先進的慢性肺阻塞疾病特徵辨識：嵌入式系統檢測平台的深度解析

簡易檢索 / 詳目顯示

回結果列表

研究生：	鄭皓姿 Cheng, Hao-Tzu
論文名稱：	先進的慢性肺阻塞疾病特徵辨識：嵌入式系統檢測平台的深度解析 Advanced COPD Feature Recognition: Insights from an Embedded System Detection Platform
指導教授：	周百祥 Chou, Pai H.
口試委員:	韓永楷 Hon, Wing-Kai 謝孫源 Hsieh, Sun-Yuan 李皇辰 Lee, Huang-Chen
學位類別：	碩士 Master
系所名稱：	電機資訊學院 - 資訊系統與應用研究所 Institute of Information Systems and Applications
論文出版年：	2024
畢業學年度：	112
語文別：	英文
論文頁數：	43
中文關鍵詞：	慢性肺阻塞疾病、穿戴式裝置、呼吸偵測、共振峰、深度學習、微型機器學習
外文關鍵詞：	COPD, Wearable device, Respiratory sound detection, Resonance peak, Deep learning, TinyML
相關次數：	點閱：79 下載：0
分享至:	分享至facebook 分享至twitter

查詢本校圖書館目錄查詢臺灣博碩士論文知識加值系統勘誤回報

為降低慢性肺阻塞疾病（COPD）患者的死亡率，本研究開發了一種穿戴式呼吸偵測裝置，期望藉此提高COPD的早期偵測機會。裝置原型設計中，研究使用兩個麥克風分別放置於使用者的上呼吸道與左右肺，收集肺部呼吸音訊。接著，通過共振峰抓取和深度學習技術，分析並找出COPD患者呼吸時的音訊特徵，並對音訊進行分類以判斷是否罹患COPD。
在論文中，首先介紹本研究的硬體設施，再將針對裝置使用系統進行伺服器端和本地端
的比較。伺服器端將深度學習模型儲存於伺服器，而本地端則採用TinyML技術，讓模型可直接儲存於微控制器（microcontrollerunit，MCU）。
在論文的最後部分，介紹了本研究的成果。在訊號前處理方面，本研究使用帶通濾波器去除非主流頻率成分，並使用線性預測（linearprediction）找出頻率共振峰。接著，利用卷積神經網路（convolutional neural network，CNN）提取共振峰特徵。相較於一般的音訊處理方法，本研究的方法使特徵更為清晰，且提供了更精確的數據，使模型辨識更加準確和迅速。此外，這種方法也滿足了穿戴式裝置對低延遲的需求。

This thesis proposes the use of a wearable 3-channel respiratory sound monitor for COPD detection. The wearable device collects respiratory sound from three microphones: one over the upper respiratory tract and two over the lower left and right lungs on the user’s back. Through resonance peak extraction and deep learning techniques, the study analyzes and identifies the acoustic features of COPD patients during breathing and classified the audio data to determine whether the user has COPD.
We compare the system usage between the server side and the local side. The server side stores the deep learning model on the server, while the local side uses TinyML technology, allowing the model to be stored directly on a microcontroller unit (MCU). We use a band-pass filter to remove non mainstream frequency components and employe linear prediction to identify frequency resonance peaks. Subsequently, we use a convolutional neural network (CNN) to extract resonance peak features. Compared to general audio processing methods, our approach makes the features clearer and provides more accurate data, enabling more precise and rapid model identification. Moreover, this method also meets the low-latency requirements of wearable devices.

Introduction 1
1 Background ............................................. 1
1.1 COPD: Causes and Symptoms ............................ 1
1.2 COPD Diagnosis and Classification .................... 2
1.3 Wearable Devices for COPD ............................ 4
2 Motivation ............................................. 5
3 Contributions .......................................... 6
3.1 COPDfeature extraction method ........................ 6
3.2 Simple respiratory monitoring ........................ 6
4 Thesis Organization .................................... 7

Background and Related Work 8
1 Background: Detection and Risk Management .............. 8
1.1 Underutilization of Spirometry ........................8
1.2 Early Symptoms Identification of COPD ................ 9
1.3 Risk Factors of COPD ................................. 9
2 Background: TinyML Application in Embedded Systems ..... 10
2.1 Integrating TinyML into Embedded Systems ............. 10
2.2 The Application of TinyML in Contemporary Medical Science 10
3 Related Work ........................................... 11
3.1 Existing COPD Wearable Detection Devices ............. 11
3.2 One-Channel Sound Logger ............................. 11
3.3 Multi-Channel Sound Logger- LEOSound ................. 12
3.4 Feature Extraction from Respiratory Sound ............ 13

Methodology 14
1 Hardware Design ........................................ 14
1.1 Microcontroller Unit ................................. 14
1.2 Peripheral Devices ................................... 15
2 Interaction between Hardware and Software .............. 15
2.1 Universal Asynchronous Receiver/Transmitter .......... 16
2.2 On-Server Interface .................................. 16
2.3 Local interface ...................................... 17
3 Software ............................................... 19
3.1 Data Pre-processing .................................. 19

Machine Learning 22
1 Database ............................................... 22
1.1 R.A.L.E. Database .................................... 23
1.2 ICBHI Database ....................................... 23
2 Synthetic Minority Oversampling Technique .............. 23
3 Convolutional Neural Network ........................... 24
4 Support Vector Machine ................................. 25

Results and Evaluation 26
1 Implementation of Existing Devices ..................... 26
2 Operation of Existing Devices .......................... 28
3 Results of Data-prepossess ............................. 28
3.1 Outcomes of Data Denoising ........................... 28
3.2 Outcomes of resonance peaks .......................... 29
4 Model Performance ...................................... 31
4.1 Performance of CNN model ............................. 31
4.2 Performance of SVM ................................... 34

Conclusions and Future Work 37
1 Conclusions ............................................ 37
2 Future Work ............................................ 37
2.1 Mobile Device Connection ............................. 38
2.2 Other Sensing Technology ............................. 38
2.3 Validation of COPD Detection Device .................. 38
                                

[3M,24] 3M, Littmann. 3M™ Littmann® Lightweight II S.E. Stethoscope. https://www.littmann.com/3M/en_US/p/d/b00037528/, 2024.
[AH71] B.S. Atal and S.L. Hanaver. Speech analysis and synthesis by linear prediction of the speech wave. The Journal of the Acoustical Society of America, 2(50), Aug 1971.
[AKA20] Gokhan Altan, Yakup Kutlu, and Novruz Allahverdi. Deep learning on computerized analysis of chronic obstructive pulmonary disease. IEEE Journal of Biomedical and Health Informatics, 24(5):1344–1350, 2020.
[AKS+23] M.A.Althobiani, B. Khan, A. J. Shah, Y. Ranjan, R. G. Mendes, A. Folarin, S. Mandal, J. C. Porter, and J. R. Hurst. Clinicians’ perspectives of wearable technology to detect and monitor exacerbations of chronic obstructive pulmonary disease: Mixed-method survey.
[Ard23] ArduinoModules. KY-037 high sensitivity sound detection module. https://arduinomodules.info/ky-037-high-sensitivity-sound detection-module/, Mar 2023. International Journal of Chronic Obstructive Pulmonary Disease, 2023(18):1401–1412, 07 2023.
[Bio17] Biomedical Health Informatics. ICBHI Respiratory Sound Database (the respiratory sound database- ICBHI 2017 Challenge). https://paperswithcode.com/dataset/icbhi-respiratory-sound-database, 2017.
[BPP+20] C.L. Bentley, L. Powell, S. Potter, J. Parker, G.A. Mountain, Y.K. Bartlett, J. Farwer, C. O’Connor, J. Burns, R.L. Cresswell, H.D. Dunn, and M.S. Hawley. The use of a smartphone app and an activity tracker to promote physical activity in the management of chronic obstructive pulmonary disease: randomized controlled feasibility study. JMIR Mhealth Uhealth., 8(6:e16203), 2020.
[CR20] J.Y. Choi and C.K. Rhee. Diagnosis and treatment of early chronic obstructive lung disease (COPD). J Clin Med, 9(11), Oct 2020.
[dHL+19] J. de Miguel-Díez, J. Hernández-Vázquez, A. López-de Andrés, A. Álvaro-Meca, V. Hernández-Barrera, and R. Jiménez-García. Analysis of environmental risk factors for chronic obstructive pulmonary disease exacerbation: A case-crossover study (2004-2013). PloS one vol., 5(14), May 2019.
[DRV22] Maha S. Diab and Esther Rodriguez-Villegas. Embedded machine learning using microcontrollers in wearable and ambulatory systems for health and care applications: A review. IEEE Access, 10:98450–98474, 2022.
[Hua23] Bo-Yan Huang(黃柏諺). Wearable stethoscope for continuous monitoring of respiratory diseases and wearer environment, 2023.
[JM14] C. Jácome and A. Marques. Computerized respiratory sounds in patients with COPD: A systematic review. COPD: Journal of Chronic Obstructive Pulmonary Disease, 12(1):104–112, 06 2014.
[KBW+14] U. Koehler, U. Brandenburg, A. Weissflog, K. Sohrabi, and V. Groß. LEOSound, an innovative procedure for acoustic long-term monitoring of asthma symptoms (wheezing and coughing) in children and adults. Pneumologie (Stuttgart, Germany), 68(4):277—281, April 2014.
[LCC+23] C. H. Lin, S. L. Cheng, C. Z. Chen, C. H. Chen, S. H. Lin, and H. C. Wang. Current progress of COPD early detection: Key points and novel strategies. International Journal of Chronic Obstructive Pulmonary Disease, 2023(18):1511–1524, 07 2023.
[LE23] Shashank Bangalore Lakshman and Nasir U. Eisty. Software engineering approaches for TinyML based IoT embedded vision: a systematic literature review. In Proceedings of the 4th International Workshop on Software Engineering Research and Practice for the IoT, SERP4IoT ’22, page 33–40, New York, NY, USA, 2023. Association for Computing Machinery.
[LLTY17] S-H. Li, B-S. Lin, C-H. Tsai, and C-T. Yang. Design of wearable breathing sound monitoring system for real-time wheeze detection. Sensors, 17(1), January 2017.
[MKBT23] Marc Miravitlles, Konstantinos Kostikas, Nikoletta Bizymi, and Nikolaos Tzanakis. A novel figure and algorithm for the gold abe classification. Archivos de Bronconeumología, 59(11):702–704, 2023.
[MKM+80] M. Mori, K. Kinoshita, H. Morinari, T. Shiraishi, S. Koike, and S. Murao. Waveform and spectral analysis of crackles. Thorax, 11(35):843–850, Nov 1980.
[R.A08] R.A.L.E. Repository. The R.A.L.E. Repository, hear respiratory sounds. http://www.rale.ca/, 2008.
[sci24] scikit-learn Developers. Plot classification boundaries with different SVM Kernels. https://scikit learn.org/stable/auto_examples/svm/plot_svm_kernels.html,2024.
[Sil24] Silicon Labs. EFM32GG S1 GG11 and GG12 32-bit Microcontrollers.https://www.silabs.com/mcu/32-bit-microcontrollers/efm32-giant-gecko-gg11/device.efm32gg11b420f2048gm64?tab=specs, 2024.
[ST 21] STMicroelectronics. STM32F427XX Datasheet, May 2021. Available at https://www.st.com/en/microcontrollers-microprocessors/stm32f427-437/documentation.html.
[ST 24] ST Microelectronics. Programmable Bluetooth LE 5.2 Wireless SoC. https://www.st.com/en/wireless-connectivity/bluenrg-1.html, 2024
[TCJ+16] T. Thap, H. Chung, C. Jeong, K. E. Hwang, H. R. Kim, K. H. Yoon, and J. Lee. High resolution time-frequency spectrum-based lung function test from a smartphone microphone. Sensors (Basel, Switzerland), 8(16), Agu 2016.
[UKC+22] C.Urban, A. Kiefer, R. Conradt, M. Kabesch, C. Lex, A. Zacharasiewicz, and S. Kerzel. Validation of the leosound® monitor for standardized detection of wheezing and cough in children. Pediatric Pulmonology, 57(2):551–559, Feb 2022.
[WLH+21] C. T. Wu, G. H. Li, C. T. Huang, Y. C. Cheng, C. H. Chen, J. Y. Chien, P. H. Kuo, L. C. Kuo, and F. Lai. Acute exacerbation of a chronic obstructive pulmonary disease prediction system using wearable device data, machine learning, and deep learning: Development and cohort study. JMIR Mhealth Uhealth, 9(5):e22591, May 2021.
[三23] 三軍總醫院. 慢性阻塞性肺病(COPD). https://wwwv.tsgh.ndmctsgh.edu.tw/PaIn/191/10040/29086/300, 12 2023.
[簡03] 簡郁芬，葉昱麟，張富勝. 慢性肺阻塞疾病的分期與治療新觀念. 台灣庭醫學醫學會,28(5):121–127, 05 2003.
[蕭04] 蕭光明. 慢性阻塞性肺疾病與睡眠 (COPD and sleep). https://www1.vghtpe.gov.tw/msg/%E6%85%A2%E6%80%A7%E9%98%BB%E5%A1%9E%E6%80%A7%E8%82%BA%E7%96%BE%E7%97%85%E8%88%87%E7%9D%A1%E7%9C%A0930316.htm, 03 2004.
[陳05] 陳煌麒主治醫師. 高醫醫訊 https://www.kmuh.org.tw/www/kmcj/data/9411/13.htm, 11 2005.

簡易檢索 / 詳目顯示

相關論文