
4.5.1倒傳遞類神經網路架構    53
4.5.2倒傳遞靈敏度    59
4.5.3學習向量量化類神經網路架構    61
4.6  系統之分類器設計    63
第五章  實驗結果與討論      65
5.1  壓電麥克風頻率響應檢測    65
5.2  數位放大濾波電路    67
5.3  特徵值轉換    69
5.4  小波基底測試    69
5.4.1 Daubechies母小波測試結果    73
5.4.2 Coiflet母小波測試結果    74
5.4.3 Symlet母小波測試結果    76
5.4.4母小波測試之結果討論    77
5.5  類神經網路架構性能模擬    78
5.6  網路準確率評估    79
5.6.1倒傳遞網路準確率評估    79
5.6.2學習向量量化網路準確率評估    83
5.7  接受器操作特性曲線分析    86
5.8  二階段式分類器    87
5.9  人機介面功能    88
第六章  結論與未來展望      90
參考文獻      92
    參考文獻:參 考 文 獻
[1]http://www.doh.gov.tw/cht2006/index_populace.aspx，行政院衛生署統計資料，上網日期:中華民國九十九年。
[2]H. Polat, I. Guler, “A simple computer-based measurement and analysis system of pulmonary auscultation sounds,” Journal of Medical System, pp.665-667, 2004.
[3]J. E. Earis, B. M. G. Cheetham, “Current methods used for computerized respiratory sound analysis,” European Respiratory Review, pp.586-590, 2000.
[4]J. C. Chien, M. C. Huang, Y. D. Lin, F. C. Chong, “A study of heart sound and lung sound separation by independent component analysis technique,” Proceedings of the 28th Annual International Conference of The IEEE Engineering in Medicine and Biology Society, pp.5708-5711, 2006.
[5]Z. Moussavi, D. Flores, G. Thomas, “Heart sound cancellation based on multi-scale products and linear prediction,” Proceedings of the 26th Annual International Conference of The IEEE Engineering in Medicine and Biology Society, pp.3840-3843, 2004.
[6]X. Lu, M. Bahoura, “An integrated automated system for crackles extrac-tion and classification,” Biomedical Signal Processing and Control, pp.244-254, 2008.
[7]H. T. Falk, W. Y. Chan, “Modulation filtering for heart and lung sound se-paration from breath sound recordings,” Proceedings of the 30th Annual International Conference of The IEEE Engineering in Medicine and Biology Society, pp.1859-1862, 2008.
[8]M. T. Pourazad, Z. Moussavi, F. Farahmand, R. K. Ward, “Heart sounds separation from lung sounds using independent component analysis,” Proceedings of the 27th Annual International Conference of The IEEE Engineering in Medicine and Biology Society, pp. 2736-2739, 2005.
[9]J. Gnitecki, Z. Moussavi, H. Pasterkamp, “Recursive least squares adaptive noise cancellation filtering for heart sound reduction in lung sounds recordings,” Proceedings of the 27th Annual International Conference of The IEEE Engineering in Medicine and Biology Society, pp.2416-2419, 2003.
[10]L. Y. Wang, H. Wang, H. Zheng, G. Yin, “Multi-sensor lung sound extrac-tion via time-shared channel identification and adaptive noise cancella-tion,” Proceedings of the 43th Annual International Conference of The IEEE Decision and Control, pp.3599-3604, 2004.
[11]P. A. Mastorocostas, J. B. Theocharis, “A dynamic fuzzy filter for separa-tion    of discontinuous adventitious sounds from vesicular sounds,” Computer in Biology and Medicine, pp.60-69, 2007.
[12]D. Stavrakoudis, P. Mastorocostas, J. Theocharis, “A pipelined recurrent fuzzy neural filter for the separation of lung sounds,” Proceedings of the 43th Annual International Conference of The IEEE Fuzzy Systems, pp.1-6, 2007.
[13]A. Kandaswamy, C. S. Kumar, R. P. Ramanathan, S. Jayaraman, N. Mal-murugan, “Neural classification of lung sounds using wavelet coeffi-cients,” Computer in Biology and Medicine, pp.523-37, 2004.
[14]Z. Dokur “Respiratory sound classification by using an incremental supervised neural network,” Springer-Verlag London Limited, pp.309-319, 2008.
[15]R. Folland, E. Hines, R. Dutta, P. Boilot, “Comparison of neural network predictor in the classification of tracheal-bronchial breath sounds by res-piratory auscultation,” Artificial Intelligence in Medicine, pp.211-220, 2004.
[16]S. Matsunaga, K. Yamauchi, M. Yamashita, S. Miyahara, “Classification between normal and abnormal respiratory sounds based on maximum li-kelihood approach based on maximum likelihood approach,” Proceedings of the Annual International Conference of The IEEE Acoustics, Speech, and Signal Processing, pp.517-520, 2009.
[17]A. Jain, J. Vepa, “Lung sound analysis for wheeze episode detection,” Proceedings of the30th Annual International Conference of The IEEE Engineering in Medicine and Biology Society, pp.2582-2585, 2008.
[18]朱家瑜譯，「理學檢查與健康評估」，藝軒圖書出版社，2001。
[19]F. John, J. A. Murray, “Textbook of respiratory medicine,” Philadelphia: W. B. Saunders, 1994.
[20]M. Misiti, Y. Misiti, G. Oppenheim, J. M. Poggi, “Wavelet: a new tool for signal analysis,” The Math Works, Inc., 2010.
[21]S. Kumar, “Neural Networks - A classroom approach,” McGraw-Hill, Inc., 2005.
[22]M. T. Hagan, H. B. Demuth, M. H. Beale, “Neural network design,” MHB, Inc. 2004.
[23]顧力栩、沈晉惠譯，「人工智慧:智慧型系統導論」，全華圖書股份有限公司，2007。
[24]汪惠健譯，「類神經網路設計」，新加坡商湯姆生亞洲私人有限公司，2004。
[25]張文恭、江昭皚譯，「運算放大器原理與應用」，格致圖書公司，1991。
[26]R. A. Hilgers, “Distribution-free confidence bound for ROC curves. me-thods of information in medicine,” pp.96-101, 1991.
[27]http://www.biopac.com/contact-microphone-specifications, BIOPAC Sys-tem, Inc., 上網日期:中華民國九十九年。
[28]劉文彩，「壓電式微麥克風之製程與分析」，逢甲大學機械工程研究所碩士論文，中華民國九十七年一月。
[29http://sine.ni.com/nips/cds/view/p/lang/zhs/nid/14903,National Instruments PXI-4472B, 2008.
[30]羅華強編著，「類神經網路-Matlab的應用」，高立圖書有限公司，2005。
[31]L. Steven, “Understanding Lung Sound,” Saunders, 2002.
[32]http://www.stethographics.com, stethographics, Inc., 上網日期:中華民國九十九年。