EXTRACTION OF MULTIPLE PHYSIOLOGICAL PARAMETERS FROM A SINGLE-CHANNEL PHOTOPLETHYSMOGRAPHIC WAVEFORM USING MACHINE LEARNING

Taiwo Samuel Ajani; Akinwumi Abimbola Amusan; Agbotiname Lucky  Imoize

doi:10.4314/njt.2026.5675

Authors

T. S. Ajani Department of Electrical & Electronics Engineering, University of Lagos, NIGERIA,
A. A. Amusan Department of Electrical & Electronics Engineering, University of Lagos, NIGERIA,
A. L. Imoize Department of Electrical & Electronics Engineering, University of Lagos, NIGERIA,

DOI:

https://doi.org/10.4314/njt.2026.5675

Keywords:

cardiovascular health, k-fold cross-validation, Machine learning, Photoplethysmogram, Physiological parameters

Abstract

In today’s world, cardiovascular health challenges constitute the foremost causes of death due to a lack of timely medical attention. This concerns the present non-continuous, uncomfortable, and expensive clinical techniques for assessing cardiovascular health. Meanwhile, photoplethysmographic (PPG) signals contain valuable information concerning the human cardiovascular system. They may be exploited to extract physiological parameters, which can provide simple, convenient, and continuous monitoring of cardiovascular health. Although, some works have attempted to determine crucial physiological parameters from the photoplethysmogram, they do not extract these parameters simultaneously with high estimation accuracy. Consequently, this work performs a pulse wave analysis of the photoplethysmogram using four machine-learning (ML) techniques (XGBoost, feedforward ANN, RNN, and Bi-LSTM) to simultaneously estimate key physiological parameters (heart rate (HR) and blood pressure (BP)) to monitor the cardiovascular system. The ML models are developed and evaluated using a publicly available dataset, which contains a handcrafted 46-feature dataset extracted from a 3D input consisting of the photoplethysmogram (PPG), velocity photoplethysmogram (VPG), and acceleration photoplethysmogram (APG) signals. The performance evaluation of the machine learning models was implemented using k-fold cross-validation and typical error indices, such as mean absolute error (MAE), mean squared error (MSE), and root mean square error (RMSE) visualized through cross-validation plots, test set metric plots, and parity plots. This study revealed that the ML models achieved a mean absolute error (MAE) of less than 0.5 for heart rate, and less than 1 for diastolic, systolic, and arterial blood pressures, respectively, outperforming existing research.

References

[1] Elgendi, M. PPG Signal Analysis: An Introduction Using MATLAB, Boca Raton, FL: CRC Press, 2021.

[2] Kumar, S., Yadav, S., and Kumar, A. “Blood pressure measurement techniques, standards, technologies, and the latest futuristic wearable cuff-less know-how,” Sensors and Diagnostics, 3 (2), pp. 181–202, 2024. doi: 10.1039/d3sd00201b.

[3] Ding, X.-R., Zhang, Y.-T., Liu, J., and Dai, W.-X “Continuous cuffless blood pressure estimation using pulse transit time and photoplethysmogram intensity ratio,” IEEE Transactions on Biomedical Engineering, 63 (5), pp. 964–972, 2016. doi: 10.1109/TBME.2015.2480679.

[4] Fuadah, Y. N., and Lim, K. M. “Advances in cardiovascular signal analysis with future directions: a review of ML and DL models for cardiovascular disease classification based on ECG, PCG, and PPG signals,” Biomedical Engineering Letters, 15 (4), pp. 619-660, 2025. doi: 10.1007/s13534-025-00473-9.

[5] Shuzan, M. N. I., Chowdhury, M. H., Chowdhury, M. E., Murugappan, M., Bhuiyan, E. H., Ayari, M. A., and Khandakar, A. “Machine learning – based respiration rate and blood oxygen saturation estimation using photoplethysmogram signals,” Bioengineering, 10 (2), article 167, pp. 1–15, 2023. doi: 10.3390/bioengineering10020167.

[6] Kang, T. W., Lee, J., Kwon, Y., Kim, J., Lee, H., Kim, S., Sim, J. Y., and Yeo, W.-H. “Recent progress in the development of flexible wearable electrodes for electrocardiogram monitoring during exercise,” Advanced NanoBiomed Research, 4 (8), 2300169, pp. 1–25, 2024. doi: 10.1002/anbr.202300169.

[7] Allen, J. “Photoplethysmography and its application in clinical physiological measurement,” Physiological Measurement, 28 (3), pp. 1–39, 2007. doi: 10.1088/0967-3334/28/3/R01.

[8] Jayroop, R., Zahra, S., Raafat, A., and Assim, S. “Atrial fibrillation classification with smart wearables using short-term heart rate variability and deep convolutional neural networks,” Sensors, 21(21), 7233, pp. 1–24, 2021. doi: 10.3390/s21217233.

[9] Kanoga, S., Hoshino, T., Kamei, S., Kobayashi, T., Ohmori, T., Uchiyama, M., and Tada, M. “Comparison of seven shallow and deep regressors in continuous blood pressure and heart rate estimation using single-channel photoplethysmograms under three evaluation cases,” Biomedical Signal Processing and Control, 85, article 105029, pp. 1–18, 2023. doi: 10.1016/j.bspc.2023.105029.

[10] Monalisa, S. R., Rajarshi, G., and Kaushik, D. S. “BePCon: A photoplethysmography-based quality-aware continuous beat-to-beat blood pressure measurement technique using deep learning,” IEEE Transactions on Instrumentation and Measurement, 71, Art no. 2519709, pp. 1-9, 2022. doi: 10.1109/TIM.2022.3212750.

[11] Yoon, G., Lee, J. Y., Jeon, K. J., Park, K. K., and Kim, H. S. “Development of a compact home health monitor for telemedicine,” Telemedicine Journal and e-Health, 11 (6), pp. 660–667, 2005. doi: 10.1089/tmj.2005.11.660.

[12] Yen, C., and Liao, C. “Blood pressure and heart rate measurements using photoplethysmography with modified LRCN,” Computers, Materials & Continua, 71 (1), pp. 1973–1986, 2022. doi: 10.32604/cmc.2022.022679.

[13] Slapničar, G., Mlakar, N., and Lustrek, M. “Blood pressure estimation from photoplethysmogram using a spectro-temporal deep neural network,” Sensors, 19 (15), 3420, pp. 1–17, 2019. doi: 10.3390/s19153420.

[14] Kim, M., Sung, M. D., Jung, J., Cho, S. P., Park, J., Soh, S., Joo, H. C., and Chung, K. S. “Wearable ECG-PPG Deep Learning Model for Cardiac Index-Based Noninvasive Cardiac Output Estimation in Cardiac Surgery Patients,” Sensors (Basel, Switzerland), 26 (2), 735, pp. 1–20, 2026. doi: 10.3390/s26020735.

[15] Xing, J., Fang, X., Bai, J., Cui, L., Zhang, F., and Xu, Y. “Study on Multimodal Sensor Fusion for Heart Rate Estimation Using BCG and PPG Signals,” Sensors, 26 (2), 548, pp. 1–17, 2026. doi: 10.3390/s26020548.

[16] Sunwoo, J., Einalou, Z., Dadgostar, M., Renna, M., Wu, K. C., Otic, N., Martin, A., Starkweather, Z., Oh, Y., Qu, J. Z., and Franceschini, M. A. “Continuous, noninvasive blood pressure estimation using forehead NIRS-PPG and LSTM-CNN deep learning,” Biomedical Signal Processing and Control, 113 (B), 108943, pp. 1–X, 2026. doi: 10.1016/j.bspc.2025.108943.

[17] Zeynali, M., Alipour, K., Tarvirdizadeh, B., and Ghamari, M. “Non-invasive blood glucose monitoring using PPG signals with various deep learning models and implementation using TinyML,” Scientific Reports, 15 (1), 581, pp. 1–23, 2025. doi: 10.1038/s41598-024-84265-8.

[18] Panwar, M., Gautam, A., Biswas, D., and Acharyya, A. “PP-Net: A deep learning framework for PPG-based blood pressure and heart rate estimation,” IEEE Sensors Journal, 20 (17), pp. 10000–10011, 2020. doi: 10.1109/JSEN.2020.2990864.

[19] Kanoga, S., Hoshino, T., Kamei, S., Kobayashi, T., Ohmori, T., Uchiyama, M., and Tada, M., “Preprocessed MIMIC III waveform database matched subset,” 2023, Available: https://drive.google.com/drive/folders/182xF0Y8NPiDGownNcxUfE4D-BPwv-bvO

[20] Li, Z., and He, W. “A continuous blood pressure estimation method using photoplethysmography by GRNN-based model,” Sensors, 21 (21), 7207, pp. 1–15, 2021. doi: 10.3390/s21217207.

[21] Leitner, J., Chiang, P., and Dey, S. “Personalized blood pressure estimation using photoplethysmography: A transfer learning approach,” IEEE Journal of Biomedical and Health Informatics, 25 (9), pp. 2194–2204, 2021, doi: 10.1109/JBHI.2021.3085526.