gms | German Medical Science

49. Jahrestagung der Deutschen Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie (gmds)
19. Jahrestagung der Schweizerischen Gesellschaft für Medizinische Informatik (SGMI)
Jahrestagung 2004 des Arbeitskreises Medizinische Informatik (ÖAKMI)

Deutsche Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie
Schweizerische Gesellschaft für Medizinische Informatik (SGMI)

26. bis 30.09.2004, Innsbruck/Tirol

A Signal Processing Pipeline for Noninvasive Imaging of Ventricular Preexcitation

Meeting Abstract (gmds2004)

  • corresponding author presenting/speaker Gerald Fischer - Institut für Biomedizinische Signalverarbeitung und Bildgebung, UMIT - Private Universität für Gesundheitswissenschaften, Medizinische Informatik und Technik, Innsbruck, Österreich
  • Friedrich Hanser - Institut für Biomedizinische Signalverarbeitung und Bildgebung, UMIT - Private Universität für Gesundheitswissenschaften, Medizinische Informatik und Technik, Innsbruck, Österreich
  • Christoph Hintermüller - Institut für Biomedizinische Signalverarbeitung und Bildgebung, UMIT - Private Universität für Gesundheitswissenschaften, Medizinische Informatik und Technik, Innsbruck, Österreich
  • Michael Seger - Institut für Biomedizinische Signalverarbeitung und Bildgebung, UMIT - Private Universität für Gesundheitswissenschaften, Medizinische Informatik und Technik, Innsbruck, Österreich
  • Bernhard Pfeifer - Institut für Biomedizinische Signalverarbeitung und Bildgebung, UMIT - Private Universität für Gesundheitswissenschaften, Medizinische Informatik und Technik, Innsbruck, Österreich
  • Robert Modre - Institut für Biomedizinische Signalverarbeitung und Bildgebung, UMIT - Private Universität für Gesundheitswissenschaften, Medizinische Informatik und Technik, Innsbruck, Österreich
  • Leonhard Wieser - Institut für Biomedizinische Signalverarbeitung und Bildgebung, UMIT - Private Universität für Gesundheitswissenschaften, Medizinische Informatik und Technik, Innsbruck, Österreich
  • Bernhard Tilg - Institut für Biomedizinische Signalverarbeitung und Bildgebung, UMIT - Private Universität für Gesundheitswissenschaften, Medizinische Informatik und Technik, Innsbruck, Österreich
  • Siegrid Egger - University Hospital, Innsbruck, Österreich
  • Thomas Berger - University Hospital, Innsbruck, Österreich
  • Franz X Roithiger - University Hospital, Innsbruck, Österreich
  • Florian Hintringer - University Hospital, Innsbruck, Österreich

Kooperative Versorgung - Vernetzte Forschung - Ubiquitäre Information. 49. Jahrestagung der Deutschen Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie (gmds), 19. Jahrestagung der Schweizerischen Gesellschaft für Medizinische Informatik (SGMI) und Jahrestagung 2004 des Arbeitskreises Medizinische Informatik (ÖAKMI) der Österreichischen Computer Gesellschaft (OCG) und der Österreichischen Gesellschaft für Biomedizinische Technik (ÖGBMT). Innsbruck, 26.-30.09.2004. Düsseldorf, Köln: German Medical Science; 2004. Doc04gmds098

Die elektronische Version dieses Artikels ist vollständig und ist verfügbar unter: http://www.egms.de/de/meetings/gmds2004/04gmds098.shtml

Veröffentlicht: 14. September 2004

© 2004 Fischer et al.
Dieser Artikel ist ein Open Access-Artikel und steht unter den Creative Commons Lizenzbedingungen (http://creativecommons.org/licenses/by-nc-nd/3.0/deed.de). Er darf vervielf&aauml;ltigt, verbreitet und &oauml;ffentlich zug&aauml;nglich gemacht werden, vorausgesetzt dass Autor und Quelle genannt werden.


Gliederung

Text

Introduction

In patients having overt ventricular preexcitation an anterogradely conducting accessory atrioventricular (AV) or atriofascicular pathway forms a second input to ventricular activation. Clinically, the vast majority of patients presents with an accessory AV pathway, which is manifested by a typical delta-wave on the ECG. This variant of ventricular preexcitation is known as the Wolff-Parkinson-White (WPW) syndrome. Its prevalence is estimated to be 0.1% to 0.3% [1].

Noninvasive Imaging of the ventricular activation sequence has recently been subject of successful validation in humans [2]. This method is of potential interest for the interventional curative treatment of WPW as it can potentially identify the location of the AV pathway as the target substrate for radio frequency catheter ablation. For a clinical application the method needs a to a high degree automated preprocessing of multi-channel ECG mapping data. The central point of this study is the development of a signal processing pipeline tailored to the potential application of imaging ventricular preexcitation.

Methods

Eight patients (3 female) underwent radio frequency ablation of an AV pathway. During the electrophysiological study an intravenous bolus injection of adenosine (12 mg in patients with a body weight below 50 kg, 18 mg in others) was applied for confirming the presence of an anterograde conducting accessory pathway. Adenosine has a to a high degree specific effect on the AV-node, blocking the conduction in this pathway. Thus, under the effect of adenosine the AV pathway is the only input for ventricular activation yielding clinical evidence for a preexcitation syndrome. Due to the short half time of adenosine (< 1.5 s) typically only a short train of beats reflects an altered morphology due to a single ventricular input. A catheter in the right ventricular apex was used for backup pacing (coupling interval 1500 ms) to prevent from ventricular arrest in the case of a complete AV-block. At the time of drug admission a 62 channel ECG map recording was started and a segment of 20 s was recorded.

The signal processing pipeline should automatically identify the beats during the short AV nodal block as they contain most information of the accessory pathway location. It has to perform the following tasks: beat identification (R-peak detection), signal classification, extraction of the time interval of ventricular depolarization and baseline correction. In this abstract we will mainly focus on signal classification. The R-peak detector is the algorithm described in [3], modified for the application to multi-channel data. The time interval of ventricular depolarization is assessed from the root mean square (rms) plot [4] using empirically determined threshold values. For an R-R interval > 600 ms the baseline correction is carried out automatically 250 ms prior to the R-peak.

Signal classification is based on a distance measure D. For each QRS-complex a feature matrix is created from all samples φ mn in the interval 40 ms prior and 60 ms after the R-peak. In order to minimize the influence of possibly drifting channels the direct current content of all M samples per channel n is set to zero. Writing φ* mn for the elements of the reference beat and φ mn for the elements of the actual beat we define the distance measure D:

(1) Equation 1

We term beats with an normal WPW-morphology 'W'-beats (mixed conduction AV-pathway and AV-node). Beats during adenosine induced AV-block we term 'A'-beats. We found that for comparing two 'W' beats D is typically < 40 μV due to the small beat to beat variability. Comparing an 'A'-beat with a 'W'-morphology reference D is in the order of 100 μV to 200 μV, with small beat to beat variations (about 20 μV) but large inter-individual variations. Comparing a beat with a different morphology (ventricular ectopics or ventricular pacing) with a 'W'-morphology reference D was always larger than the distance value found when comparing 'W' and 'A' morphology in the same individuum. Biophysically, this can be explained having in mind that for both 'A' and 'W'-beats the accessory pathway is the input for first ventricular activation. Thus, the difference in the morphology is smaller in this case, compared to beats where the first activation originates from a different location. Based on this observations the following algorithm was developed which automatically classifies the beats in 'W' and 'A'-morphology leaving a subset of beats (e.g. ectopics) unclassified. The classification of 'W'-beats is trivial (D i ≤ 40 μV). The search for 'A'-beats is performed using a candidate list, which includes all beats differing from an estimated mean 'A'-'W' distance Δ A by less than a threshold value d. In an iterative process Δ A is updated and the threshold is halved refining the search. The search is stopped when d ≤ 40 μV. The values used in this algorithm have been selected retrospectively:

Select a 'W' reference beat by an human expert.

Compute D i for each beat.

Classify all beats with D i ≤ 40 μ V as 'W'-beats.

Initialize search for 'A'-beats: Δ A =160 μV, d=160 μV.

repeat until: d ≤ 40 μV do:

Include all beats with Δ A - d/2 ≤ D i ≤ Δ A + d/2 in the 'A'-beat candidate list.

Update Δ A computing the mean distance measure of the candidate list.

Halve threshold d.

od repeat

Classify all beats in the candidate list as 'A'-beats.

Results

In all 8 recordings 176 R-peaks were identified by a human expert. Exactly the same 176 R-peaks were found by the automatic R-peak detector (100% sensitivity and detection rate). For 'A'-beats the depolarization interval should be prolonged as the activation via the normal conduction system is absent. This gives the possibility of testing the classification algorithm independently from the distance measure D. Table 1 [Tab. 1] lists the results (number of classified beats n, duration T and standard deviation σ).

For all patients the prolongation of ventricular depolarization could be demonstrated by a single tailed students t-test with statistical significance (p<0.001). Additionally 9 ventricular ectopics (VE) and 4 paced beats (VP) were observed in the recordings. All of them were correctly left unclassified by our algorithm.

Conclusion

The implemented signal processing pipeline is a powerful tool for selecting target beats for noninvasive activation imaging in WPW patients. It robustly identifies and classifies beats. The small beat to beat variations in the depolarization interval detection indicate accurate identification of the time window of interest.

Acknowledgement

This study was supported by the Austrian Science Fund (FWF) under grants START Y-144-N04 and P16579-N04.


References

1.
Yee R, Klein GJ, Prystowsky E: The Wolff-Parkinson-White syndrome and related variants. In: Zipes DP, Jalife J, Editors. Cardiac Electrophysiology - From Cell to Bedside. W.B. Saunders, Philadelphia, 1999: 845-861.
2.
Tilg B, Fischer G, Modre R, Hanser F, Messnarz B, Schocke M, Kremser C, Berger T, Hintringer F, Roithinger FX. Model-based imaging of cardiac electrical excitation in humans. IEEE T Med Imag 2002 21:1031-9.
3.
Benitez D, Gaydecki PA, Zaidi A, Fitzpatrick AP. The use of the Hilbert transform in ECG signal analysis. Comput Biol Med 2001 31:399-406.
4.
Gozolits S, Fischer G, Berger T, Hanser F, Abou-Harb M, Tilg B, Pachinger O, Hintringer F, Roithinger FX. The global P-wave duration in the 65-Lead ECG: single and dual site pacing in the structurally normal human atria. J Cardiovasc Electrophysiol 2002 13:1240-5.