A-EGM Signal Processing

Analysis of intracardial electrograms


Focused on analysis of atrial electrograms recorded during cathether radiofrequency ablation of  Atrial Fibrillation.

Automated assessment of endocardial electrograms fractionation in human

Figure 1. Pre and Postablation maps recorded during radiofrequency catheter ablation (RFA).

Author and responsible person: Václav Křemen, Ph.D.

Cooperation with:

Department of Cardiology, IK+EM Prague, Czech republic.
Department of Cardiology, VFN Prague, Czech republic.

This is supported by the research projects #MSM 6840770012 "Transdisciplinary Biomedical Engineering Research II from Ministry of Education, Youth and Sports of the Czech Republic".

Your activity is welcome

We appreciate you cooperation

whether you are student or researcher or company

If this article "captured" you and You are interested to cooperate, don't hesitate to write an email to author:

vaclav.kremen (at) fel.cvut.cz


Current  Situation of the Studied Problem

Computer-based methods for analysis and interpretation of biological signals have been subject of intense research. Applications that perform signal processing and diagnostic  interpretation of signals are widely spread. One example is a visual analysis of long term electrocardiogram (ECG) signal (sometimes called Holter signal) or analysis of sleep electroencephalogram (EEG) that is tedious time-consuming and operator dependent.

It is obvious that automated systems for biological signals processing such as noise removal, patterns detection, arrhythmias classification and clustering considerably reduces the amount of time the medicine needs to spend and help to improve or support the judgement of physicians that perform the signal analysis.

Significant progress has also been achieved in the field of curative ablation of atrial fibrillation (AF) in recent years. While empirical isolation of pulmonary veins is usually an effective strategy in  paroxysmal AF, targeting extrapulmonary substrate within left (right) atrium is often necessary in the case of persistent/permanent AF [1]. Both areas with high dominant frequency of atrial electrograms (A-EGMs) [13] and areas with complex fractionated atrial electrograms (CFAEs) [8] were shown to play a role in the maintenance of the arrhythmia. In order to identify those sites, great effort has been made to describe the patterns of activation in AF [3] and to quantify general characteristics of A-EGMs either in time- or frequency-domain [9], [7], [12] and [16].

Recently, two software algorithms were implemented in commercially available mapping systems. The algorithm, incorporated in CARTO system (Biosense-Webster), was used for CFAEs visualization during real-time left atrium mapping in the feasibility study by Scherr et. al [14]. Another algorithm, integrated in EnSite NavX system (St. Jude Medical), was used for targeting CFAEs during left atrium ablation prior to pulmonary vein isolation [17]. Although the automated detection of CFAEs was in reasonable agreement with the classification performed by investigators [17] and [18], detailed study of the relationship between subjective perception of A-EGMs and their computerized quantification has never been performed. Moreover, currently used software algorithms require initial setting of specific input parameters making them, at least to some extent, operator-dependent.

Proposed Methodology of A-EGM Evaluation

Methodology of A-EGM processing


Figure 2. Methodology of A-EGM  processing, suggested by our current work.


Using the robust signal preprocessing algorithms, exhaustive optimization algorithms and testing of several signal processing, feature extraction and selection methods and classifiers we are still developing and suggesting following methodology of signal processing (Figure 2) and evaluation tailored to help to navigate the operator during AF ablation in real time.

The methodology consists of, currently for A-EGM signal, unique usage of following steps:

  1. Analysis and preprocessing of the signal using for this purpose specially designed wavelet filtering algorithm.
  2. Application of wavelet transform for signal preprocessing step and feature extraction steps.
  3. Feature extraction algorithms, that enable to extract more than 40 features to describe the A-EGM signal.
  4. Feature selection methods, either self-standing or wrapped into the classifiers, enabled to select most appropriate feature for the given task of evaluation of complexity of the A-EGM signal.
  5. Optimization, which is now hidden, but was extensively used during design and testing phase of the thesis.
  6. Classification, regression or evaluation of A-EGMs level of fractionation using several methods of artificial intelligence or statistical pattern recognition.

Described methodology of the signal processing allowes us to construct several systems that can effectively work with the A-EGM signal in realtime. Figure 2 shows the whole signal path from its recording by catheter touching the endocardial tissue, through the robust wavelet based filtering method, feature extraction and selection phase towards the regression or classification. The result of regression or classification can be then used as a number to be mapped into real electro-anatomical maps during the ablation of AF.

Figure 3. Different levels of fractionation of A-EGM signals recorded during RFA of AF.


We present biological signal processing that uses Wavelet transform and Statistical Pattern Recognition tools. Both have been extensively used in many different fields such as speech processing or ECG signal processing. Regarding intracardial signals area, the concept and use ofWavelet transform is new and a small number of contributions has appeared relatively compared to those in the rest of fields. In the field of CFAEs the presented signal processing approach and completely new methodology of A-EGMs processing and description has never been used in such comprehensive form. These operator-independent and fully automatic algorithms for A-EGMs complexity description is the first usage of such novel approaches in A-EGMs processing and analysis and may be easily incorporated into mapping systems to facilitate CFAEs identification and help to guide AF substrate ablation.

The presented work concerns a very specific type of A-EGM signals from the bipolar tip of a catheter recorded during electrophysiological mapping procedure during AF. Such specific signals needed to be specially recorded for the study under specific conditions because nowadays there is no such A-EGM signals database in the world that could be used for the study purposes. Forming a dataset for the study was one of the main parts of the work, showing completely new approach to finding new AF ablation strategies and A-EGM signals evaluation or to evolve recent AF ablation strategies. We consider, that the success of forming the dataset was confirmed by the partial results during all stages of the work and by the final results of classification.


Figure 4. Feature extraction (extractor) using a program, that works similary like NavX algorithm described by Scherr et al. [14].


A-EGM preprocessing phase is almost a prerequisite either in supervised or unsupervised learning and signal evaluation. Therefore, we focus on important steps of A-EGM processing: signal  denoising. The methods are being developed under the framework of wavelet transform which is very suitable for biological signals analysis. We conclude that the proposed approach of signal processing and evaluation works very sufficiently, which is obvious from the results of feature extraction and selection, and A-EGM classification. In case of A-EGM noise reduction we implement threshold technique based on the first five details of wavelet transform and thresholds found by PSO algorithm and developed optimization methodology. Here proposed filtering algorithm is able to remove the noise and to keep, for the consecutive signal evaluation, useful information. We conclude that the designed and implemented method outperformes the classical approaches as median or elliptic filters and that visual inspection shows the results, which look very clean (This does not have such big importance for the task of signal evaluation, but it is a nice side effect of the signal preprocessing.).

Furthermore we have been developing  a techniques to extract more A-EGM features that are based on several possible information dimensions (degree of freedom) of the A-EGM signal, for example entropy, DF and CFAEs based features, as well as time and frequency domain analysis features. We introduce the unique wavelet transform based algorithm for searching of FSs in A-EGM signal, sequentially followed by extraction algorithms that alow to mine the information of level of fractionation of the signal based on local electrical activity, which is automatically found by this algorithm [2].


Figure 5. Performance of several classifiers tested during classification phase of A-EGM evaluation.


We describe and test more than 40 features, and based on several selection criteria we selected the most important 15 features that entered the sequential steps of features evaluation and signal classification/evaluation. All the described steps to design new A-EGM filtering and feature extraction algorithms were comprehensively supported by in the thesis described optimization procedures that enabled to find best possible setup of the algorithms to maximize the effectiveness of the A-EGM classification/evaluation steps that were performed in next step of the work. Therefore the optimization forms an important bridge between signal preprocessing and evaluation to get ultimate results. The suggested and performed methodology of A-EGM preprocessing and description (feature extraction) was tested, its usability was confirmed and extended by several classification approaches. We found out and tested new regression models and classifiers of A-EGM signal to describe the level of fractionation (complexity) of the signal. Using the dataset, we showed that the models constructed here, using selected features, performed well and highly outperform still known approaches of A-EGM description, that used single feature only (DF and CFAEs). The constructed algorithms itself can be used self-contained on off-line analysis in real-time and can offer various and independent projection of the A-EGM signals during ablation of AF.


In conclusion, we propose a novel algorithms for automated and operator independent assessment of A-EGMs fractionation to facilitate CFAEs identification and to guide AF substrate ablation. Because of the low computational costs it can be easily incorporated into real-time mapping systems provided it will be first validated off-line in larger and independent A-EGMs sample. By now, its clinical value is unknown and warrants further investigation.


[1] M. Haissaguerre, M. Wright, M. Hocini, and P. Jais. The substrate maintaining persistent atrial fibrillation. Circ. Arrhythmia Electrophysiol., 1:2–5, 2008.

[2] V. Křemen, L. Lhotská, R. Čihák, V. Vančura, J. Kautzner, and D. Wichterle. A new approach to automated assessment of endocardial electrograms fractionation in human left atrium during atrial fibrillation. Physiol. Measurement, 2008.

[3] K. Kumagai. Patterns of activation in human atrial fibrillation. Heart Rhythm, 4(3):Suppl. S7–S12, 2008.

[4] S. Lazar, S. Dixit, F. Marchlinski, D. Callans, and E. Gerstenfeld. Presence of left-to-right atrial frequency gradient in paroxysmal but not persistent atrial fibrillation in humans. Circulation, 110:3181–3186, 2004.

[5] Y. Lin, C. Tai, T. Kao, S. Chang, W. Wongcharoen, L. Lo, T. Tuan, A. Udyavar, Y. Chen, S. Higa, K. Ueng, and S. Chen. Consistency of complex fractionated atrial electrograms during atrial fibrillation. Heart Rhythm, 5:406–12, 2008.

[6] Y. Lin, C. Tai, T. Kao, H. Tso, S. Higa, H. Tsao, S. Chang, M. Hsieh, and S. Chen. Frequency analysis in different types of paroxysmal atrial fibrillation. J. Am. Coll. Cardiol., 47:1401–1407, 2006.

[7] L. Mainardi, V. Corino, L. Lombardi, C. Tondo, M. Mantica, F. Lombardi, and S. Cerutti. Assessment of the dynamics of atrial signals and local atrial period series during atrial fibrillation: effects of isoproterenol administration. Biomed. Eng. Online, 3(37), 2004.

[8] K. Nademanee, J. McKenzie, E. Kosar, M. Schwab, B. Sunsaneewitayakul, T. Vasavakul, C. Khunnawat, and T. Ngarmukos. A new approach for catheter ablation of atrial fibrillation: mapping of the electrophysiologic substrate. J Am Coll Cardiol, 43:2044–53, 2004. 1,

[9] J. Ng, J. Gold, and J. Goldberger. Understanding and interpreting dominant frequency analysis of af electrograms. Journal of Cardiovascular Electrophysiology, pages 680–685, 2007. 1,

[10] J. Ng, A. Kadish, and J. Goldberger. Effect of electrogram characteristics on the relationship of dominant frequency to atrial activation rate in atrial fibrillation. Heart Rhythm, pages 1295–1305, 2006.

[11] M. Porter, W. Spear, J. Akar, R. Helms, N  Brysiewicz, P. Santucci, and D. Wilber. Prospective study of atrial fibrillation termination during ablation guided by automated detection of fractionated electrograms. J. Cardiovasc. Electrophysiol., 2008.

[12] F. Ravelli, L. Faes, L. Sandrini, F. Gaita, R. Antolini, M. Scaglione, and  G. Nollo. Wave similarity mapping shows the spatiotemporal distribution of fibrillatory wave complexity in the human right atrium during paroxysmal and chronic atrial fibrillation. J. Cardiovasc. Electrophysiol., 16:1071–1076, 2005. 1,

[13] P. Sanders, O. Berenfeld, M. Hocini, P. Jais, R. Vaidyanathan, L. Hsu, S. Garrigue, Y. Takahashi, M. Rotter, F. Sacher, C. Scavee, R. Ploutz-Snyder, J. Jalife, and M. Haissguerre. Spectral analysis identifies sites of high frequency activity maintaining atrial fibrillation in humans. Circulation, 112:789–797, 2005.

[14] D. Scherr, D. Dalal, A. Cheema, A. Cheng, C. Henrikson, D. Spragg, J. Marine, R. Berger, H. Calkins, and J. Dong. Automated detection and characterization of complex fractionated atrial electrograms in human left atrium during atrial fibrillation. Heart Rhythm, 4:10131020, 2007.

[15] M. Stiles, A. Brooks, B. John, Shashidhar, L. Wilson, P. Kuklik, H. Dimitri, D. Lau, R. Roberts-Thomson, L. Mackenzie, S. Willoughby, G. Young, and P. Sanders. The effect of electrogram duration on quantification of complex fractionated atrial electrograms and dominant frequency. J. Cardiovasc. Electrophysiol., 19:252–258, 2008.

[16] Y. Takahashi, M. O´ Neill, M. Hocini, R. Dubois, S. Matsuo, S. Knecht, S. Mahapatra, K. Lim, P. Jais, A. Jonsson, F. Sacher, P. Sanders, T. Rostock, P. Bordachar, J. Clementy, G. Klein, and M.  Haissaguerre. Characterization of electrograms associated with termination of chronic atrial fibrillation by catheter ablation. J. Am. Coll. Cardiol., 51:1003– 1010, 2008.

[17] A. Verma, P. Novak, L. Macle, B. Whaley, M. Beardsall, Z. Wulffhart, and Y. Khaykin. A prospective multicenter evaluation of ablating complex fractionated electrograms (CFEs) during atrial fibrillation (AF) identified by an automated mapping algorithm: acute effects on AF and efficacy as an adjuvant strategy. Heart Rhythm, 5:198–205, 2008.

[18] J.Wu, H. Estner, A. Luik, E. Ucer, T. Reents, A. Pflaumer, B. Zrenner, G. Hessling, and I. Deisenhofer. Automatic 3D mapping of complex fractionated atrial electrograms (CFAE) in patients with paroxysmal and persistent atrial fibrillation. J. Cardiovasc. Electrophysiol., 2008.

Our publications

In journals:

V. Křemen, R. Čihák, V. Vančura, J. Kautzner, L. Lhotská, and D. Wichterle. A new approach to automated assessment of endocardial electrograms fractionation in human left atrium during atrial fibrillation. Physiol Meas, submitted 2008. IOP Publishing.

Abstract  - D. Wichterle, V. Křemen, R. Čihák, V. Vančura, J. Kautzner. A novel method to identify Complex Fractionated Atrial Electrograms. J Am Coll Cardiol, 51 (10 Suppl. 1):A10. 2008.

Abstract - D. Wichterle, V. Křemen, R. Čihák, J. Kautzner. Comparison of two methods for identifying Complex Fractionated Atrial Electrograms. Journal of Cardiovascular Electrophysiology, 18 (Suppl. 2):35. 2007.

Abstract - D. Wichterle, V. Křemen, R. Čihák, V. Vančura, J. Kautzner. A subjective perception of Complex Fractionated Atrial Electrograms: Implications for electroanatomic mapping of atrial fibrillation. G Ital Aritmol Cardiostim, 10 (3 Suppl. 2):66. 2007.