急性低血壓(Acute Hypotension)可能導致心臟衰竭、急性腎衰竭等併發症，常常在短時間內威脅患者的生命。如何提前準確地預測急性低血壓已經成為了醫學界和工程界廣泛關注的臨床熱門議題。如果能提前預警可能發生的低血壓事件，臨床醫生便能盡早採取相應的措施以避免風險的發生與影響。
近年來，得益於機器學習(Machine Learning, ML)、深度學習(Deep Learning, DL)的發展。這些技術得以被應用至各種不同的醫療場景，作為創新、有效且可行的解決方案。主流研究透過從侵入式量測的血壓提取特徵，將其用作預測未來可能發生的急性低血壓。而另一個受到廣泛關注的研究領域則是透過光體積變化描記圖法(Photoplethysmography, PPG)、心電描記術(Electrocardiography, ECG/EKG)等非侵入式生理訊號估計動脈壓(Arterial Blood Pressure, ABP)。
綜合上述，本研究以基於非侵入式訊號的動脈壓估計方法作為基礎，進一步開發了非侵入式的急性低血壓連續監測方法。將深度學習中，電腦視覺領域常用於影像分割的經典模型Unet修改適用於1維表示的時序波形資料，結合遷移學習、多任務學習等方法，從相關生理訊號中學習對應的潛在映射關係。並將擬合完成的模型用於自動化特徵提取，透過該特徵表示對未來可能發生的急性低血壓事件做預測。
本研究的實驗設計基於真實加護病房(Intensive Care Unit, ICU)臨床收集的大型資料庫MIMIC，以模擬實際應用場景的方式，探討了基於非侵入式訊號之連續低血壓監測方法的可行性與限制，並嘗試透過簡單有效的校準(Calibration)方式，改善模型的泛用性能。理想測試環境的實驗結果顯示本研究提出的方法具有相當程度的潛力，而所提出的校準方法也有助於改善實際應用時的模型預測性能。

Acute hypotension can lead to complications such as heart failure and acute kidney injury, often posing a life-threatening risk to patients in a short timeframe. The accurate prediction of acute hypotension has become a widely discussed clinical issue in both the medical and engineering communities. Early warning of potential hypotensive events could enable clinicians to take timely measures to mitigate risks and their impact.
In recent years, advancements in machine learning (ML) and deep learning (DL) have opened doors for innovative, effective, and feasible solutions across various healthcare scenarios. Mainstream research has focused on extracting features from invasively measured blood pressure for predicting future episodes of acute hypotension. Another extensively studied area involves estimating arterial blood pressure (ABP) through non-invasive physiological signals like photoplethysmography (PPG) and electrocardiography (ECG/EKG).
This study builds upon non-invasive ABP estimation methods and further develops a non-invasive continuous monitoring approach for acute hypotension. It adapts the classic Unet model, commonly used in computer vision for image segmentation, to 1D representations of temporal waveform data within deep learning. Techniques such as transfer learning and multitask learning are employed to learn underlying mapping relationships from relevant physiological signals. The fitted model is used for automated feature extraction, which is then utilized for predicting future acute hypotensive episodes.
The experimental design of this study is based on a large database, MIMIC, collected from real Intensive Care Unit (ICU) clinical settings, simulating practical application scenarios. It investigates the feasibility and limitations of continuous non-invasive monitoring methods based on physiological signals and attempts to enhance the model's generalizability through a straightforward and effective calibration process. The results in an ideal testing environment demonstrate a significant potential for the proposed method in this study, and the calibration techniques suggested also contribute to improving the model's predictive performance in real-world applications.

論文審定書 i
摘要 ii
Abstract iii
目錄 iv
圖次 vii
表次 viii
第一章　緒論 1
1.1 研究背景 1
1.2 研究動機 1
1.3 研究目的 3
第二章　文獻探討 4
2.1 血壓估計(Blood Pressure Estimation) 4
2.1.1 收縮壓/舒張壓/平均動脈壓(SBP/DBP/MAP) 4
2.1.2 連續動脈壓(ABP)波形 4
2.2 實時低血壓事件預測(Real-Time Hypotension Prediction) 5
2.2.1 基於侵入式生理訊號 5
2.2.2 基於非侵入式生理訊號 6
2.3 深度學習方法 6
2.3.1 卷積神經網路(CNN) 7
2.3.2 一維卷積神經網路(1D CNN) 8
2.3.3 殘差網路(ResNet) 8
2.3.4 U-net 8
2.3.5 多任務學習(Multi-task Learning, MTL) 9
第三章　研究資料與方法 10
3.1 資料集建立 10
3.1.1 原始資料來源 10
3.1.2 紀錄選擇 11
3.1.3 可用時間段提取 11
3.1.4 訊號濾波 11
3.1.5 血壓特徵值提取及標註 12
3.1.6 低血壓事件標註 12
3.1.7 排除無效的訊號片段 15
3.1.8 彙總訓練、驗證及測試子集 15
3.2 深度學習架構(Deep Learning Architecture) 17
3.2.1 輸入卷積塊(In-convolution Block) 18
3.2.2 殘差塊(Residual Block) 18
3.2.3 輸出卷積塊(Out-convolution Block) 19
3.2.4 編碼器(Encoder) 19
3.2.5 解碼器(Decoder) 19
3.2.6 血壓估計器／低血壓分類器(BP Estimator／HPO Classifier) 19
3.2.7 校準(Calibration) 20
第四章 實驗結果與討論 22
4.1 實驗流程設計 22
4.1.1 生理訊號建模 22
4.1.2 低血壓預測 23
4.1.3 校準 24
4.2 評估指標 24
4.2.1 血壓估計方法評估 24
4.2.2 低血壓事件預測方法評估 26
4.3 實驗結果 28
4.3.1 評估血壓估計 28
4.3.2 急性低血壓事件預測評估 32
4.4 綜合討論 33
4.4.1 SBP之高變異性 34
4.4.2 受試者生理訊號的過擬合 35
第五章　結論 37
第六章　參考文獻 38

[1] K. Maheshwari et al., "The relationship between ICU hypotension and in-hospital mortality and morbidity in septic patients," Intensive care medicine, vol. 44, no. 6, pp. 857-867, 2018.
[2] V. Salmasi et al., "Relationship between intraoperative hypotension, defined by either reduction from baseline or absolute thresholds, and acute kidney and myocardial injury after noncardiac surgery: a retrospective cohort analysis," Anesthesiology, vol. 126, no. 1, pp. 47-65, 2017.
[3] M. E. Chowdhury et al., "Wearable real-time heart attack detection and warning system to reduce road accidents," Sensors, vol. 19, no. 12, p. 2780, 2019.
[4] M. E. Chowdhury et al., "Real-time smart-digital stethoscope system for heart diseases monitoring," Sensors, vol. 19, no. 12, p. 2781, 2019.
[5] A. Chandrasekhar, C.-S. Kim, M. Naji, K. Natarajan, J.-O. Hahn, and R. Mukkamala, "Smartphone-based blood pressure monitoring via the oscillometric finger-pressing method," Science translational medicine, vol. 10, no. 431, p. eaap8674, 2018.
[6] M. Elgendi et al., "The use of photoplethysmography for assessing hypertension," NPJ digital medicine, vol. 2, no. 1, pp. 1-11, 2019.
[7] H. Lee et al., "Toward all-day wearable health monitoring: An ultralow-power, reflective organic pulse oximetry sensing patch," Science advances, vol. 4, no. 11, p. eaas9530, 2018.
[8] Y. Liang, Z. Chen, R. Ward, and M. Elgendi, "Photoplethysmography and deep learning: enhancing hypertension risk stratification," Biosensors, vol. 8, no. 4, p. 101, 2018.
[9] J. Wang, "Automated detection of atrial fibrillation and atrial flutter in ECG signals based on convolutional and improved Elman neural network," Knowledge-Based Systems, vol. 193, p. 105446, 2020.
[10] T. Pereira et al., "Photoplethysmography based atrial fibrillation detection: a review," NPJ digital medicine, vol. 3, no. 1, p. 3, 2020.
[11] L. Peter, N. Noury, and M. Cerny, "A review of methods for non-invasive and continuous blood pressure monitoring: Pulse transit time method is promising?," Irbm, vol. 35, no. 5, pp. 271-282, 2014.
[12] P. Hubner et al., "On detection of spontaneous pulse by photoplethysmography in cardiopulmonary resuscitation," The American Journal of Emergency Medicine, vol. 38, no. 3, pp. 526-533, 2020.
[13] G. Martínez, N. Howard, D. Abbott, K. Lim, R. Ward, and M. Elgendi, "Can photoplethysmography replace arterial blood pressure in the assessment of blood pressure?," Journal of clinical medicine, vol. 7, no. 10, p. 316, 2018.
[14] A. Esmaili, M. Kachuee, and M. Shabany, "Nonlinear cuffless blood pressure estimation of healthy subjects using pulse transit time and arrival time," IEEE Transactions on Instrumentation and Measurement, vol. 66, no. 12, pp. 3299-3308, 2017.
[15] G. Thambiraj, U. Gandhi, V. Devanand, and U. Mangalanathan, "Noninvasive cuffless blood pressure estimation using pulse transit time, Womersley number, and photoplethysmogram intensity ratio," Physiological Measurement, vol. 40, no. 7, p. 075001, 2019.
[16] T. Vandenberk et al., "Clinical validation of heart rate apps: mixed-methods evaluation study," JMIR mHealth and uHealth, vol. 5, no. 8, p. e7254, 2017.
[17] A. S. Zadi, R. Alex, R. Zhang, D. E. Watenpaugh, and K. Behbehani, "Arterial blood pressure feature estimation using photoplethysmography," Computers in biology and medicine, vol. 102, pp. 104-111, 2018.
[18] M. Simjanoska, M. Gjoreski, M. Gams, and A. Madevska Bogdanova, "Non-invasive blood pressure estimation from ECG using machine learning techniques," Sensors, vol. 18, no. 4, p. 1160, 2018.
[19] G. Slapničar, N. Mlakar, and M. Luštrek, "Blood pressure estimation from photoplethysmogram using a spectro-temporal deep neural network," Sensors, vol. 19, no. 15, p. 3420, 2019.
[20] M. H. Chowdhury et al., "Estimating blood pressure from the photoplethysmogram signal and demographic features using machine learning techniques," Sensors, vol. 20, no. 11, p. 3127, 2020.
[21] X. Fan, H. Wang, Y. Zhao, Y. Li, and K. L. Tsui, "An Adaptive Weight Learning-Based Multitask Deep Network for Continuous Blood Pressure Estimation Using Electrocardiogram Signals," Sensors, vol. 21, no. 5, p. 1595, 2021.
[22] C. Sideris, H. Kalantarian, E. Nemati, and M. Sarrafzadeh, "Building continuous arterial blood pressure prediction models using recurrent networks," in 2016 IEEE International Conference on Smart Computing (SMARTCOMP), 2016: IEEE, pp. 1-5.
[23] N. Ibtehaz and M. S. Rahman, "Ppg2abp: Translating photoplethysmogram (ppg) signals to arterial blood pressure (abp) waveforms using fully convolutional neural networks," arXiv preprint arXiv:2005.01669, 2020.
[24] S. Mahmud et al., "NABNet: a nested attention-guided BiConvLSTM network for a robust prediction of blood pressure components from reconstructed arterial blood pressure waveforms using PPG and ECG signals," Biomedical Signal Processing and Control, vol. 79, p. 104247, 2023.
[25] K. Qin, W. Huang, and T. Zhang, "Deep generative model with domain adversarial training for predicting arterial blood pressure waveform from photoplethysmogram signal," Biomedical Signal Processing and Control, vol. 70, p. 102972, 2021.
[26] M. A. Mehrabadi, S. A. H. Aqajari, A. H. A. Zargari, N. Dutt, and A. M. Rahmani, "Novel Blood Pressure Waveform Reconstruction from Photoplethysmography using Cycle Generative Adversarial Networks," arXiv preprint arXiv:2201.09976, 2022.
[27] N. Aguirre, E. Grall-Maës, L. J. Cymberknop, and R. L. Armentano, "Blood pressure morphology assessment from photoplethysmogram and demographic information using deep learning with attention mechanism," Sensors, vol. 21, no. 6, p. 2167, 2021.
[28] 紀韋辰, "結合TCN與GRU之深度網路由PPG及ECG預測非侵入式連續動脈壓波形," 碩士, 資訊管理學系研究所, 國立中山大學, 高雄市, 2022. [Online]. Available: https://hdl.handle.net/11296/36c5nh
[29] J. Lee and R. G. Mark, "An investigation of patterns in hemodynamic data indicative of impending hypotension in intensive care," Biomedical engineering online, vol. 9, no. 1, pp. 1-17, 2010.
[30] F. Hatib et al., "Machine-learning algorithm to predict hypotension based on high-fidelity arterial pressure waveform analysis," Anesthesiology, vol. 129, no. 4, pp. 663-674, 2018.
[31] B. Saugel, K. Kouz, P. Hoppe, K. Maheshwari, and T. W. Scheeren, "Predicting hypotension in perioperative and intensive care medicine," Best Practice & Research Clinical Anaesthesiology, vol. 33, no. 2, pp. 189-197, 2019.
[32] M. Cherifa, A. Blet, A. Chambaz, E. Gayat, M. Resche-Rigon, and R. Pirracchio, "Prediction of an acute hypotensive episode during an ICU hospitalization with a super learner machine-learning algorithm," Anesthesia & Analgesia, vol. 130, no. 5, pp. 1157-1166, 2020.
[33] M. C. Moghadam, E. M. K. Abad, N. Bagherzadeh, D. Ramsingh, G.-P. Li, and Z. N. Kain, "A machine-learning approach to predicting hypotensive events in ICU settings," Computers in biology and medicine, vol. 118, p. 103626, 2020.
[34] M. C. Moghadam, E. Masoumi, N. Bagherzadeh, D. Ramsingh, and Z. N. Kain, "Supervised machine-learning algorithms in real-time prediction of hypotensive events," in 2020 42nd Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC), 2020: IEEE, pp. 5468-5471.
[35] S. Lee, M. Lee, S.-H. Kim, and J. Woo, "Intraoperative Hypotension Prediction Model Based on Systematic Feature Engineering and Machine Learning," Sensors, vol. 22, no. 9, p. 3108, 2022.
[36] M. C. Moghadam, E. Masoumi, S. Kendale, and N. Bagherzadeh, "Predicting hypotension in the ICU using noninvasive physiological signals," (in eng), Comput Biol Med, vol. 129, p. 104120, Feb 2021, doi: 10.1016/j.compbiomed.2020.104120.
[37] I. Gratz et al., "The application of a neural network to predict hypotension and vasopressor requirements non-invasively in obstetric patients having spinal anesthesia for elective cesarean section (C/S)," BMC anesthesiology, vol. 20, no. 1, pp. 1-15, 2020.
[38] S. Lee et al., "Deep learning models for the prediction of intraoperative hypotension," British Journal of Anaesthesia, vol. 126, no. 4, pp. 808-817, 2021.
[39] Y. LeCun, L. Bottou, Y. Bengio, and P. Haffner, "Gradient-based learning applied to document recognition," Proceedings of the IEEE, vol. 86, no. 11, pp. 2278-2324, 1998.
[40] A. Krizhevsky, I. Sutskever, and G. E. Hinton, "Imagenet classification with deep convolutional neural networks," Advances in neural information processing systems, vol. 25, 2012.
[41] D. Kalita. "Basics of CNN in Deep Learning." https://www.analyticsvidhya.com/blog/2022/03/basics-of-cnn-in-deep-learning/ (accessed September 2nd, 2023).
[42] K. He, X. Zhang, S. Ren, and J. Sun, "Deep residual learning for image recognition," in Proceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 770-778.
[43] O. Ronneberger, P. Fischer, and T. Brox, "U-net: Convolutional networks for biomedical image segmentation," in Medical Image Computing and Computer-Assisted Intervention–MICCAI 2015: 18th International Conference, Munich, Germany, October 5-9, 2015, Proceedings, Part III 18, 2015: Springer, pp. 234-241.
[44] S. Ruder, "An overview of multi-task learning in deep neural networks," arXiv preprint arXiv:1706.05098, 2017.
[45] M. G. Moody B, Villarroel M, Clifford GD, Ikaro S. MIMIC-III Waveform Database Matched Subset (version 1.0). [Online]. Available: https://physionet.org/content/mimic3wdb-matched/1.0/
[46] A. E. Johnson et al., "MIMIC-III, a freely accessible critical care database," Scientific data, vol. 3, no. 1, pp. 1-9, 2016.
[47] A. L. Goldberger et al., "PhysioBank, PhysioToolkit, and PhysioNet: components of a new research resource for complex physiologic signals," circulation, vol. 101, no. 23, pp. e215-e220, 2000.
[48] A. Johnson, Pollard, T., & Mark, R. (2016). MIMIC-III Clinical Database (version 1.4). [Online]. Available: https://physionet.org/content/mimiciii/1.4/
[49] T. Athaya and S. Choi, "An estimation method of continuous non-invasive arterial blood pressure waveform using photoplethysmography: A U-Net architecture-based approach," Sensors, vol. 21, no. 5, p. 1867, 2021.
[50] Y. Liang, M. Elgendi, Z. Chen, and R. Ward, "An optimal filter for short photoplethysmogram signals," Scientific data, vol. 5, no. 1, pp. 1-12, 2018.
[51] M. Elgendi, "The use of photoplethysmography for assessing hypertension. npj Digit. Med. 2," ed, 2019.
[52] D.-K. Kim, Y.-T. Kim, H. Kim, and D.-J. Kim, "Deepcnap: A deep learning approach for continuous noninvasive arterial blood pressure monitoring using photoplethysmography," IEEE Journal of Biomedical and Health Informatics, vol. 26, no. 8, pp. 3697-3707, 2022.
[53] C. Yan et al., "Novel deep convolutional neural network for cuff-less blood pressure measurement using ECG and PPG signals," in 2019 41st Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), 2019: IEEE, pp. 1917-1920.
[54] W. Wang, P. Mohseni, K. L. Kilgore, and L. Najafizadeh, "PulseDB: A large, cleaned dataset based on MIMIC-III and VitalDB for benchmarking cuff-less blood pressure estimation methods," Frontiers in Digital Health, vol. 4, p. 1090854, 2023.
[55] F. Scholkmann, J. Boss, and M. Wolf, "An efficient algorithm for automatic peak detection in noisy periodic and quasi-periodic signals," Algorithms, vol. 5, no. 4, pp. 588-603, 2012.
[56] M. Poullis, "New formula to calculate mean aor tic pressure?," The Lancet, vol. 353, no. 9169, p. 2075, 1999.
[57] G. B. Moody and L.-w. H. Lehman, "Predicting acute hypotensive episodes: The 10th annual physionet/computers in cardiology challenge," in 2009 36th Annual Computers in Cardiology Conference (CinC), 2009: IEEE, pp. 541-544.
[58] S.-H. Kim, L. Li, C. Faloutsos, H.-J. Yang, and S.-W. Lee, "HeartCast: predicting acute hypotensive episodes in intensive care units," Statistical Methodology, vol. 33, pp. 1-13, 2016.
[59] S. Bhattacharya, V. Huddar, V. Rajan, and C. K. Reddy, "A dual boundary classifier for predicting acute hypotensive episodes in critical care," PloS one, vol. 13, no. 2, p. e0193259, 2018.
[60] J. Enevoldsen and S. T. Vistisen, "Performance of the Hypotension Prediction Index may be overestimated due to selection bias," Anesthesiology, vol. 137, no. 3, pp. 283-289, 2022.
[61] L. Zhang et al., "Developing personalized models of blood pressure estimation from wearable sensors data using minimally-trained domain adversarial neural networks," in Machine Learning for Healthcare Conference, 2020: PMLR, pp. 97-120.
[62] M. Kachuee, M. M. Kiani, H. Mohammadzade, and M. Shabany, "Cuffless blood pressure estimation algorithms for continuous health-care monitoring," IEEE Transactions on Biomedical Engineering, vol. 64, no. 4, pp. 859-869, 2016.
[63] S. S. Mousavi, M. Firouzmand, M. Charmi, M. Hemmati, M. Moghadam, and Y. Ghorbani, "Blood pressure estimation from appropriate and inappropriate PPG signals using A whole-based method," Biomedical Signal Processing and Control, vol. 47, pp. 196-206, 2019.
[64] S. Baek, J. Jang, and S. Yoon, "End-to-end blood pressure prediction via fully convolutional networks," Ieee Access, vol. 7, pp. 185458-185468, 2019.
[65] S. Hochreiter and J. Schmidhuber, "Long short-term memory," Neural computation, vol. 9, no. 8, pp. 1735-1780, 1997.
[66] R. Pascanu, T. Mikolov, and Y. Bengio, "On the difficulty of training recurrent neural networks," in International conference on machine learning, 2013: Pmlr, pp. 1310-1318.
[67] H. Liang, W. He, and Z. Xu, "A deep learning method for continuous noninvasive blood pressure monitoring using photoplethysmography," Physiological Measurement, vol. 44, no. 5, p. 055004, 2023.
[68] S. Mahmud et al., "A shallow U-Net architecture for reliably predicting blood pressure (BP) from photoplethysmogram (PPG) and electrocardiogram (ECG) signals," Sensors, vol. 22, no. 3, p. 919, 2022.
[69] R. Caruana, "Multitask learning," Machine learning, vol. 28, pp. 41-75, 1997.
[70] J. Li, S. Qiu, Y.-Y. Shen, C.-L. Liu, and H. He, "Multisource transfer learning for cross-subject EEG emotion recognition," IEEE transactions on cybernetics, vol. 50, no. 7, pp. 3281-3293, 2019.
[71] E. O’Brien et al., "The British Hypertension Society protocol for the evaluation of blood pressure measuring devices," J hypertens, vol. 11, no. Suppl 2, pp. S43-S62, 1993.
[72] ANSI/AAMI SP10:2002/(R)2008 and A1:2003/(R)2008 and A2:2006/(R)2008－Manual, Electronic,Or Automated Sphygmomanometers. [Online]. Available: https://webstore.ansi.org/Standards/AAMI/ansiaamisp1020022008a12003a2