176

Characterization of Onset Mechanism and WaveformAnalysis in Patients with Atrial Fibrillation Usinga High-Resolution Noncontact Mapping System

STEFAN WEBER, M.D., GJIN NDREPEPA, M.D., MICHAEL SCHNEIDER, M.D.,BERNADETTE GEISSLER, JURGEN SCHREIECK, M.D., MARTIN KARCH, M.D.,

SEBASTIAN SCHMIEDER, M.D., ISABEL DEISENHOFER, M.D.,BERNHARD ZRENNER, M.D., ALBERT SCHOMIG, M.D., and CLAUS SCHMITT, M.D.

From Deutsches Herzzentrum Munchen and 1. Medizinische Klinik, Klinikum rechts der Isar,Technische Universitat Munchen, Munich, Germany

Characterization of AF with Noncontact Mapping. Introduction: Information on the spatiotem-poral organization of atrial activity at the onset of atrial fibrillation (AF) is limited.

Methods and Results: The study consisted of 26 consecutive patients (22 men and 4 women; mean age56 ± 9 years) with AF in whom the left atrium (LA) was mapped using a noncontact mapping system. At theonset of AF, the AF cycle lengths and wavefronts were analyzed at the site of origin of the triggering atrialpremature complex (APC) and five predefined sites within the LA (superior, anterior, posterior, lateral, andseptal walls). If repetitive activity was observed at the site of origin of APCs, triggered AF episodes wereconsidered as focally driven. APCs that induced AF had shorter coupling intervals than APCs that did notinduce AF (300 ± 41 msec vs 392 ± 64 msec, P < 0.001). Immediately after AF onset, repetitive firing wascrucial for maintenance of arrhythmia in 52 (80%) of 65 AF episodes. In 13 AF onset episodes (20%), AFwas maintained by other mechanisms. The number of LA wavefronts after AF onset was lower in focallydriven AF episodes compared with episodes in which no focally driven activity was observed (1.9 ± 0.6 v.2.3 ± 2.3 wavefronts, P < 0.05). After the onset of AF, the posterior wall of the LA showed the earliestdisorganized activity (after 5.2 ± 3.1 cycles).

Conclusion: In the majority of AF episodes (80%), repetitive firing from the triggering foci may play animportant role in maintaining AF immediately after arrhythmia onset. In 20% of the episodes, AF at earlystages seems to be maintained by other mechanisms. The capability of APCs to induce AF depends on thecoupling interval and the focus localization. The posterior wall of the LA shows the earliest disorganizationof wavefronts at the onset of AF. (J Cardiovasc Electrophysiol, Vol. 14, pp. 176-181, February 2003)

atrial fibrillation, noncontact mapping, left atrium

Introduction

Recent studies have considered the pulmonary vein(PV)-left atrium (LA) complex to be the most common originof focal activity that triggers the onset of atrial fibrillation(AF).1,2 However, little is known about the activation se-quence and the spatiotemporal organization of wavefronts atthe beginning of AF.3 Furthermore, it is still unclear whetherthe triggering ectopy simply puts the fibrillatory activity intomotion at the beginning of AF or if it plays a role in main-taining the arrhythmia. The information obtained with con-ventional mapping tools provides only a sketchy view of LAactivation at the time of AF onset.4,5 In this study, we useda three-dimensional noncontact mapping system (EnSite,Endocardial Solutions, St. Paul, MN, USA) to perform wave-front analysis in different regions of the LA at the onset ofAF. The system enables evaluation of regional and global LAactivity during onset of AF and can detect continuation of the

Address for correspondence: Stefan Weber M.D., Deutsches HerzzentrumMunchen, Lazarettstrasse 36, D80636, Munchen, Germany. Fax: 49-89-12184593; E-mail: webers@dhm.mhn.de

Manuscript received 18 July 2002; Accepted for publication 27 November2002.

firing from triggering foci after establishment of fibrillatoryactivation (conduction) within the atria.

The aim of this study was to characterize the electrophys-iologic processes that evolve at the onset of AF, with specialemphasis on the spatiotemporal organization of the wave-fronts, AF onset mechanism, and possible role of continua-tion of ectopic firing in maintenance of AF.

Methods

Study Population

The study population consisted of 26 consecutive patients(22 men and 4 women; mean age 56 ± 9 years, range 31–69) with paroxysmal (n = 22) or persistent (n = 4) AF.Patients with paroxysmal AF had a history of documentedself-terminating episodes of AF that were recorded by 7-daycontinuous Holter monitoring. Patients with persistent AFwere continuously in AF for at least 3 months before elec-trophysiologic study. Structural heart disease was present in14 patients (coronary artery disease in 4, hypertension withleft ventricular hypertrophy and/or reduced left ventricularfunction in 13, and cardiomyopathy in 1). In 12 patients, nostructural disease was documented. All patients included inthis study were treated unsuccessfully with at least two antiar-rhythmic drugs. All antiarrhythmic drugs were discontinued

Weber et al. Characterization of AF with Noncontact Mapping 177

for ≥5 plasma half-lives before electrophysiologic study.LA longitudinal and transverse echocardiographic diameterswere 52 ± 9 mm and 38 ± 5 mm, respectively. All patientsunderwent transesophageal echocardiography to rule out LAthrombi.

Noncontact Mapping

The noncontact mapping system (EnSite 3000, Endocar-dial Solutions) and its components and principles were de-scribed previously.6-8 In brief, the noncontact system consistsof a multielectrode array (MEA) mounted on a catheter. TheMEA enables collection of endocardial far-field potentialsfrom the surrounding chamber. The potentials are recorded,amplified, and processed by the system. Electrical activity ofthe surrounding endocardial walls is calculated using the in-verse equation of La Place and the boundary element method.By using virtual electrodes that can be positioned at any pointof interest on the virtual reconstructed geometry, endocardialactivity can be visualized at any location as a series of unipolarelectrograms. Based on this information, color-coded isopo-tential maps of the propagating wavefronts can be displayed.

Virtual Reconstruction of LA Anatomy

All patients included in this study gave written informedconsent. The noncontact mapping catheter and the conven-tional mapping/ablation catheter were deployed in the LA viatransseptal approach. The noncontact mapping catheter wasstabilized via a guidewire positioned in one of the left-sidedPVs (mostly the left upper PV). The conventional mappingcatheter was dragged around the endocardial contour of theLA to collect anatomic points to define the endocardial sur-face and reconstruct the geometry of the LA. The conven-tional mapping catheter was inserted in all four PVs, and theostial regions were marked on the reconstructed geometrymodel (Fig. 1).

Mapping of AF Onset and Ectopic Triggers

AF cycle lengths and wavefronts were analyzed at the siteof origin of the triggering atrial premature complex (APC)and five predefined regions in the LA (superior, anterior, pos-terior, lateral, and septal wall). For anterior, superior, and pos-terior LA regions, three unipolar electrograms were recordedat the septal, mid-portion, and lateral aspects of each wall.For septal and lateral regions, unipolar electrograms wererecorded from anterior, superior, posterior, and mid-portionaspects of each of the walls. AF cycle lengths (fibrillationintervals) were measured until no sharp atrial potential waspresent or up to 10 cycles after the initiating APC ensued.Activity was considered to be disorganized if virtual unipo-lar electrograms had multiple peaks of different amplitudesthat were not separated by an isoelectric line.

Ectopic activity was analyzed using the reconstructedisopotential maps. Spontaneous APC that did or did not in-duce AF were traced back until the earliest electrical activitywas defined. The spots of earliest activity (foci) were markedon the virtual geometry of the LA. Activity at the originof the triggering APC was traced for multiple activations inorder to define active firing after fibrillatory activation pat-terns were observed in the LA. AF episodes lasting morethan 30 seconds were analyzed. Propagation of the wave-fronts was followed, and the wavefront divisions (generationof daughter waveforms) were searched for and analyzed. If

more than three simultaneous wavefronts were observed, fur-ther wavefront divisions could not be defined accurately andthe activity was classified as multiple wavelet activity. In pa-tients with absent or rare ectopic activity at the time of theprocedure, atrial stimulation and orciprenaline infusion wereused to provoke APC and AF onset episodes. In all patientswho were in AF at the beginning of the procedure or whodeveloped AF episodes that lasted more than 15 minutes, in-ternal cardioversion was performed using a single lead system(AlertR, EPMedSystems, West Berlin, NJ, USA). A balloon-guided defibrillation catheter was inserted intravenously andpositioned with the defibrillation electrode arrays in the leftbranch of the pulmonary artery and the lateral wall of theright atrium, respectively.9

Statistical Analysis

Data are given as mean ± SD, range or percentage. Atwo-tailed, unpaired Student’s t-test was used to perform in-tergroup and intragroup comparisons of the continuous vari-ables. Comparison of more than two groups was performedwith one-way analysis of variance. Differences were consid-ered statistically significant at P < 0.05.

Results

Anatomic Location of Arrhythmogenic Foci

All patients included in the study showed APC either spon-taneously or under the effect of provocative maneuvers. In19 patients (73%), APC triggered episodes of AF. In the re-maining 7 patients (27%), only APCs were observed; no AFepisodes occurred. The noncontact mapping system tracedback the APC to the origin of the earliest electrical activ-ity and identified a total of 40 separate anatomic foci, 28 ofwhich generated AF episodes and 12 generated only APC.A single arrhythmogenic focus was observed in 15 patients(58%), whereas multiple foci (two or more) were observedin 11 patients (42%). Of the patients with a single focus, 10had foci located in the PV (left upper 4, right upper 3, leftlower 2, and right lower 1)3 and 5 patients had extra-PV foci(posterior region 3, posteroseptal region 1, and anteroseptalregion 1). In the group of patients with more than one focus,8 patients had 2 firing foci and 3 patients showed 3 sepa-rate arrhythmogenic foci. Six patients showed 2 extra-PVfoci, 1 patient showed 2 PV foci, and 1 patient showed 1 PVfocus and 1 extra-PV focus. In each of the remaining 3 pa-tients, 3 foci of different locations were observed. A total of65 AF episodes triggered by 28 LA foci occurred. Twentysix AF onsets occurred within 30 seconds after internalcardioversion.

Analysis of Cycle Length at the Beginning of AF

The AF cycle lengths of the consecutive activations fromthe first APC up to the 10th activation or the initiation ofdisorganized activity were measured. A total of 6,050 cy-cle lengths were used for analysis. The distribution of theAF cycle lengths showed an overall septal-to-lateral gradientwith shorter AF cycle lengths in the septal region (185 ±38 msec vs 194 ± 41 msec for patients with paroxysmalAF and 175 ± 24 msec vs 179 ± 25 msec for patientswith persistent AF, P < 0.001 for both comparisons). Thisgradient in the distribution of AF cycle lengths immedi-ately after the onset of AF was found both in patients with

178 Journal of Cardiovascular Electrophysiology Vol. 14, No. 2, February 2003

Figure 1. Bipolar electrogram recorded from the mappingcatheter and virtual electrograms obtained by noncontact mappingsystem during a spontaneous onset of atrial fibrillation triggeredby an atrial premature complex emerging from the right upper pul-monary vein (RUPV). The upper panel shows the isopotential mapof the left atrium with earliest activity just emerging from the RUPVregion. The positions of all four pulmonary veins were marked onthe reconstructed model of the left atrium. The lower panel showssurface ECG lead II, bipolar electrogram (Map 1/2), and virtualunipolar electrograms (virtual 6 to virtual 8) recorded from thesame region. The first two beats are sinus beats. The third beat isan atrial premature complex that initiated atrial fibrillation. Bothtypes of recordings show a gradual reduction in electrogram am-plitude, a gradual increase in electrogram duration, and initiationof electrogram fractionation (the last beats of the tracing). LLPV= left lower pulmonary vein; LUPV = left upper pulmonary vein;RLPV = right lower pulmonary vein.

intermittent and persistent AF. In patients with persistent AF,mean cycle length was 11% shorter than in patients withintermittent AF (Table 1). The cycle length from the initi-ating APC up to the 10th cycle after AF onset shortenedgradually. Comparison of patients with persistent AF andthose with intermittent AF showed an earlier and more rapiddecline in AF cycle length in patients with persistent AF(Fig. 2). APCs that induced AF had significantly shorter cou-pling intervals than APCs that did not induce AF (300 ±41 msec vs 392 ± 64 msec, P < 0.001; Fig. 3). Of the APCsthat induced AF, those of extra-PV origin showed signifi-cantly shorter coupling intervals as compared to APCs of PVorigin (285 ± 47 msec vs 311 ± 33 msec, P < 0.001).

Figure 3. Initiation of atrial fibrillation (AF) from a focus lo-cated in mid-posterior wall of the left atrium (LA). The upperpanel shows the isopotential map (posterior aspect of the LA).The lower panel shows surface ECG lead II and three virtualelectrograms (virtual 6 to virtual 8) that were recorded from therespective positions in the LA map. The first two atrial prematurecomplexes (APCs) with a cycle length of 320 ms did not induce AF.The third APC with a cycle length of 281 ms initiated AF. Degener-ation into AF was observed after the fourth activation. LLPV leftlower pulmonary vein; LUPV = left upper pulmonary vein; RLPVright lower pulmonary vein; RUPV = right upper pulmonaryvein.

Spatiotemporal Organization of Wavefronts

Analysis of spatiotemporal organization of waveforms atthe early stages of AF in different regions of the LA showedsignificant local differences. After the onset of AF, the firstLA region that showed disorganized activity was the poste-rior wall (after 5.2 ± 3.1 cycles), followed by septal (6.0 ±3.0 cycles), lateral (6.8 ± 3.0 cycles), superior (7.0 ± 2.8 cy-cles), and anterior (7.1 ± 2.8 cycles) (P < 0.0001). With theexception of the anterior wall of the LA, all other LA regionsdisorganized earlier during AF episodes initiated by APCsof extra-PV origin compared with AF episodes initiated byAPC of PV origin (Table 2 and Figs. 3 and 4).

Weber et al. Characterization of AF with Noncontact Mapping 179

Figure 2 Atrial fibrillation (AF) cycle lengths at the onset of AF in patientswith paroxysmal AF (upper graph) and persistent AF (lower graph). In thepatients with persistent AF, cycle lengths at AF onset were shorter than inpatients with paroxysmal AF.

Persistence of Focal Firing after Onset of AF

In 52 (80%) of 65 AF episodes, the ectopic foci that trig-gered AF continued to fire repetitively after the onset of AF.These episodes were defined as focally driven AF episodes.In 13 AF episodes (20%), no repetitive firing from distinctfoci could be observed. In 52 focally driven AF episodes, dis-tinct repetitive firing from the triggering foci lasted for 4.0 ±2.2 cycles. The number of LA wavefronts after initiation ofAF was lower in focally driven AF episodes compared withepisodes in which no focally driven activity was observed(1.9 ± 0.6 vs 2.3 ± 2.3 wavefronts, P < 0.05). Disorganizedactivity in the LA was observed later in focally driven AFepisodes compared with AF episodes in which no repetitivefiring was observed (7.6 ± 2.6 vs 4.6 ± 2.5 cycles, P < 0.001).At the beginning of AF, generation of more than three wave-fronts (Fig. 5) was seen earlier in the AF episodes in whichno persistence of repetitive activity was detectable (2.6 ± 1.9vs 4.3 ± 1.7 cycles after AF initiation, P < 0.005).

TABLE 1

Regional Differences of Atrial Fibrillation Cycle Length in Patientswith Paroxysmal and Persistent Atrial Fibrillation at the Beginning

of Atrial Fibrillation

Paroxysmal AF Persistent AFLeft Atrial Region (n = 15) (n = 4)

Lateral wall 194 ± 41∗ 179 ± 25Anterolateral 191 ± 43 180 ± 30Superolateral 191 ± 41 177 ± 24Posterolateral 199 ± 43 177 ± 24Mid-lateral 190 ± 41 177 ± 23

Septal wall 185 ± 38 175 ± 24Anteroseptal 185 ± 38 175 ± 24Superoseptal 188 ± 40 176 ± 24Posteroseptal 184 ± 40 174 ± 25Mid-septal 186 ± 39 176 ± 25

Mid-anterior wall 189 ± 42 177 ± 23Mid-superior wall 190 ± 40 176 ± 24Mid-posterior wall 198 ± 44 176 ± 24

All values are given in milliseconds.∗P < 0.05 for all comparisons between atrial fibrillation (AF) cycle lengthsin paroxysmal and persistent AF groups.

TABLE 2.

Occurrence of Disorganized Atrial Activity after Atrial FibrillationOnset in Different Regions of the Left Atrium with Respect to

Trigger Localization

Pulmonary ExtrapulmonaryLeft Atrial Region Vein Onset Vein Onset

Anterior wall 6.9 ± 3.0∗ 7.2 ± 2.8Superior wall 7.4 ± 2.9 6.8 ± 2.7Posterior wall 5.8 ± 3.1 4.7 ± 2.9Lateral wall 7.3 ± 3.1 6.4 ± 2.7Septal wall 6.8 ± 2.7 5.4 ± 3.0

Values are measured in number of cycles after onset of atrial fibrillation.∗P < 0.05 for all comparisons between pulmonary vein onset and extrapul-monary vein onset.

Discussion

Understanding the underlying mechanisms of AF initia-tion and maintenance is one of the challenges of contempo-rary electrophysiology. This study focused on the analysis ofatrial ectopic activity, waveform behavior, and activation se-quences at the onset of AF using a high-resolution noncontactmapping system.

Regional Differences in the LA in Paroxysmal andPersistent AF

Our study showed that there are substantial regional dif-ferences in AF cycle lengths and the degree of organizationof electrical activity immediately after AF onset. The ob-served septal-to-lateral gradient of the distribution of AF cy-cle lengths at AF onset reflects these regional differences andcoincides with results of other mapping studies.10-13 MeanAF cycle lengths (fibrillation intervals) of the rhythm at AFonset were 11% shorter in patients with persistent AF thanin patients with paroxysmal AF. Parallel to the decrease inAF cycle lengths, we observed faster generation of disorga-nized activity and earlier development of multiple waveformsin the group of patients with persistent AF. These observa-tions coincide with an experimental study by Sih et al.,14

who reported a nearly 25% increase in the disorganizationobserved in canine LA in chronic AF compared with acuteAF. The decrease in cycle length and the increase in degreeof disorganization in patients with persistent AF could be ex-plained by remodeling induced by long-standing fibrillatoryactivation.15 These findings may explain to a certain extentthe difficulties in restoring sinus rhythm and the early recur-rences of AF in the patients with persistent AF comparedwith the patients with paroxysmal AF.

Characterization of AF Initiation and Repetitive Activity

The role of APCs in the initiation of AF has already beendetermined.16-18 However, our data showed that the capabil-ity of APCs to induce AF depends on several factors, in-cluding the coupling interval and localization of the ectopicfoci. Foci located inside the PVs initiated AF even thoughthe APCs had longer coupling intervals than the foci locatedin other parts of the atria. This leads to the speculation thatAPCs entering the LA through PV-LA muscle connectionsare more prone to induce AF compared with APCs that enterthe LA via other propagation routes.

180 Journal of Cardiovascular Electrophysiology Vol. 14, No. 2, February 2003

Figure 4. Initiation of atrial fibrillation from a focus located inthe left lower pulmonary vein (LLPV). The upper panel shows theisopotential map [posterior aspect of the left atrium (LA)] with theearliest activity just emerging from the LLPV ostium. The lowerpanel shows surface ECG lead II and three virtual electrograms(virtual 6 to virtual 8) that were recorded from the respectivepositions in the LA map (posterior aspect of the LA). Again oneAPC with a cycle length of 383 ms did not induce AF. The secondAPC with a cycle length of 336 ms triggered the onset of AF.Regular repetitive activity was recorded from that region. LUPV= left upper pulmonary vein; RLPV = right lower pulmonaryvein; RUPV = right upper pulmonary vein.

The rapidity and the time of disorganization of the wave-fronts at AF onset depended on the localization of triggeringactivity. The active triggers (focally driven activity) after ini-tiation of AF may serve as a mechanism of AF maintenanceearly in the initiation process and could explain relatively reg-ular rhythms observed at the early stages of AF onset. Theserelatively regular rhythms depend on repetitive firing focalactivity. In AF episodes in which no repetitive firing couldbe observed at onset of AF, disorganization of waveformswas observed earlier than in AF episodes initially driven byrepetitive focal activity.

Figure 5. Generation of multiple wavelet activity at the onsetof atrial fibrillation. Isopotential map of the posterior aspectof the LA is shown. A: Initial activation wavefront starts nearthe ostial regions of the left pulmonary veins (marked by 1).B: The wavefront is divided into two wavefronts: one wavefront(arrow 2) propagates downward through the lateral wall andreaches the inferoposterior region. The second wavefront (arrow3) separates and moves through the posterior wall in a lateral-to-medial direction. C: Wavefronts 2 and 3 continue to propagatein inferoseptal and posterosuperior directions, respectively. D:From wavefront 2, a fourth wavefront (marked by arrow 4) isobserved. LIPV = left inferior pulmonary vein; LSPV = leftsuperior pulmonary vein; RIPV = right inferior pulmonary vein;RSPV = right superior pulmonary vein.

Implications of the Study

Our study has implications concerning the mechanisms re-sponsible for maintenance of AF, especially during the earlystages after the triggering APC. Using noncontact mapping,we were able to detect the continuation of repetitive firingfrom foci for a certain amount of time after the triggeringAPC, a finding that underscores the importance of repeti-tive firing for maintenance of AF immediately after onset.These findings are in accordance with studies by Haissaguerreet al.1 and Jais et al.,19 who demonstrated the importance of

Weber et al. Characterization of AF with Noncontact Mapping 181

repetitive firing at the beginning of paroxysmal AF. How-ever, it appears that in some patients (20% in our study), othermechanisms enable maintenance of AF at the beginning of ar-rhythmia. Since disorganized and complex activity emergedearlier in this group of patients than in the patients with focallydriven AF episodes, we hypothesize that multiple waveletsgenerated early after the triggering APC maintain the AFprocess. The fact that the posterior wall of the LA showedmultiple wavefronts earlier than the other LA areas could beexplained by the complex structure of the region.20 Whetherthese patients with early generation of multiple wavefrontsrepresent a group of patients with more advanced atrial dis-ease compared with patients with focally driven AF remainsto be seen.

Conclusion

Data from this study suggest that in the majority of AFepisodes (80%), repetitive firing from the triggering foci mayplay an important role in maintaining AF immediately afterarrhythmia onset. In 20% of episodes, AF during the earlystages seems to be maintained by other mechanisms. Thepower of APC to induce AF depends on the coupling in-terval and focus localization. The posterior wall of the LAshows the earliest disorganization of wavefronts at the onsetof AF.

References

1. Haissaguerre M, Jais P, Shah DC, Takahashi A, Hocini M, Quiniou G,Garrigue S, Le Mouroux A, Le Metayer P, Clementy J: Spontaneous ini-tiation of atrial fibrillation by ectopic beats originating in the pulmonaryveins. N Engl J Med 1998;339:659-666.

2. Kumagai K, Khrestian C, Waldo AL: Simultaneous multisite map-ping studies during induced atrial fibrillation in the sterile pericardi-tis model: Insights into the mechanism of its maintenance. Circulation1997;95:511-521.

3. Skanes AC, Mandapati R, Berenfeld O, Davidenko JM, Jalife J: Spa-tiotemporal periodicity during atrial fibrillation in the isolated sheepheart. Circulation 1998;98:1236-1248.

4. Jais P, Haissaguerre M, Shah DC, Chouairi S, Clementy J: Regional dis-parities of endocardial atrial activation in paroxysmal atrial fibrillation.Pacing Clin Electrophysiol 1996;19(Pt II): 1998-2003.

5. Lin JL, Lai LP, Tseng YZ, Lien WP, Huang SKS: Global distribution ofatrial ectopic foci triggering recurrence of atrial tachyarrhythmia after

electrical cardioversion of long-standing atrial fibrillation: A bi-atrialbasket mapping study. J Am Coll Cardiol 2001;37:904-910.

6. Schilling RJ, Kadish AH, Peters NS, Goldberger J, Davies DJ: Endo-cardial mapping of atrial fibrillation in the human right atrium using anon-contact catheter. Eur Heart J 2000;21:550-564.

7. Schilling RJ, Peters NS, Davies DW: Feasibility of a noncontact catheterfor endocardial mapping of human ventricular tachycardia. Circulation1999;99:2543-2552.

8. Schilling RJ, Peters NS, Davies DW: Simultaneous endocardial map-ping in the human left ventricle using a noncontact catheter: Comparisonof contact and reconstructed electrograms during sinus rhythm. Circu-lation 1998;98:887-898.

9. Karch MR, Schmieder S, Ndrepepa G, Schneider MA, Zrenner B,Schmitt C: Internal atrial defibrillation during electrophysiological stud-ies and focal atrial fibrillation ablation procedures. Pacing Clin Electro-physiol 2001;24:1464-1469.

10. Konings KT, Kirchhof CJ, Smeets, Wellens HJ, Penn OC, Allessie MA:High-density mapping of electrically induced atrial fibrillation in hu-mans. Circulation 1994;89:1665-1680.

11. Markides V, Schilling RJ, Chow AWC, Peters NS, Davies DW: Identi-fication of the mechanisms of initiation of atrial fibrillation using non-contact mapping. (Abstract) Circulation 2001;104:II-158.

12. Saksena S, Prakash A, Krol RB, Shankar A: Regional endocardial map-ping of spontaneous and induced atrial fibrillation in patients with heartdisease and refractory atrial fibrillation. Am J Cardiol 1999;84:880-889.

13. Ndrepepa G, Karch MR, Schneider MA, Weyerbrock S, Schreieck J,Deisenhofer I, Zrenner B, Schomig A, Schmitt C: Characterization ofparoxysmal and persistent atrial fibrillation in the human left atriumduring initiation and sustained episodes. J Cardiovasc Electrophysiol2002;13:525-532.

14. Sih HJ, Zipes DP, Berbari EJ, Adams DE, Olgin JE: Differences inorganization between acute and chronic atrial fibrillation in dogs. J AmColl Cardiol 2000;36:924-931.

15. Wijffels MC, Kirchhof CJ, Dorland R, Allessie MA: Atrial fibrillationbegets atrial fibrillation: A study in awake chronically instrumentedgoats. Circulation 1995;92:1954-1968.

16. Killip T, Gault J: Mode of onset of atrial fibrillation in man. Am HeartJ 1965;70:172.

17. Capucci A, Santarelli A, Boriani G, Magnani B: Atrial premature beatscoupling interval determines lone paroxysmal atrial fibrillation onset.Int J Cardiol 1992;36:87-93.

18. Kolb C, Nurnberger S, Ndrepepa G, Zrenner B, Schomig A, SchmittC: Modes of initiation of paroxysmal atrial fibrillation from analysisof spontaneously occurring episodes using a 12-lead Holter monitoringsystem. Am J Cardiol 2001;88:853-857.

19. Jais P, Haissaguerre M, Shah DC, Chouairi S, Gencel L, Hocini M,Clementy J: A focal source of atrial fibrillation treated by discrete ra-diofrequency ablation. Circulation 1997;95:572-576.

20. Nathan H, Eliakim M: The junction between the left atrium and thepulmonary veins: An anatomic study of human hearts. Circulation1966;34:412-421.