Advertisement

Somnologie

, Volume 21, Issue 2, pp 110–120 | Cite as

Extended algorithm for real-time pulse waveform segmentation and artifact detection in photoplethysmograms

  • C. Fischer
  • M. Glos
  • T. Penzel
  • I. Fietze
Original Article

Abstract

Background

Photoplethysmography can be used for measuring oxygen saturation or assessing autonomic function. Artifacts can render the photoplethysmogram (PPG) useless. Thus, algorithms capable of identifying artifacts are important. However, the published algorithms are limited in their abilities and study design. Therefore, the authors developed a novel embedded algorithm for pulse waveform (PWF) segmentation and artifact detection in real-time.

Objectives

The previous PWF analysis was not able to detect a diastolic peak, which prevents analyses like arterial stiffness. Furthermore, the algorithm shows room for improvements if the first part of the pulse wave is disturbed. To overcome these limitations, the authors extended the PWF analysis.

Materials and methods

The extended PWF analysis was validated as before with the data records from 63 subjects acquired in a sleep laboratory, ergometry laboratory, and intensive care unit. The output of the algorithm was compared with harmonized experts’ annotations with a total duration of 31.5 h.

Results

The comparison of the artifacts detection performance between the extended and the original PWF analysis shows a reduced sensitivity from 99.6 to 99.5%, but increased specificity from 90.5 to 91.6%, precision from 98.5 to 98.6%, accuracy from 98.3 to 98.4%, Cohen’s kappa coefficient from 0.927 to 0.932, and F‑measure from 0.990 to 0.991. Furthermore, the PWF analysis is now able to detect diastolic peaks.

Conclusion

The proposed novel extended PWF analysis seems to be a suitable method for real-time annotation of the PPG, and detection of pulse wave amplitude, pulse wave duration, rise time, pulse propagation time as well as their variations.

Keywords

Oxygen saturation Autonomic function Blood volume Diastolic peaks Vascular stiffness 

Erweiterter Algorithmus zur Echtzeit-Pulswellensegmentierung und Artefakterkennung in Photoplethysmogrammen

Zusammenfassung

Hintergrund

Die Photoplethysmographie kann verwendet werden, um die Sauerstoffsättigung zu messen oder das vegetative Nervensystem zu untersuchen. Artefakte können das Photoplethymogramm (PPG) bis zur Unkenntlichkeit verzerren. Daher sind Algorithmen wichtig, die in der Lage sind, Artefakte zu erkennen. Die bisher publizierten Algorithmen sind aber in ihren Fähigkeiten und ihrem Studiendesign limitiert. Deshalb entwickelten die Autoren einen neuartigen integrierten Algorithmus zur Segmentierung der Pulswellenform (PWF) und Artefakterkennung in Echtzeit.

Ziel der Arbeit

Die vorangegangene PWF-Analyse konnte keinen diastolischen Peak erkennen, wodurch z. B. Untersuchungen der arteriellen Steifigkeit nicht möglich waren. Zudem zeigte der Algorithmus Verbesserungspotenzial bei im Anfangsbereich gestörten Pulswellen. Um diese Einschränkungen zu beseitigen, erweiterten die Autoren die PWF-Analyse.

Material und Methoden

Die erweiterte PWF-Analyse wurde wie zuvor mit den Datensätzen von 63 Personen validiert, die in einem Schlaflabor, Ergometrielabor und auf Intensivstation erfasst worden waren. Die Ergebnisse des Algorithmus wurden mit den abgeglichenen Expertenauswertungen mit einer Gesamtdauer von 31,5 h verglichen.

Ergebnisse

Der Vergleich der Leistung bei der Artefakterkennung zwischen der erweiterten und der Original-PWF-Analyse zeigt eine reduzierte Sensitivität von 99,6 % auf 99,5 %, während sich die Spezifität von 90,5 % auf 91,6 %, der positive Vorhersagewert von 98,5 % auf 98,6 %, die Vertrauenswahrscheinlichkeit von 98,3 % auf 98,4 %, der Kappa-Koeffizient nach Cohen von 0,927 auf 0,932 und das F‑Maß von 0,990 auf 0,991 verbesserten. Darüber hinaus lassen sich nun mit der PWF-Analyse diastolische Peaks erkennen.

Schlussfolgerung

Die neue, erweiterte PWF-Analyse scheint eine brauchbare Methode für die Echtzeitauswertung von PPG und die Bestimmung der Pulswellenamplitude, Pulswellendauer, Anstiegszeit, Pulsausbreitungszeit sowie deren Variationen zu sein.

Schlüsselwörter

Sauerstoffsättigung Autonome Funktion Blutvolumen Diastolische Spitzenwerte Vaskuläre Steifigkeit 

Notes

Compliance with ethical guidelines

Conflict of interest

The development of the original PWF analysis was supported by the German Federal Ministry of Education and Research (BMBF) and MCC GmbH & Co. KG. During the database recording and the development of the original PWF analysis, C. Fischer was in an employment relationship with the company MCC GmbH & Co. KG. For the development of the extended PWF anlaysis C. Fischer, M. Glos, T. Penzel, and I. Fietze declare that they have no competing interests.

All procedures followed were in accordance with the ethical standards of the responsible committee on human experimentation (institutional and national). Informed consent was obtained from all patients included in the study.

References

  1. 1.
    Association for the Advancement of Medical Instrumentation (2008) ANSI/AAMI/ISO EC57:1998/(R)2008. AAMI, ArlingtonGoogle Scholar
  2. 2.
    Aboy M, McNames J, Thong T et al (2005) An automatic beat detection algorithm for pressure signals. IEEE Trans Biomed Eng 52:1662–1670. doi: 10.1109/tbme.2005.855725 CrossRefPubMedGoogle Scholar
  3. 3.
    Adachi H, Mikami A, Kumano-go T et al (2003) Clinical significance of pulse rate rise during sleep as a screening marker for the assessment of sleep fragmentation in sleep-disordered breathing. Sleep Med 4:537–542CrossRefPubMedGoogle Scholar
  4. 4.
    Allen J (2007) Photoplethysmography and its application in clinical physiological measurement. Physiol Meas 28:R1–R39. doi: 10.1088/0967-3334/28/3/r01 CrossRefPubMedGoogle Scholar
  5. 5.
    Bronshtein IN, Semendyayev KA, Musiol G, Mühlig H (1999) Taschenbuch der Mathematik, 4th edn. Harri Deutsch, Frankfurt a. M.Google Scholar
  6. 6.
    Cary JM, Krugmeier R, Newman B et al (1984) Coughs cause systemic blood flow. Thorax 39:192–195. doi: 10.1136/thx.39.3.192 CrossRefPubMedPubMedCentralGoogle Scholar
  7. 7.
    Chan M, Estève D, Fourniols J‑Y et al (2012) Smart wearable systems: current status and future challenges. Artif Intell Med 56:137–156. doi: 10.1016/j.artmed.2012.09.003 CrossRefPubMedGoogle Scholar
  8. 8.
    Cohen J (1960) A coefficient of agreement for nominal scales. Educ Psychol Meas 20:37–46. doi: 10.1177/001316446002000104 CrossRefGoogle Scholar
  9. 9.
    Dawber TR, Thomas HE, McNamara PM (1973) Characteristics of the dicrotic notch of the arterial pulse wave in coronary heart disease. Angiology 24:244–255CrossRefPubMedGoogle Scholar
  10. 10.
    Elgendi M (2012) Standard terminologies for photoplethysmogram signals. Curr Cardiol Rev 8:215–219. doi: 10.2174/157340312803217184 CrossRefPubMedPubMedCentralGoogle Scholar
  11. 11.
    Elgendi M (2012) On the analysis of fingertip photoplethysmogram signals. Curr Cardiol Rev 8:14–25CrossRefPubMedPubMedCentralGoogle Scholar
  12. 12.
    Fischer C, Dömer B, Wibmer T, Penzel T (2016) An algorithm for real-time pulse waveform segmentation and artifact detection in photoplethysmograms. IEEE J Biomed Health Inform:1–1. doi: 10.1109/jbhi.2016.2518202 PubMedGoogle Scholar
  13. 13.
    Fleming PR (1957) The mechanism of the pulsus bisferiens. Br Heart J 19:519–524CrossRefPubMedPubMedCentralGoogle Scholar
  14. 14.
    General Services Administration (1996) Federal standard 1037 C. Government institutesGoogle Scholar
  15. 15.
    Hertzman AB (1938) The blood supply of various skin areas as estimated by the photoelectric plethysmograph. Am J Physiol Regul Integr Comp Physiol 124:328–340Google Scholar
  16. 16.
    Hripcsak G, Rothschild AS (2005) Agreement, the f‑measure, and reliability in information retrieval. J Am Med Inform Assoc 12:296–298. doi: 10.1197/jamia.m1733 CrossRefPubMedPubMedCentralGoogle Scholar
  17. 17.
    Iber C, Ancoli-Israel S, Chesson AL Jr, Quan SF (2007) The AASM manual for the scoring of sleep and associated events, 1st edn. American Academy of Sleep Medicine, WestchesterGoogle Scholar
  18. 18.
    Karlen W, Turner M, Cooke E et al (2010) CapnoBase: signal database and tools to collect, share and annotate respiratory signalsGoogle Scholar
  19. 19.
    Klass DW (1995) The continuing challenge of artifacts in the EEG. Am J EEG Technol 35:239–269Google Scholar
  20. 20.
    Lewis T (1907) The pulsus bisferiens. Br Med J 1:918–920CrossRefPubMedPubMedCentralGoogle Scholar
  21. 21.
    Mannheimer PD (2007) The light-tissue interaction of pulse oximetry. Anesth Analg 105:10–17. doi: 10.1213/01.ane.0000269522.84942.54 CrossRefGoogle Scholar
  22. 22.
    Murray WB, Foster PA (1996) The peripheral pulse wave: information overlooked. J Clin Monit 12:365–377. doi: 10.1007/s11517-005-0008-y CrossRefPubMedGoogle Scholar
  23. 23.
    Penzel T, Kantelhardt JW, Lo C‑C et al (2003) Dynamics of heart rate and sleep stages in normals and patients with sleep apnea. Neuropsychopharmacology 28(Suppl 1):48–53. doi: 10.1038/sj.npp.1300146 CrossRefGoogle Scholar
  24. 24.
    Petterson MT, Begnoche VL, Graybeal JM (2007) The effect of motion on pulse oximetry and its clinical significance. Anesth Analg 105:78–84. doi: 10.1213/01.ane.0000278134.47777.a5 CrossRefGoogle Scholar
  25. 25.
    Saeed M, Villarroel M, Reisner AT et al (2011) Multiparameter intelligent monitoring in intensive care II: a public-access intensive care unit database. Crit Care Med 39:952–960. doi: 10.1097/ccm.0b013e31820a92c6 CrossRefPubMedPubMedCentralGoogle Scholar
  26. 26.
    Shelley KH, Jablonka DH, Awad AA et al (2006) What is the best site for measuring the effect of ventilation on the pulse oximeter waveform? Anesth Analg 103:372–377. doi: 10.1213/01.ane.0000222477.67637.17 CrossRefPubMedGoogle Scholar
  27. 27.
    Shin KG, Ramanathan P (1994) Real-time computing: a new discipline of computer science and engineering. Proc IEEE 82:6–24. doi: 10.1109/5.259423 CrossRefGoogle Scholar
  28. 28.
    Strijbos J‑W, Martens RL, Prins FJ, Jochems WMG (2006) Content analysis: What are they talking about? Comput Educ 46:29–48. doi: 10.1016/j.compedu.2005.04.002 CrossRefGoogle Scholar
  29. 29.
    Sweeney KT, Ward TE, McLoone SF (2012) Artifact removal in physiological signals-practices and possibilities. IEEE Trans Inf Technol Biomed 16:488–500. doi: 10.1109/titb.2012.2188536 CrossRefPubMedGoogle Scholar
  30. 30.
    Tobin RM, Pologe JA, Batchelder PB (2002) A characterization of motion affecting pulse oximetry in 350 patients. Anesth Analg 94:54–61Google Scholar

Copyright information

© Springer Medizin Verlag GmbH 2017

Authors and Affiliations

  1. 1.Roche Diabetes Care GmbHMannheimGermany
  2. 2.Interdisziplinäres Schlafmedizinisches ZentrumCharité – Universitätsmedizin BerlinBerlinGermany
  3. 3.International Clinical Research CenterSt. Anne’s University Hospital BrnoBrnoCzech Republic
  4. 4.BruchsalGermany

Personalised recommendations