本論文提出了一個血壓預測演算法，可以透過單一個脈音訊號回歸成連續血壓波形，其演算法考慮過去一段時間資訊生成連續特徵，並以非線性機器學習模型回歸成連續血壓波形以及逐搏的血壓數值（收縮壓、舒張壓、平均動脈壓以及脈壓）。
在血壓預測的任務上為了增加其難度，訊號收集的過程中請受測者執行閉氣用力的動作（Valsalva maneuver），促使血壓在短時間產生劇烈變化。收錄的訊號來源為兩個部分，橈動脈上的脈音訊號（pulse audiogram,PAG）以及透過手指收集的高精度非侵入式連續血壓波形（non-invasive blood pressure ，NIBP）。
初步分析脈音訊號與連續血壓波形的相關性，比較連續十秒區間的變化，脈音訊號峰值與收縮壓相關係數高達0.86，脈音訊號震幅與脈壓的相關係數為0.76。透過初步的分析可以證實脈音訊號與連續血壓波形具有高度相關性，以此為基礎發展連續血壓預測演算法。
連續血壓預測演算法首先將訊號做預處理，使用濾波、正規化方式去除高頻雜訊以及減少個體間的差異，並產生時序特徵，保留過去一段時間內連續波形變化，在機器學習模型上使用了 Linear least squares、Ensemble of bagging decision trees、Support vector machine 以及 Artificial neural network 訓練回歸模型，驗證部分則是選用了更嚴格的方式（Leave one measure out）來評估模型性能。 最佳結果顯示，Support vector machine 模型對於連續血壓預測的平均絕對誤差為3.76 ，標準差為 3.79。 對於逐搏血壓預測，最好的結果顯示在Artificial neural network模型中平均動脈壓的表現其平均絕對誤差和標準差分別為 2.57和 3.44。
相較於原始脈音訊號，透過非線性模型轉換後的訊號和連續血壓波形相關性分析中，收縮壓、舒張壓、平均動脈壓以及脈壓逐博變化趨勢呈現一致性，逐搏血壓相關性有顯著提升。
總結上述的表現，可以證明使用單一脈音訊號可以有效回歸連續血壓波形，透過時序特徵選用上，也驗證了能夠反應高度變異的血壓特徵，在高變異血壓估計應用上優於現有的文獻結果。

This thesis proposes a blood pressure prediction algorithm that can directly regress a single pulse audiogram signal into a continuous blood pressure waveform, and this work takes into account information from the past period of time to generate continuous features and applies non-linear machine learning approaches to predict continuous blood pressure waveforms and beat-to-beat blood pressure values (systolic blood pressure, diastolic blood pressure, mean arterial pressure, and pulse pressure).
In order to increase the difficulty of the blood pressure prediction task, the subjects were asked to perform a forceful attempt of exhalation against a closed airway (Valsalva maneuver) during the data collection which caused blood pressure to change drastically in a short time. The recorded signal consisted of two parts, one was the pulse audiogram signal collected on the radial artery and the other was a high-precision non-invasive continuous blood pressure waveform (NIBP) collected through the finger.
Preliminary observation and analysis of the correlation between the pulse audiogram signal and the continuous blood pressure waveform, comparing the changes in the continuous 10-second interval, the correlation coefficient between the peak value of the pulse audiogram signal and the systolic blood pressure was 0.86, and the correlation coefficient between the amplitude of the pulse audiogram signal and the pulse pressure was 0.76. Through the preliminary analysis, it can be verified that the pulse audiogram signal was highly correlated with the continuous blood pressure waveform. Based on this, we developed the continuous blood pressure prediction algorithm.
In the continuous blood pressure prediction algorithm, the first step of preprocessing was using filtering and normalization methods to remove high-frequency noise and reduce inter-individual differences. After that, generated time series features to retain continuous waveform changes information in the past period of time.
The machine learning approaches, the linear least squares, ensemble of bagging decision trees, support vector machine, and artificial neural network were used to train the regression model, and the verification part uses a stricter method (leave one segment out) to evaluate the model performance. The best results showed that the mean absolute error (MAE) was 3.76 and the standard deviation was 3.79 for continuous blood pressure prediction by the support vector machine model. For beat-to-beat BP prediction, the best results showed the mean absolute error and standard deviation of 2.57 and 3.44 for mean arterial pressure in the artificial neural network model, respectively.
Compared with the original pulse audiogram signal, in the correlation analysis between the signal converted by the nonlinear model and the continuous blood pressure waveform, the trend of systolic blood pressure, diastolic blood pressure, mean arterial pressure and pulse pressure showed consistency, and beat-to-beat blood pressure correlation was significantly improved.
Summarizing the above performance, it can be demonstrated that the continuous blood pressure waveform can be effectively regressed by using a single pulse audiogram signal. Through the generation of continuous time series features, it is also verified that it can reflect highly variable blood pressure features. This work outperforms the existing literature results on the application of high blood pressure estimation.

[1] S. Mendis, S. Davis, and B. Norrving, “Organizational update: the world health organization global status report on noncommunicable diseases 2014; one more landmark step in the combat against stroke and vascular disease,” Stroke, vol. 46, no. 5, pp. e121-e122, 2015.
[2] T. G. Pickering, J. E Hall, L. J Appel, B. E Falkner, J. Graves, M. N. Hill, D. W. Jones, T.Kurtz, S. G. Sheps, and E. J. Roccella, “Recommendations for blood pressure measurement in humans and experimental animals: part 1: blood pressure measurement in humans: a statement for professionals from the Subcommittee of Professional and Public Education of the American Heart Association Council on High Blood Pressure Research,” Hypertension, vol. 45, no. 1, pp. 142-161, 2005.
[3] P. J. Collignon and A. S. o. I. C. A. Sepsis, “Intravascular catheter associated sepsis: a common problem,” Medical journal of Australia, vol. 161, no. 6, pp. 374-378, 1994.
[4] L. Geddes and G. LA, “Pulse arrival time as a method of obtaining systolic and diastolic blood pressure indirectly,” 1981
[5] W. Chen, T. Kobayashi, S. Ichikawa, Y. Takeuchi, and T. Togawa, “Continuous estimation of systolic blood pressure using the pulse arrival time and intermittent calibration,” Medical and Biological Engineering and Computing, vol. 38, no. 5, pp. 569-574, 2000.
[6] G. Zhang, M. Gao, D. Xu, N. B. Olivier, and R. Mukkamala, “Pulse arrival time is not an adequate surrogate for pulse transit time as a marker of blood pressure,” Journal of applied physiology, vol. 111, no. 6, pp. 1681-1686, 2011.
[7] M. Forouzanfar, S. Ahmad, I. Batkin, H. R. Dajani, V. Z. Groza, and M. Bolic, “Model-based mean arterial pressure estimation using simultaneous electrocardiogram and oscillometric blood pressure measurements,” IEEE Transactions on Instrumentation and Measurement, vol. 64, no. 9, pp. 2443-2452, 2015.
[8] L. Geddes, M. Voelz, C. Babbs, J. Bourland, and W. Tacker, “Pulse transit time as an indicator of arterial blood pressure,” psychophysiology, vol. 18, no. 1, pp. 71-74, 1981.
[9] M. Nitzan, B. Khanokh, and Y. Slovik, “The difference in pulse transit time to the toe and finger measured by photoplethysmography,” Physiological measurement, vol. 23, no. 1, p. 85, 2001.
[10] C. Poon and Y. Zhang, “Cuff-less and noninvasive measurements of arterial blood pressure by pulse transit time,” in 2005 IEEE engineering in medicine and biology 27th annual conference, 2006: IEEE, pp. 5877-5880.
[11] T. Ma and Y.-T. Zhang, “A correlation study on the variabilities in pulse transit time, blood pressure, and heart rate recorded simultaneously from healthy subjects,” in 2005 IEEE Engineering in Medicine and Biology 27th Annual Conference, 2006: IEEE, pp. 996-999.
[12] N. Kumar, A. Agrawal, and S. Deb, “Cuffless BP measurement using a correlation study of pulse transient time and heart rate,” in 2014 International Conference on Advances in Computing, Communications and Informatics (ICACCI), 2014: IEEE, pp. 1538-1541.
[13] Y. Chen, C. Wen, G. Tao, M. Bi, and G. Li, “Continuous and noninvasive blood pressure measurement: a novel modeling methodology of the relationship between blood pressure and pulse wave velocity,” Annals of biomedical engineering, vol. 37, no. 11, pp. 2222-2233, 2009.
[14] Y. Chen, C. Wen, G. Tao, and M. Bi, “Continuous and noninvasive measurement of systolic and diastolic blood pressure by one mathematical model with the same model parameters and two separate pulse wave velocities,” Annals of biomedical engineering, vol. 40, no. 4, pp. 871-882, 2012.
[15] M. Elgendi, R. Fletcher, Y. Liang, N. Howard, N. H. Lovell, D. Abbott, K. Lim and R. Ward., “The use of photoplethysmography for assessing hypertension,” NPJ digital medicine, vol. 2, no. 1, pp. 1-11, 2019.
[16] 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.
[17] X. Teng and Y. Zhang, “Continuous and noninvasive estimation of arterial blood pressure using a photoplethysmographic approach,” in Proceedings of the 25th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (IEEE Cat. No. 03CH37439), 2003, vol. 4: IEEE, pp. 3153-3156.
[18] Y. Kurylyak, F. Lamonaca, and D. Grimaldi, “A Neural Network-based method for continuous blood pressure estimation from a PPG signal,” in 2013 IEEE International instrumentation and measurement technology conference (I2MTC), 2013: IEEE, pp. 280-283.
[19] A. D. Choudhury, R. Banerjee, A. Sinha, and S. Kundu, “Estimating blood pressure using Windkessel model on photoplethysmogram,” in 2014 36th annual international conference of the IEEE engineering in medicine and biology society, 2014: IEEE, pp. 4567-4570.
[20] A. Gaurav, M. Maheedhar, V. N. Tiwari, and R. Narayanan, “Cuff-less PPG based continuous blood pressure monitoring—A smartphone based approach,” in 2016 38th annual international conference of the IEEE engineering in medicine and biology society (EMBC), 2016: IEEE, pp. 607-610.
[21] M. Liu, L.-M. Po, and H. Fu, “Cuffless blood pressure estimation based on photoplethysmography signal and its second derivative,” International Journal of Computer Theory and Engineering, vol. 9, no. 3, p. 202, 2017.
[22] D. Fujita, A. Suzuki, and K. Ryu, “PPG-based systolic blood pressure estimation method using PLS and level-crossing feature,” Applied Sciences, vol. 9, no. 2, p. 304, 2019.
[23] X. Xing and M. Sun, “Optical blood pressure estimation with photoplethysmography and FFT-based neural networks,” Biomedical optics express, vol. 7, no. 8, pp. 3007-3020, 2016.
[24] 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.
[25] M. Panwar, A. Gautam, D. Biswas, and A. Acharyya, “PP-Net: A deep learning framework for PPG-based blood pressure and heart rate estimation,” IEEE Sensors Journal, vol. 20, no. 17, pp. 10000-10011, 2020.
[26] K.-H. Huang, F. Tan, T.-D. Wang, and Y.-J. Yang, “A highly sensitive pressure-sensing array for blood pressure estimation assisted by machine-learning techniques,” Sensors, vol. 19, no. 4, p. 848, 2019.
[27] J. Kim, E. F. Chou, J. Le, S. Wong, M. Chu, and M. Khine, “Soft wearable pressure sensors for beat‐to‐beat blood pressure monitoring,” Advanced healthcare materials, vol. 8, no. 13, p. 1900109, 2019.
[28] S.-Y. Yoo, J.-E. Ahn, G. Cserey, H.-Y. Lee, and J.-M. Seo, “Reliability and validity of non-invasive blood pressure measurement system using three-axis tactile force sensor,” Sensors, vol. 19, no. 7, p. 1744, 2019.
[29] T.-W. Wang and S.-F. Lin, “Wearable piezoelectric-based system for continuous beat-to-beat blood pressure measurement,” Sensors, vol. 20, no. 3, p. 851, 2020.
[30] B. D. Alvis, M. Polcz, M. Miles, D. Wright, M. Shwetar, P. Leisy, R. Forbes, R. Fissell, J. Whitfield, S. Eagle, C. Brophy and K. Hocking, “Non-invasive venous waveform analysis (NIVA) for volume assessment in patients undergoing hemodialysis: an observational study,” BMC nephrology, vol. 21, no. 1, pp. 1-8, 2020.
[31] P. Rwei, C. Qian, A. Abiri, Y. Zhou, E. F. Chou, W. C. Tang, M. Khine, “Soft Iontronic Capacitive Sensor for Beat‐to‐Beat Blood Pressure Measurements,” Advanced Materials Interfaces, p. 2200294, 2022.
[32] E. O'Brien, R. Asmar, L. Beilin, Y. Imai, G. Mancia, T. Mengden, M. Myers, P. Padfield, P. Palatini, G. Parati, T. Pickering, J. Redon, J. Staessen, G. Stergiou and P. Verdecchia, “Practice guidelines of the European Society of Hypertension for clinic, ambulatory and self blood pressure measurement,” Journal of hypertension, vol. 23, no. 4, pp. 697-701, 2005.
[33] B. Rosner, N. Cook, R. Portman, S. Daniels, and B. Falkner, “Blood pressure differences by ethnic group among United States children and adolescents,” Hypertension, vol. 54, no. 3, pp. 502-508, 2009.
[34] G. S Stergiou, B. Alpert, S. Mieke, R. Asmar, N. Atkins, S. Eckert, G. Frick, B. Friedman, T. Graßl, T. Ichikawa, J. P Ioannidis, P. Lacy, R. McManus, A. Murray, M. Myers, P. Palatini, G. Parati, D. Quinn, J. Sarkis, A. Shennan, T. Usuda, J. Wang, C. O Wu and E. O'Brien, “A universal standard for the validation of blood pressure measuring devices: Association for the Advancement of Medical Instrumentation/European Society of Hypertension/International Organization for Standardization (AAMI/ESH/ISO) Collaboration Statement,” Hypertension, vol. 71, no. 3, pp. 368-374, 2018.
[35] 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.
[36] Z.-D. Liu, J.-K. Liu, B. Wen, Q.-Y. He, Y. Li, and F. Miao, “Cuffless blood pressure estimation using pressure pulse wave signals,” Sensors, vol. 18, no. 12, p. 4227, 2018.
[37] L. Pstras, K. Thomaseth, J. Waniewski, I. Balzani, and F. Bellavere, “The Valsalva manoeuvre: physiology and clinical examples,” Acta physiologica, vol. 217, no. 2, pp. 103-119, 2016.
[38] J.-Y. Chen, C.-C. K. Lin, C.-W. Lin, F.-M. Yu, K.-J. Li, and L.-M. Tsai, “Development of Radial Artery Pulse Audiogram Sensing System for Fast Detection of Atrial Fibrillation and Pulse Amplitude Variation,” IEEE Access, vol. 8, pp. 178770-178781, 2020.
[39] ADInstruments NZ Limited., “Human NIBP Nano- Owners Guide,” 2022. [Online]. Available:https://m-cdn.adinstruments.com/owners-guides/Human%20NIBP%20Nano%20Owners%20Guide.pdf
[40] Alex Yartsev, “Physiology of the Valsalva manoeuvre,” 2020. [Online]. Available: https://derangedphysiology.com/main/cicm-primary-exam/required-reading/cardiovascular-system/Chapter%20505/physiology-valsalva-manoeuvre
[41] W.-Q. Lin 1, H.-H. Wu, C.-S. Su, J.-T. Yang, J.-R. Xiao, Y.-P. Cai, X.-Z. Wu and G.-Z. Chen, “Comparison of Continuous Noninvasive Blood Pressure Monitoring by TL-300 With Standard Invasive Blood Pressure Measurement in Patients Undergoing Elective Neurosurgery,” Journal of Neurosurgical Anesthesiology, vol. 29, no. 1, pp. 1-7, 2017
[42] J. Miles, “R-Squared, Adjusted R-Squared,” Encyclopedia of Statistics in Behavioral Science, 2005.
[43] J. Benesty, J. Chen, Y. Huang, and I. Cohen, “Pearson correlation coefficient,” in Noise reduction in speech processing: Springer, 2009, pp. 1-4.
[44] C. J. Willmott and K. Matsuura, “Advantages of the mean absolute error (MAE) over the root mean square error (RMSE) in assessing average model performance,” Climate research, vol. 30, no. 1, pp. 79-82, 2005.
[45] X. Wan, W. Wang, J. Liu, and T. Tong, “Estimating the sample mean and standard deviation from the sample size, median, range and/or interquartile range,” BMC medical research methodology, vol. 14, no. 1, pp. 1-13, 2014.
[46] N. Selvaraj and H. Reddivari, "Feasibility of Noninvasive Blood Pressure Measurement using a Chest-worn Patch Sensor," (in eng), Annu Int Conf IEEE Eng Med Biol Soc, vol. 2018, pp. 1-4, Jul 2018
[47] Y. Ota, A. Kokubo, S. Yamashita and K. Kario, "Development of Small and Lightweight Beat-By-Beat Blood Pressure Monitoring Device Based on Tonometry," 2021 43rd Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC), Mexico, 2021, pp. 5455-5458
[48] J. R. Martina, Berend E. Westerhof, Jeroen van Goudoever, Edouard M. F. H. de Beaumont, Jasper Truijen, Yu-Sok Kim, Rogier V. Immink, Dorothea A. Jöbsis, Markus W. Hollmann, Jaap R. Lahpor, Bas A. J. M. de Mol and Johannes J. van Lieshout, "Noninvasive continuous arterial blood pressure monitoring with Nexfin®," The Journal of the American Society of Anesthesiologists, vol. 116, no. 5, pp. 1092-1103, 2012.
[49] B. P. M. Imholz, W. Wieling, G. A. van Montfrans, and K. H. Wesseling, "Fifteen years experience with finger arterial pressure monitoring: assessment of the technology," Cardiovascular Research, vol. 38, no. 3, pp. 605-616, 1998
[50] J. B. Ramsey, "Tests for specification errors in classical linear least‐squares regression analysis," Journal of the Royal Statistical Society: Series B (Methodological), vol. 31, no. 2, pp. 350-371, 1969.
[51] T. G. Dietterich, "An experimental comparison of three methods for constructing ensembles of decision trees: Bagging, boosting, and randomization," Machine learning, vol. 40, pp. 139-157, 2000.
[52] W. S. Noble, "What is a support vector machine?, " Nature biotechnology, vol. 24, no. 12, pp. 1565-1567, 2006.
[53] A. K. Jain, J. Mao, and K. M. Mohiuddin, "Artificial neural networks: A tutorial," Computer, vol. 29, no. 3, pp. 31-44, 1996.