Selection of amplitude-temporal and spectral parameters of PPG signals for diagnostics based on machine learning methods
DOI:
https://doi.org/10.64915/RADAP.2026.104.63-73Keywords:
non-invasive diagnostics, physiological signal analysis, biosignal processing, feature extraction, feature selection, frequency-domain analysis, phase spectrum, photoplethysmography (PPG), pulse wave analysis, machine learningAbstract
The problem of increasing the efficiency of non-invasive express diagnostics of the functional state of the organism based on the analysis of photoplethysmographic (PPG) signals using machine learning methods is considered.
Based on the analysis of the morphological features of the pulse wave, which contain important information about hemodynamic processes, the feasibility of switching from the traditional analysis of the amplitude-time characteristics of the signal to a complex approach is justified, which includes additional extraction of spectral-phase parameters within individual cardiocycles. A method of signal preprocessing is proposed, which includes band-pass filtering, normalization and segmentation, as well as an approach to ``soft'' filtering, which allows preserving diagnostically significant features of the signal shape. The system of amplitude-time parameters (duration of anacrota, intervals between characteristic points, area under the curve, etc.) and spectral-phase characteristics obtained using Fourier series expansion is considered in detail.
Particular attention is paid to the procedure for selecting informative features for machine learning models. A multi-step approach is proposed, which includes eliminating the dependence on heart rate, multicollinearity analysis, assessing the importance of features using the permutation importance method, and interpreting the results using the PCA and SHAP methods. It was found that the greatest contribution to the accuracy of the models is provided by spectral-phase parameters, in particular group delay and phase shifts of harmonics, which are directly related to the physiological properties of the vascular system.
As a result, a set of features was formed that combines the temporal and spectral characteristics of the signal and provides improved classification and prediction quality. The results obtained confirm the prospects of using complex analysis of PPG signals in combination with machine learning methods for medical diagnostics and human condition monitoring.
References
1. Kyriacou P. and Allen J. (2021) Photoplethysmography: Technology, Signal Analysis and Applications, 1st ed. Elsevier.
2. Khalid S. G., Zhang J., Chen F. and Zheng D. (2018) Blood pressure estimation using photoplethysmography only: Comparison between different machine learning approaches, Journal of Healthcare Engineering, vol. 2018, 1548647, doi: 10.1155/2018/1548647
3. Visvanathan A., Sinha A. and Pal A. (2013) Estimation of blood pressure levels from reflective photoplethysmograph using smart phones, 13th IEEE International Conference on BioInformatics and BioEngineering, pp. 1-5. doi: 10.1109/BIBE.2013.6701568
4. Ali N. F. and Atef M. (2022) Lstm multi-stage transfer learning for blood pressure estimation using photoplethysmography, Electronics, vol. 11, 22, doi: 10.3390/electronics11223749
5. El-Hajj C. and Kyriacou P. (2020) A review of machine learning techniques in photoplethysmography for the non-invasive cuff-less measurement of blood pressure, Biomedical Signal Processing and Control, vol. 58, p. 101 870. doi: 10.1016/j.bspc.2020.101870
6. Priyadarshini R. G., Kalimuthu M., Nikesh S. and Bhuvaneshwari M. (2021) Review of ppg signal using machine learning algorithms for blood pressure and glucoseestimation, IOP Conference Series: Materials Science and Engineering, vol. 1084, pp. 012-031. doi: 10.1088/1757-899X/1084/1/012031
7. Arg¨uello-Prada E. J., Ojeda A. V. V., and Ojeda M. Y. V. (2025) Non-invasive prediction of cholesterol levels from photoplethysmogram (ppg)-based features using machinelearning techniques: A proof-of-concept study, Cogent Engineering, vol. 12, no. 1, p. 2 467 153. doi: 10.1080/23311916.2025.2467153
8. Elgendi M. (2020) textit{PPG Signal Analysis}. 1st ed. CRC Press. [Online]. Available: https://www.perlego.com/book/1974664/ppg-signal-analysis-an-introduction-using-matlab-pdf
9. Seenivasan D. and Sakthivel K. (2025) Patch-based ecg segmentation for arrhythmia classification using hybrid deep learning, Information Technology and Control, vol. 54, no. 4, pp. 1189-1205. doi: 10.5755/j01.itc.54.4.41549
10. Seitsonen E. et al. (2005) EEG spectral entropy, heart rate, photoplethysmography and motor responses to skin incision during sevoflurane anaesthesia, Acta anaesthesiologica Scandinavica, vol. 49, pp. 284-292. doi: 10.1111/j.1399-6576.2005.00654.x
11. Alty S. R., Angarita-Jaimes N.,Millasseau S. C. and Chowienczyk P. J. (2007) Predicting arterial stiffness from the digital volume pulse waveform, IEEE Transactions on Biomedical Engineering, vol. 54, pp. 2268-2275. doi: 10.1109/TBME.2007.897805
12. Awad A. A. et al. (2007) The relationship between the photoplethysmographic waveform and systemic vascular resistance, Journal of Clinical Monitoring and Computing, vol. 21, pp. 365-372. doi: 10.1007/s10877-007-9097-5
13. Sharpan O. B. (2003) Diagnostic sensitivity of the phase spectrum of the pulse signal, textit{Elektronika i svyaz'}, No. 18, pp. 91-95.
14. Sharpan O. B., Husieva O. V. and Mahliovana N. I. (2007) The program-measuring photoplethismographic system of pulse spectrometry. Bulletin of Vinnytsia Polytechnic Institute, no. 5, pp. 34-40, 2007.
15. Rinkeviˇcius M. et al. (2019) Photoplethysmogram signal morphology-based stress assessment, 2019 Computing in Cardiology (CinC), pp. 1-4. doi: 10.22489/CinC.2019.126
16. Elgendi M. (2016) Optimal signal quality index for photoplethysmogram signals, Bioengineering, vol. 3, no. 4. doi: 10.3390/bioengineering3040021
17. Moscato S., Lo Giudice S., Massaro G., and Chiari L. (2022) Wrist photoplethysmography signal quality assessment for reliable heart rate estimate and morphological analysis, Sensors, vol. 22, 15, doi: 10.3390/s22155831
18. Yousef Q., Reaz M.B.I. and Ali M. (2012) The analysis of ppg morphology: Investigating the effects of aging on arterial compliance, Measurement Science Review, vol. 12, pp. 266-271. doi: 10.2478/v10048-012-0036-3
19. Hickey M. and Phillips J. (2015) Investigation of peripheral photoplethysmographic morphology changes induced during a hand-elevation study, Journal of Clinical Monitoring and Computing, vol. 30. doi: 10.1007/s10877-015-9761-0
20. Reguig F. B. (2016) Photoplethysmogram signal analysis for detecting vital physiological parameters: An evaluating study, 2016 International Symposium on Signal, Image, Video and Communications (ISIVC), pp. 167-173. doi: 10.1109/ISIVC.2016.7893981
21. Gonzalez-Landaeta R., Ramirez B. and Mejia J. (2022) Estimation of systolic blood pressure by random forest using heart sounds and a ballistocardiogram, Scientific Reports, vol. 12, no. 1, p. 17196, doi: 10.1038/s41598-022-22205-0
22. Benchekroun M. et al. (2022) Comparison of stress detection through ecg and ppg signals using a random forest-based algorithm, 2022 44th Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC), vol. 2022, pp. 3150-3153. doi: 10.1109/EMBC48229.2022.9870984
23. Lee S.-W., Woo D.-K., Son Y.-K. and Mah P. S. (2019) Wearable bio-signal(ppg)-basedpersonal authentication method using random forest and period setting consideringthe feature of ppg signals, J. Comput., vol. 14, pp. 283-294. doi: 10.17706/jcp.14.4.283-294.
Downloads
Published
Issue
Section
License
Copyright (c) 2026 В. С. Мосійчук, О. Б. Шарпан, О. В. Гусєва
This work is licensed under a Creative Commons Attribution 4.0 International License.
Authors who publish with this journal agree to the following terms:
- Authors retain copyright and grant the journal right of first publication with the work simultaneously licensed under a Creative Commons Attribution License that allows others to share the work with an acknowledgement of the work's authorship and initial publication in this journal.
- Authors are able to enter into separate, additional contractual arrangements for the non-exclusive distribution of the journal's published version of the work (e.g., post it to an institutional repository or publish it in a book), with an acknowledgement of its initial publication in this journal.
- Authors are permitted and encouraged to post their work online (e.g., in institutional repositories or on their website) prior to and during the submission process, as it can lead to productive exchanges, as well as earlier and greater citation of published work (See The Effect of Open Access).
