臺灣博碩士論文加值系統

談到健康議題時，血壓是一個常被提及的關鍵指標。血壓不僅是身體健康的重要指標，更是許多嚴重疾病的主要風險因素之一，高血壓往往在不引起任何明顯症狀的情況下悄悄地影響我們的身體。
本研究提出一種方法，開發一套錄製裝置用以錄製壓脈帶氣囊的共振波資料，並將所收錄製的資料以形態學之方法用於一維卷積神經網路模型訓練之中，在不降低模型準確度的前提下，減少模型參數量，進行模型準確度及大小之評估，建立一個預測個體血壓的模型，使模型在樹梅派上能夠運行。
為了驗證所提出方法在預測血壓值的準確性，參考了ANSI/AAMI/ISO 81060-2:2019協定進行了測試。在這項測試中，招募了16位受試者，並記錄了他們的性別、年齡、臂圍和血壓讀數。本研究結果顯示，所提出的方法在收縮壓和舒張壓的預估上與共振法相比，平均標準差分別為3.00mmHg±5.70mmHg 和3.00mmHg±4.50mmHg，符合協議規定的標準，即≤5mmHg ±8mmHg。

Blood pressure is a crucial indicator of human health. However, due to various factors influencing blood pressure values, single measurements often fail to accurately reflect an individual's true blood pressure level. To assess individual blood pressure status more accurately, multiple measurements are often needed, with the average value taken as a reference.
This study proposes a method to develop a recording device for capturing oscillometric wave data from a pressure cuff bladder. The collected data will be used to train a one-dimensional convolutional neural network model using morphological methods. The goal is to reduce the model's parameters without compromising its accuracy, and to evaluate the model's accuracy and size. Ultimately, the aim is to establish a model that can predict individual blood pressure and can be implemented on a Raspberry Pi.
To comfirm the accuracy of the proposed method in predicting blood pressure values, tests were conducted following the ANSI/AAMI/ISO 81060-2:2019 protocol. In this test, sixteen subjects were recruited, and their gender, age, arm circumference, and blood pressure readings were recorded. The results of this study show that the proposed method, compared to the Oscillometry method, has an average standard deviation of 3.00mmHg±5.70mmHg for systolic pressure and 3.00mmHg±4.50mmHg for diastolic pressure, meeting the ANSI/AAMI/ISO 81060-2:2019 protocol, i.e., ≤5mmHg ±8mmHg.

摘要 i
ABSTRACT ii
致謝 iv
目錄 v
表目錄 viii
圖目錄 ix
第一章 緒論 1
1.1 前言 1
1.2 研究動機 2
1.3 論文結構 2
第二章 背景知識與相關文獻 3
2.1 血壓與高血壓之關係 3
2.2 血壓測量 5
2.2.1 聽診法 5
2.2.2 共振法 6
2.3 卷積神經網路 7
2.4 文獻回顧 8
第三章 系統架構與實驗方法 12
3.1 系統架構 12
3.2 硬體設計 13
3.2.1 壓脈帶壓力錄製 13
3.2.2 共振波錄製 14
3.3 軟體工具 17
3.3.1 共振波信號錄製PC端接收工具 17
3.3.2 深度學習網路訓練 18
3.4 實驗設計方法 19
3.4.1 實驗流程 20
3.4.2 共振波錄製流程 21
3.4.3 共振波資料擴增 22
3.4.4 一維卷積神經網路設計 25
3.5 裝置校正 27
3.5.1 ADC校正 28
3.5.2 錄製裝置校正 30
3.6 驗證方法 31
第四章 實驗結果與討論 32
4.1 受測者選擇 32
4.2 神經網路訓練結果 34
4.2.1 學習曲線 35
4.3 血壓分佈 36
4.4 實驗結果分析 37
4.5 1D CNN模型層數選擇探討 42
4.6 1D CNN超參數Batch Size探討 44
4.7 模型參數量與平台執行之比較 45
4.8 討論 46
4.9 研究局限性 47
第五章 結論與未來展望 48
5.1 結論 48
5.2 未來展望 49
參考文獻 51
附錄 55

參考文獻
[1] 中華民國防高血壓協會, "認識高血壓," 2016. [Online]. Available:http://hypertension.org.tw/know.php. [Accessed 06 May 2024].
[2]衛生福利部國民健康署, "每22分鐘就有1人死於心臟病「用心、懂心」從8項行為做起," 2023. [Online]. Available: https://www.hpa.gov.tw/Pages/Detail.aspx?nodeid=4705&pid=17535. [Accessed 06 May 2024].
[3]衛生福利部統計處, "111 年死因統計結果摘要表," 2023. [Online]. Available: https://www.mohw.gov.tw/cp-16-74869-1.html. [Accessed 06 May 2024].
[4]D. Souza, et al. “CrossTalk Opposing View: Bradycardia in the Trained Athlete Is Attributable to a Downregulation of a Pacemaker Channel in the Sinus Node.” The Journal of Physiology, vol. 593, pp.1749–1751, 2015
[5]Informed Health. "What Is Blood Pressure and How Is It Measured?," Nih.gov, Institute for Quality and Efficiency in Health Care (IQWiG), 2019.
[6]B. Williams, et al. “2018 ESC/ESH Guidelines for the Management of Arterial Hypertension,” European Heart Journal, vol. 39, no. 33, pp. 3021–3104, 2018.
[7]L. Charles, et al. “Secondary Hypertension: Discovering the Underlying Cause,” American Family Physician, vol. 96, no. 7, pp. 453–461, 2017
[8]C. Diaconu, et al. “Drug-Induced Arterial Hypertension – a Frequently Ignored Cause of Secondary Hypertension: A Review,” Acta Cardiologica, vol. 73, no. 6, pp. 511–517, 2018.
[9]D. Perloff, et al. "Human Blood Pressure Determination by Sphygmomanometry," Circulation, vol. 88, no. 5, pp. 2460–2470, 1993.
[10] 王瑋廷, "使用正向充氣估測方式之血壓量測裝置," 碩士論文,國立臺北科技大學, 2009.
[11]J. Carr, et al. "Introduction to Biomedical Equipment Technology," 2003.
[12]S. Kiranyaz, et al. "1D Convolutional Neural Networks and Applications: A Survey," Mechanical Systems and Signal Processing, vol. 151,p. 107398, 2021
[13]J. Sanchez, N. Soltani, R. Chamarthi et al., “A Novel 1D-Convolution Accelerator for Low-Power Real-time CNN Processing on the Edge,” 2018 IEEE High Performance extreme Computing Conference (HPEC), pp. 1-8, 2018.
[14]A. Argha, et al. “Artificial Intelligence Based Blood Pressure Estimation from Auscultatory and Oscillometric Waveforms: A Methodological Review.” IEEE Reviews in Biomedical Engineering, vol. 15, pp. 152–168,2022
[15]S. Lee, et al. "Deep Belief Networks Ensemble for Blood Pressure Estimation," IEEE Access, vol. 5, pp. 9962–9972, 2017.
[16]M. Choi, et al. "Oscillometry-Based Blood Pressure Estimation Using Convolutional Neural Networks," IEEE Access, vol. 10, pp. 56813–56822, 2022.
[17]N. Delfan, et al. "Oscillometric Blood Pressure Measurement Using a Hybrid Deep Morpho-Temporal Representation Learning Framework," ArXiv (Cornell University), 2023.
[18]Metrodyne Microsystem, "MPS-3117 datasheet," [Online]. Available: https://www.lcsc.com/datasheet/lcsc_datasheet_1811071040_Metrodyne-Microsystem-MPS-3117-006GC-A3_C242002.pdf. [Accessed 06 May 2024].
[19]Texas Instruments, "LM324 datasheet," [Online]. Available: https://www.jameco.com/Jameco/Products/ProdDS/838284.pdf. [Accessed 06 May 2024].
[20]Koge Micro Tech, "KPM14A datasheet," [Online]. Available: https://koge-europe.com/wp-content/uploads/datasheet-KPM14A.pdf. [Accessed 06 May 2024].
[21]STMicroelectronics, "STM32F429I-DISCOVERY Data brief," [Online]. Available: https://www.st.com/resource/en/data_brief/32f429idiscovery.pdf. [Accessed 06 May 2024].
[22]F. Santucci, et al. "Precordial Vibrations: A Review of Wearable Systems, Signal Processing Techniques, and Main Applications," Sensors, vol. 22, no. 15, p. 5805, 2022.
[23]T. Saramäki, et al. "Finite Impulse Reponse Filter Design," Researchportal.tuni.fi, John Wiley & Sons, 1993.
[24]維基百科, "有限脈衝響應," [Online]. Available: https://zh.wikipedia.org/zh-tw/%E6%9C%89%E9%99%90%E8%84%88%E8%A1%9D%E9%9F%BF%E6%87%89. [Accessed 06 May 2024].
[25]衛生福利部國民健康署, "國民營養健康狀況變遷調查成果報告2017-2020年," 2022. [Online]. Available: https://www.hpa.gov.tw/Pages/ashx/File.ashx?FilePath=~/File/Attach/15562/File_18775.pdf. [Accessed 06 May 2024].
[26]Z. Li, et al. "A Survey of Convolutional Neural Networks: Analysis, Applications, and Prospects. " IEEE Transactions on Neural Networks and Learning Systems, vol. 33, no. 12, pp. 1–21, 2021.
[27]M. Lin, et al. “Network in Network.” ArXiv.org, 2013
[28]ANSI Standard, "ANSI/AAMI/ISO 81060-2:2019," 2019. [Online]. Available: https://webstore.ansi.org/standards/aami/ansiaamiiso810602019. [Accessed 06 May 2024].
[29]維基百科, "最小平方法," [Online]. Available: https://zh.wikipedia.org/zh-tw/%E6%9C%80%E5%B0%8F%E4%BA%8C%E4%B9%98%E6%B3%95. [Accessed 06 5 2024].
[30]J. Zhuang, et al. “AdaBelief Optimizer: Adapting Stepsizes by the Belief in Observed Gradients,” Advances in Neural Information Processing Systems, vol. 33, pp. 18795–18806, 2020.
[31]X. Ying. et al. “An Overview of Overfitting and Its Solutions.” Journal of Physics: Conference Series, vol. 1168, no. 2, p. 022022, 2019
[32]夏嘉鎂, "基於卷積神經網路混合模型探討卷積深度與數量的影響,“ 碩士論文,高雄醫學大學, 2019.
[33]D. MASTERS, et al. “Revisiting small batch training for deep neural networks,” arXiv preprint, 2018.