Method for Simultaneous Epicardial and Endocardial Mappingof In Vivo Canine Heart: Application to Atrial Conduction

Properties and Arrhythmia MechanismsKATAYOUN DERAKHCHAN, PH.D.,1,7 DANSHI LI, M.D., PH.D.,1

MARC COURTEMANCHE, PH.D.,1 BENJAMIN SMITH,6 JUDITH BROUILLETTE,6

PIERRE L. PAGE, M.D.,1,3,5 and STANLEY NATTEL, M.D.1,2,4,7

From the 1Research Center and the Departments of 2Medicine and 3Surgery, Montreal Heart Institute; the Departments of 4Medicineand 5Surgery, University of Montreal; and the Departments of 6Physiology and 7Pharmacology and Therapeutics,

McGill University, Montreal, Quebec, Canada

Endocardial/Epicardial Mapping of AF. Introduction: It has been suggested that the three-dimensional structure of the atria may be crucial in arrhythmogenesis; however, previous in vivo atrialactivation mapping studies have been limited to either endocardial or epicardial approaches.

Methods and Results: To investigate the role of endocardial and epicardial structures and theirinteraction in atrial conduction and arrhythmias, we used � ve epicardial plaques and two intra-atrialballoon arrays to record a total of 368 unipolar electrograms from the entire epicardial and endocardialsurface of both atria. During regular 1:1 pacing from the right atrial appendage, right atrial endocardialactivation spread considerably faster than epicardial (total activation time 45 6 12 msec vs 60 6 19 msec,respectively [mean 6 SD]; P < 0.05), pointing to preferential conduction over structures like the cristaterminalis and pectinate muscles. No such differences were noted in the left atrium. Transseptal spreadoccurred via discrete anterior and posterior pathways, causing separate breakthroughs in anterior andposterior atrial regions, respectively. Dissociation between septal pathways played a role in reentry duringvagal atrial � brillation. In 2 of 4 dogs with atrial � brillation associated with congestive heart failure, singlemacroreentrant circuits involving endocardial and epicardial components were revealed during thearrhythmia.

Conclusion: We conclude that activation mapping using simultaneous recording from both epicardialand endocardial surfaces provides potentially important insights into the mechanisms of atrial conductionand arrhythmogenesis. (J Cardiovasc Electrophysiol, Vol. 12, pp. 548-555, May 2001)

mapping, cardiac arrhythmias, heart disease, reentrant mechanisms, cardiac conduction

Introduction

Atrial � brillation (AF) is the most common sustainedcardiac arrhythmia, and its treatment presently is unsatis-factory.1 To improve therapeutic approaches to AF, it isessential to have a better understanding of basic underlyingmechanisms. Although there has been considerable im-provement of our understanding of AF mechanisms over thelast 5 years, our knowledge still is limited, partly because ofinadequate mapping tools. One of the most fundamentalchallenges in understanding AF is unraveling the anatomic-functional aspects of the underlying reentrant mechanism.Previous in vivo mapping studies used either endocardial orepicardial mapping arrays, essentially treating the atria astwo-dimensional structures.2 -9 However, the atria present acomplex geometry with prominent endocardial structures

(particularly right atrial [RA] pectinate muscles and cristaterminalis), obstacles (such as venae cavae and pulmonaryveins), and limited interatrial connections (Bachmann’ sbundle and atrial septum). This complex three-dimensionalarchitecture is believed to be important in atrial conductionand arrhythmias.1 0 -1 2

To de� ne better the role of atrial structure in electricalpropagation during normal atrial activation and during AF,we developed a new technique for simultaneous recordingof endocardial and epicardial activation of the in situ dogheart. In conjunction with custom-made mapping and im-aging software, the results allow for simultaneous visual-ization of atrial endocardial and epicardial activation invivo. Using these methods, we aimed to address the follow-ing speci� c questions: (1) Are there preferential pathwaysof atrial activation? (2) How do impulses propagate betweenatria? (3) Can simultaneous endocardial and epicardial map-ping provide insights into mechanisms of reentry duringexperimental AF?

Methods

Experimental Models and General Methods

Twelve mongrel dogs of either sex weighing 25 to 30 kgwere studied. Eight acute dogs were used to study theactivation pattern during sinus rhythm, during pacing atvarious atrial sites, and during AF in the presence of vagal

This study was supported by funds from the Medical Research Council ofCanada, the Quebec Heart Foundation, and the Mathematics of InformationTechnology and Complex Systems Network of Centers of Excellence. Drs.Derakhchan and Li are Heart and Stroke Foundation of Canada ResearchFellows.

Address for correspondence: Stanley Nattel, M.D., Research Center, Mon-treal Heart Institute, 5000 Belanger Street East, Montreal, Quebec H1T1C8, Canada. Fax: 514-376-1355; E-mail: nattel@icm.umontreal.ca

Manuscript received 3 July 2000; Accepted for publication 1 February2001.

548 Reprinted with permission fromJOURNAL OF CARDIOVASCULAR ELECTROPHYSIOLOGY, Volume 12, No. 5, May 2001Copyright ©2001 by Futura Publishing Company, Inc., Armonk, NY 10504-0418

stimulation. Another four conditioned dogs were used tostudy sustained AF induced in a substrate associated withcongestive heart failure (CHF), as previously described.1 3

For production of heart failure, dogs were anesthetizedwith sodium pentobarbital 30 mg/kg intravenously (IV),followed by IV boluses of 4 mg/kg as needed. Arti� cialrespiration was maintained via an endotracheal tube con-nected to a Harvard-type mechanical ventilator. Under ster-ile technique, a unipolar endocardial tined ventricular pac-ing lead (CapSure™ SP, model 4023; Medtronic Inc.,Minneapolis, MN, USA) was inserted into the right ventric-ular apex via the right jugular vein, and the distal end of thelead was � xed in the right ventricular apex under � uoro-scopic guidance at a site with a low pacing threshold. Initialventricular capture was veri� ed using an external demandpacemaker (GBM 5880, Medtronic Inc.). The proximal endof the pacing lead was connected to a custom-modi� edVVIP pacemaker (model 8084, Medtronic Inc.), which thenwas implanted into a subcutaneous pocket in the lateralneck. To prevent VA conduction, AV nodal conduction waseliminated by radiofrequency ablation. Twenty-four hourswere allowed for lead stabilization, then the pacemaker wasprogrammed to capture the right ventricle at 240 beats/minusing 0.42-msec square-wave pulses at twice threshold cur-rent. The right ventricle was continuously stimulated at thisrate for a period of 3 weeks. The pacing rate then wasdecreased to 220 beats/min, and the ventricle was paced atthis rate for another 2 weeks. The surface ECG was checkedweekly to ensure constant pacing. The presence of heartfailure was established by a constellation of signs (lethargy,loss of appetite, dyspnea, peripheral edema, and ascites).

To study vagotonic AF, methods previously described indetail were applied.4 In brief, bipolar Te� on-coated (exceptfor the distal 1 cm) stainless steel stimulating electrodeswere inserted within the sheaths of each cervical vagalnerve parallel to the axis of vagal � bers, with an approxi-mately 2-cm electrode length within the sheath. Stimulationwas applied with 5-V, 0.1-second stimuli at a frequency twothirds that required to produce asystole for . 3 seconds.With this level of vagal stimulation, a brief burst of 10-Hzatrial pacing is suf� cient to induce AF that remains sus-tained as long as vagal stimulation is continued.

On study days, dogs were anesthetized with morphine 2mg/kg subcutaneously and a-chloralose 120 mg/kg IV bo-lus, followed by a continuous infusion of 29.25 mg/kg/hour,14 and ventilated with room air supplemented withoxygen. Body temperature was maintained at 37°C with ahomeothermic temperature control system. The left femoralartery and both femoral veins were cannulated and keptpatent with heparinized 0.9% saline solution. The left fem-oral artery cannula was used to monitor arterial pressure.The left femoral vein cannula was used to infuse a-chlora-lose. The right femoral vein cannula was perfused with 5%glucose at approximately 1 mL/min. A bilateral thoracot-omy was performed between the fourth and � fth intercostalspaces, and a pericardial cradle was created. Two bipolarTe� on-coated stainless steel electrodes were hooked intothe right atrial appendage (RAA) and left atrial appendage(LAA) for recording and stimulation. A programmablestimulator (Digital Cardiovascular Instruments) was used todeliver 2-msec pulses at twice threshold current. Arterialpressure, four standard surface ECG leads, two atrial elec-trograms, and stimulus artifacts were monitored and re-

corded. In heart failure dogs, the surface ECG was recordedbefore surgery at the terminal electrophysiologic study tocon� rm 1:1 ventricular pacing at 220 beats/min. The im-planted pacemaker then was deactivated. In vagotonic prep-arations, the vagus nerves were isolated in the neck, doublyligated and divided, and bipolar electrodes were inserted ineach nerve. An external pacemaker (GBM 5880, MedtronicInc.) was used to stimulate the ventricles at 80/min when(particularly during vagal AF) the ventricular rate becameexcessively slow.

Epicardial and Endocardial Mapping

Two mapping systems were used simultaneously torecord signals from a total of 368 sites: 240 distributed overthe epicardial surfaces of both atria and 128 electrodesdivided evenly between two balloons inserted via smallappendiceal incisions. Figure 1 shows the layout of theelectrode arrays and unipolar electrodes used for mapping.For epicardial mapping, � ve thin silicon sheets containing240 unipolar electrodes for stimulation and/or recordingwere sewn into position on the atrial epicardial surfaces.Three sheets were sewn to the lateral surfaces of the atrialappendages and to the free walls. One sheet was placedunder the atrial appendages and on Bachmann’s bundle. The� fth sheet was placed between the pulmonary veins. Theinterelectrode distance between each unipolar electrode andneighboring sites was between 3.1 and 4.0 mm for thearrays on the RA and LA free walls, RAA and LAA, andBachmann’s bundle, and 6 mm for the array between thepulmonary veins.

Endocardial activation was recorded by using electrodesattached to in� atable balloons, inserted within the atria viasmall appendiceal incisions ( ; 8 mm). Each balloon wasoval shaped, constructed with nylon, and had 64 unipolarelectrodes uniformly distributed on the surface. When de-� ated, the right balloon measured about 1 cm wide and 3 cmlong; the left balloon measured 1 3 2.4 cm. When in� ated,the balloons were expanded to appose closely to the endo-cardial surface of the atria. Good contact was ensured byverifying electrogram size and morphology and by gentlypalpating the atrial surfaces. The interelectrode distancebetween adjacent unipolar electrodes when the balloonswere in� ated was 8 to 9 mm for the LA balloon and 6 to 9mm for the RA balloon. The balloons were secured by apursestring suture around the point of insertion, with theinsertion process completed within 5 seconds. Balloon in-sertion did not produce any change in arterial pressure,provoke arrhythmias, or cause detectable cardiac injury. Atthe time of data recording, the balloons were in� ated brie� yto contact the entire endocardial surface of each atrium. Theballoons then were de� ated promptly, so that the recordingof endocardial activation had no detectable effect on cardiachemodynamics.

Unipolar atrial recording contacts (referenced to Wil-son’ s central terminal) were connected to multichannel re-cording systems (Electrophysiological Data Interface, Insti-tut de Genie Biomedical, Ecole Polytechnique de Montreal,Montreal, Canada) and mVAX computers (Digital DataCorp., Maynard, MA, USA). Signals were � ltered (10 to900 Hz) and digitized with 12-bit resolution and a 2-kHzsampling rate. Two mapping systems, each with the capac-

Derakhchan et al. Endocardial/Epicardial Mapping of AF 549

ity to acquire simultaneously up to 256 electrograms, wereused to acquire data. A 1-msec unipolar signal was deliv-ered by the stimulator simultaneously to each mappingsystem, to allow for analysis of endocardial and epicardialactivation times on the same time base. Data were stored ona hard disk, from which selected � les were later retrievedfor analysis using custom-made software. The maximumnegative slopes of electrogram signals were used to de� neactivation time. All computer-selected events were veri� edmanually with an interactive program.

At the termination of each experiment, the animal wassacri� ced and the heart carefully removed from the chest

cavity. The atria were opened from the appendages and theendocardial electrode positions veri� ed and marked in re-lation to anatomic landmarks and to epicardial electrodepositions.

Three-Dimensional Visualization of Activation Patterns

To relate epicardial and endocardial activation to eachother and to atrial anatomy, a three-dimensional visualiza-tion system was developed. Activation times for each elec-trode position were used to produce a three-dimensionalanimated (1-msec frames) display of activation with the use

Figure 1. Three-dimensional visualization of elec-trode arrays showing a superior view of both atriawith the appendages at the top. Unipolar electrodessites are indicated by white dots. A total of 240 elec-trodes were distributed over the epicardial surfaces ofboth atria, with the RA epicardial patch containing105 sites, the Bachmann’s bundle patch 48 sites, thepatch between the pulmonary veins 8 sites, the LAappendage 34 sites, and the LA free-wall patch 45sites. Each endocardial balloon array is representedas its real oval shape and contains 64 electrodes. Theballoon arrays are shown here as a wireframe dis-play; otherwise the electrode sites behind the plane ofvision would be obscured. For easier readability, onlythe portion of the arrays facing the observer areshown in other � gures; however, all sites were eval-uated for all mapping studies. Endo 5 endocardium;epi 5 epicardium.

Figure 2. Activation following epicar-dial RA (top) and LA (bottom) append-age stimulation. The stimulation site isshown by an asterisk. Initial RA acti-vation is at the epicardium (A). Activa-tion spread rapidly to the endocardiumand began to move down the endocar-dial surface within 6 ms (B). The im-pulse had activated over half of theendocardial surface and only about aquarter of the epicardium (C) within 15ms of initial epicardial activation. En-docardial activation approaches theapex 7 ms later (D), when less than halfof the epicardium has been activated.The LA also is activated initially in theepicardium of the appendage (E). Theendocardium is activated about 13 mslater (F), and then activation proceedsmore or less in parallel in endocar-dium and epicardium (G, H). Ant 5anterior; app 5 appendage; Post 5posterior.

550 Journal of Cardiovascular Electrophysiology Vol. 12, No. 5, May 2001

of custom-made software in Linux (Red Hat Software Inc.,Durham, NC, USA) running on a Pentium workstation. Therecording instruments were modeled to scale with nonuni-form rational B-spline (NURBS) surfaces1 5 to represent theelectrode plaques and balloons (Fig. 1). Balloon and patchelectrode positions are considered as vertices to generate amesh of second-order polynomials. Before activation, eachvertex (electrode) is colored pure blue. Upon activation, theelectrode is assigned pure red. The color between eachelectrode is determined by a second-order polynomial. Re-gions that were activated within the preceding 4 msec aredisplayed in red (Fig. 2). All other surfaces are in blue. Adarker region that often occurs between pure red and blueelectrodes is a result of the Mesa-OpenGL algorithm tryingto blend the colors. The orientation of the atrial surfaces canbe rotated freely in any direction and the endocardial sur-faces can be rendered transparent (wireframe view as in Fig.1) to visualize the endocardial surface behind the surfacefacing the reviewer.

A 2-second window of activation data was acquiredduring sinus rhythm and during 1:1 atrial pacing (cyclelength 250 msec) at the RAA and LAA, the RA and LAatrial free wall, the middle of Bachmann’s bundle (superiorseptum), and the left pulmonary vein region. Endocardialand epicardial activation recordings were obtained after 1minute of continuous pacing. AF was de� ned as a rapid( . 450 beats/min), irregular atrial rhythm with varying atrialelectrogram morphology and frequency, accompanied by acharacteristic surface ECG pattern. AF was considered sus-tained if it persisted for . 30 minutes and required electricalcardioversion for termination. A 10-second window of ac-tivation data was acquired during AF to analyze activationpatterns.

Statistical Analysis

Statistical comparisons between group means were ob-tained by Student’ s t-test. All results are expressed asmean 6 SD. Differences with P , 0.05 were consideredstatistically signi� cant.

Results

After 1:1 stimulation at the RAA, activation began in theepicardium, but as soon as it reached the endocardiumbegan to spread quite rapidly over the endocardial surface.Figure 2 (top) shows activations at selected time-pointsfollowing stimulation at the RAA epicardium. Initial acti-vation is at the lateral epicardial and then endocardial sur-faces of the RAA. Although initial activation is at theepicardium (Fig. 2A), within 6 msec activation has spreadfurther along the endocardial than epicardial surface (Fig.2B). Over the following 9 msec, over half of the endocardialsurface has been activated, compared to , 25% of the epi-cardial surface (Fig. 2C). When RA endocardial activationapproaches the apex, less than half of the epicardial surfacehas been activated (Fig. 2D). In contrast, LA activationspreads much more symmetrically along the endocardialand epicardial surfaces (Fig. 2, bottom). After initial LAAepicardial activation (Fig. 2E), the LA endocardium is ac-tivated as the impulse travels posteriorly in the epicardium(Fig. 2F), with parallel epicardial and endocardial impulsepropagation (Fig. 2G) toward the apex (Fig. 2H). In all eightnormal dogs, total RA endocardial activation time averaged

45 6 12 msec (mean 6 SD), signi� cantly less than for theRA epicardium (60 6 19 msec; P , 0.05). In contrast, whenthe LAA was paced, there was no signi� cant difference inendocardial (67 6 10 msec) versus epicardial (59 6 6 msec)activation time, nor any visually apparent differences inactivation speed. These results are consistent with the notionthat RA cable-like structures (like the crista terminalis andpectinate muscles) result in preferentially rapid conductionpathways on the RA endocardial surface.

Anterior transseptal conduction was studied during stim-ulation at the mid-point of Bachmann’s bundle. Figure 3shows a typical example of activation following mid-Bach-mann’ s bundle stimulation. Initial activation is at the middleof Bachmann’s bundle (Fig. 3A). Within 10 msec, theimpulse breaks through into both atria at the anteromedialseptal endocardial surface (Fig. 3B). The electrical wave-fronts then spread progressively over the endocardial sur-faces (Figs. 3C to 3F). The epicardial surface is � rst acti-vated anteriorly at the atrial appendages (Fig. 3D), withactivation of the posterior epicardium occurring later fromadjacent endocardial surfaces (Figs. 3E and 3F). Final ac-tivation of the epicardium results from wavefronts movinganteriorly and posteriorly in each and colliding in the mid-dle of each epicardial surface (Fig. 3F).

Figures 4 and 5 show the posterior transseptal pathwayupon stimulation of the LA and RA, respectively. Figure 4shows activation after stimulation of the posterior LA in theregion of the pulmonary veins (Fig. 4A). The impulsetravels rapidly to the posterior LA endocardium (Fig. 4B).Soon after LA endocardial septal activation, the middle ofBachmann’s bundle is activated via the septum (Fig. 4C).Initial RA activation is in the posterior endocardium (Fig.4D), at a site quite distinct from the region initially activatedby anterior septal spread (as shown in Fig. 3B).

Figure 5 shows posterior septal spread after stimulationat the posterior epicardial RA. After stimulation, activationspreads rapidly over the posterior epicardial surface (Fig.5A). The impulse then reaches the posterior endocardialsurface (Fig. 5B), from which it travels via the septum to themiddle of Bachmann’s bundle (Fig. 5C) and shortly there-after to the posterior LA endocardial surface (Fig. 5D).

Given the existence of distinct anterior and posteriorseptal pathways, it is theoretically possible for temporaldissociation between the pathways to permit reentry. Inthree dogs studied during vagal AF, reentrant cycles wereobserved that involved dissociation between the anteriorand posterior septal pathways, with each path forming aseparate component of the reentry circuit. One example isshown in Figure 6. Initial activation is at the posteriorendocardial surface of both atria (Fig. 6A). The impulsetravels anteriorly in the RA endocardium (Fig. 6B) butencounters a line of functional block at the base of the RAA(dashed line in Fig. 6B). The impulse then spreads posteri-orly in the RA endocardium and reaches the RA epicardialsurface (Fig. 6C), while spreading anteriorly in the LAtoward the LA junction of Bachmann’s bundle. The impulsethen traverses Bachmann’s bundle (Fig. 6D) to activate theanterior RA (Fig. 6E). The posterior RA then is reactivated,followed by posterior transseptal spread to the posterior LA(Fig. 6F).

We previously demonstrated evidence for macroreentryunderlying AF in dogs with CHF16 ,1 7 ; however, we wereunable to follow reentry through the entire cycle. We hy-

Derakhchan et al. Endocardial/Epicardial Mapping of AF 551

pothesized that the failure to demonstrate reentry may havebeen due to the inability of epicardial mapping alone torecord from essential components of the circuit that werelocated endocardially. In the present study, we were able torecord complete and stable macroreentrant circuits in twodogs with sustained AF initiated in the presence of CHF.Figures 7 and 8 illustrate activation during AF in one suchdog. Figure 7 shows activation maps during one cycle ofAF, and Figure 8 shows corresponding electrogram record-ings. Initial activation is at the anteromedial RAA (Fig. 7A),with the impulse then spreading across Bachmann’s bundle(Fig. 7B) and clockwise around a line of functional block(dashed line) in the RA (Fig. 7B). The impulse arrives in theposterior RA epicardium (Fig. 7C) and crosses the posterior

septal pathway to activate the posterior LA (Fig. 7D). TheRA epicardium is activated in a posterior to anterior fashionover Figures 7E through 7G, with reactivation of the RAendocardium completing reentry in Figures 7G to 7I. Thispattern of reentry was consistent throughout AF in this dog.Note that the LA is activated passively over ; 25 msec(Figs. 7D to 7F), but is quiescent over most of the cycle andappears not to comprise part of the reentrant circuit, whichis shown diagrammatically along with electrogram record-ings in Figure 8. Recordings from individual sites within theapparent reentry circuit are shown to the left, correspondingto locations a through i on the schematic at the right. Site jis just outside the circuit, at the RAA epicardium. Site k isjust outside the circuit in the RA endocardium, and its

Figure 3. Anterior transseptal conduction follow-ing stimulation at the middle of Bachmann’s bun-dle. Initial activation is at the middle of Bach-mann’s bundle (A). Within 10 ms, activation hasspread to the anteromedial surfaces of the endo-cardium of both atria (B). Further endocardialspread (C) is followed by epicardial activation (D).Activation then moves to the posterior surfaces ofthe atria (E,F). BB 5 Bachmann’s bundle. Otherabbreviations as in previous � gures. See text fordetailed discussion.

Figure 4. Posterior septal transmission from LA to RA following stimula-tion of the atrium adjacent to the pulmonary veins. Initial activation is atthe pulmonary vein patch (A), with rapid spread to the LA endocardium(B). The impulse then travels to Bachmann’s bundle via the septum (C),followed by posterior transseptal spread to the posterior RA (D). Note theclear difference between the location of RA endocardial breakthrough in Dand the location of breakthrough following anterior septal transmission inFigure 3B. Abbreviations as in previous � gures.

Figure 5. Posterior septal transmission from RA to LA following stimula-tion of the posterior right atrial epicardium. Initial activation is on theepicardial patch (A), followed by endocardial activation (B). Bachmann’sbundle then is activated via the septum (C), and the posterior LA endo-cardium is activated 6 ms later by transseptal spread (D). Note that the LAbreakthrough in D is in a position generally similar to posterior LAactivation preceding septal transmission to the RA in Figure 4C, and quitedifferent from the LA breakthrough location following anterior septaltransmission in Figure 3B. Abbreviations as in previous � gures.

552 Journal of Cardiovascular Electrophysiology Vol. 12, No. 5, May 2001

irregularity re� ects variability in its pattern of activation inrelation to the primary circuit. Sites l through o are locatedin Bachmann’s bundle and the LA endocardium, respec-tively. A surface ECG recording is shown at the top andindicates the type of � brillo-� utter atrial activity recorded inthis dog (note that QRS complexes are regular because thedog had AV block with ventricular rhythm maintained by aventricular pacemaker).

Discussion

We developed a new approach to record activation si-multaneously on the endocardial and epicardial surfaces ofboth atria in the in situ dog heart. Application of this methodallowed us to obtain insights into atrial conduction and themechanisms underlying atrial reentry in dog models of AF.

New Findings and Potential Signi� cance

Despite the in vitro evidence for potential differencesbetween endocardial and adjacent epicardial activation,1 0 invivo evidence has been lacking. In the present study, wefound that simultaneous endocardial and epicardial mappingprovide potentially new insights into atrial conduction andarrhythmias. First, we obtained evidence suggesting thatspecialized structures in the RA endocardium may permitselectively more rapid conduction on the RA endocardialcompared with epicardial surface. Second, we were able toshow discrete pathways of transseptal conduction, consis-tent with the results of other recent endocardial mappingstudies in vivo.18 ,1 9 Furthermore, we were able to show thatthe existence of these distinct pathways can provide a func-

Figure 6. Dissociation of septal pathways un-derlying a cycle of reentry during vagal AF.Initial activation was in the posterior part of theatria (A) with propagation anteriorly activatingthe RA and LA endocardial surfaces (B,C). Per-haps because of the faster endocardial spreadin the RA, the RA surface is activated morerapidly and the wavefront encounters a line ofrefractoriness in the anterior RA (C). However,the impulse travels up the LA endocardium toactivate Bachmann’s bundle (D) and thenspreads to the RA anterior endocardium (E).Posterior transseptal reactivation of the LA epi-cardium (F) follows, to complete the cycle. Ab-breviations as in previous � gures.

Figure 7. Macroreentry underlying atrial � brillation in a dogwith heart failure. Panels A to I show sequential activation, withthe times indicated at the top corresponding to the running timefrom the onset of activation and to the times indicated for theindividual recordings shown in Figure 8. The bright red zoneswere activated within the 4 ms preceding the running time indi-cated. Abbreviations as in previous � gures. See text fordiscussion.

Derakhchan et al. Endocardial/Epicardial Mapping of AF 553

tional substrate for intra-atrial reentry during AF. Our � nd-ings are consistent with previous studies pointing to animportant role of the septum in AF20 and contribute to apotential mechanistic rationale for procedures aiming tocontrol AF by ablation of septal tissue2 1 or by atrial septalpacing.2 2

Since the early part of the 20th century, one of themechanisms proposed to underlie AF was a single macro-reentrant circuit with irregular conduction to the other partsof the atria.2 3 This idea fell out of fashion, with the pre-dominant conceptual format for AF becoming multiple cir-cuit2 3 or a closely related notion, multiple wavelet,24 reen-try. Recently, there has been renewed interest in the conceptof primary macroreentrant circuits with � brillatory conduc-tion on the basis of a variety of experimental observa-tions.1 6 ,25 In the present study, we found that consistentsingle macroreentrant circuits could be recorded during AFin two dogs with CHF. The surface ECG pattern was one of“� brillo-� utter” or coarse AF, with P wave irregularity dueto varying relative timing of activation at different sitesoutside the primary reentrant circuit. This type of activitycould be much more prevalent in clinical AF than presentlyrealized (particularly early after the onset of arrhythmia),with a component of tachycardia-related remodeling pro-ducing a subsequently more disorganized atrial activationpattern at the time of clinical presentation. Clari� cation ofthe evolution of AF mechanisms over time might haveimportant clinical and therapeutic implications.

Schuessler et al.1 0 were the � rst to analyze simulta-neously electrical activity in the epicardium and endocar-dium of isolated canine RA. Differences in activation timeof adjacent epicardial and endocardial tissues were noted,particularly in regions of greater thickness and pectinatemuscles. Doshi et al.26 performed simultaneous endocardialand epicardial mapping in isolated LA from three dogs.With this approach, they were able to exclude reentrant

activity during isoproterenol-induced tachyarrhythmias andto show a focal origin at the ligament of Marshall. Athill etal.2 7 performed endocardial mapping and epicardial micro-electrode recordings during acetylcholine-induced arrhyth-mias in 11 isolated canine RA, and followed this by com-bined endocardial and epicardial mapping in two RApreparations. With this approach, they were able to showthat cells at the core if the reentrant circuit were unexcited,consistent with the spinal wave model of reentry. Our ap-proach is complementary to this previous work, by applyingsimultaneous epicardial and endocardial in vivo to evaluateimpulse propagation and arrhythmias in the intact heart.

Potential Limitations

The main potential limitation of our approach is thelimited resolution provided by the number of electrodesavailable. Technical advances are needed to reduce furtherthe size of electrodes on the balloon arrays and the size ofthe attached cables, so that more dense electrode arrays canbe applied to the inner surfaces of either atrium. Anotherpotential limitation is the need for an atriotomy to introducethe balloon arrays. The incision needed is relatively small,however, and epicardial maps before and after balloon in-sertion con� rmed the absence of any substantial changes inatrial activation induced by balloon insertion. It would havebeen interesting to include other areas in our mappingapproach, such as the ligament of Marshall and the coronarysinus. Evaluation of the roles of these regions would be anappropriate objective in future work.

Conclusion

Simultaneous mapping of atrial endocardial and epicar-dial activation is feasible in vivo and may provide interest-

Figure 8. Electrograms during a cycleof macroreentrant atrial � brillation in adog with congestive heart failure. Top:Surface ECG recording. Left: Electro-grams recorded from the sites indicatedin the schematic diagram at the right.The vertical dashed lines show the acti-vation window illustrated in Figure 7.Activation times during the window areshown above each electrogram. Sites ato i were in the reentrant circuit. Right:Schematic diagram of the electrode ar-rays showing positions at which theelectrograms at the left were recordedand indicating the reentrant circuit.

554 Journal of Cardiovascular Electrophysiology Vol. 12, No. 5, May 2001

ing insights into fundamental mechanisms of atrial conduc-tion and arrhythmias.

Acknowledgments: The authors thank Chantal Maltais and Chantal St. Cyrfor excellent technical assistance and Annie Laprade for excellent secre-tarial help with the manuscript.

References

1. Nattel S: Newer developments in the management of atrial � brillation.Am Heart J 1995;130:1094-1106.

2. Allessie MA, Lammers WJEP, Bonke FIM, Hollen J: Experimentalevaluation of Moe’ s multiple wavelet hypothesis of atrial � brillation.In Zipes DP, Jalife J, eds: Cardiac Arrhythmias. Grune and Stratton,New York, 1985, pp. 265-276.

3. Cox JL, Canavan TE, Schuessler RB, Cain ME, Lindsay BD, Stone C,Smith PK, Corr PB, Boineau JP: The surgical treatment of atrial� brillation. II. Intraoperative electophysiologic mapping and descrip-tion of the electrophysiologic basis of atrial � utter and atrial � brilla-tion. J Thorac Cardiovasc Surg 1991;101:406-426.

4. Wang Z, Page P, Nattel S: Mechanism of � ecainide’ s antiarrhythmicaction in experimental atrial � brillation. Circ Res 1992;71:271-287.

5. Sih HJ, Berbari EJ, Zipes DP: Epicardial maps of atrial � brillationafter linear ablation lesions. J Cardiovasc Electrophysiol 1997;8:1046-1054.

6. Gaspo R, Bosch RF, Talajic M, Nattel S: Functional mechanismsunderlying tachycardia-induced sustained atrial � brillation in a chronicdog model. Circulation 1997;96:4027-4035.

7. Kumagai K, Khrestian C, Waldo AL: Simultaneous multisite mappingstudies during induced atrial � brillation in the sterile pericarditismodel. Insights into the mechanism of its maintenance. Circulation1997;95:511-521.

8. Skanes AC, Mandapati R, Berenfeld O, Davidenko JM, Jalife J:Spatiotemporal periodicity during atrial � brillation in the isolatedsheep heart. Circulation 1998;98:1236-1248.

9. Kadish A, Hauck J, Pederson B, Beatty G, Gornick C: Mapping ofatrial activation with a noncontact, multielectrode catheter in dogs.Circulation 1999;99:1906-1913.

10. Schuessler RB, Kawamoto T, Hand DE, Mitsuno M, Bromberg BI,Cox JL, Boineau JP: Simultaneous epicardial and endocardial activa-tion sequence mapping in the isolated canine right atrium. Circulation1993;88:250-263.

11. Ikeda T, Yashima M, Uchida T, Hough D, Fishbein MC, Mandel WJ,Chen PS, Karagueuzian HS: Attachment of meandering reentrant wavefronts to anatomic obstacles in the atrium. Role of the obstacle size.Circ Res 1997;81:753-764.

12. Wu TJ, Yashima M, Xie F, Athill CA, Kim YH, Fishbein MC, Qu Z,Gar� nkel A, Weiss JN, Karagueuzian HS, Chen PS: Role of pectinatemuscle bundles in the generation and maintenance of intra-atrial re-

entry: Potential implications for the mechanism of conversion betweenatrial � brillation and atrial � utter. Circ Res 1998;83:448-462.

13. Li D, Fareh S, Leung TK, Nattel S: Promotion of atrial � brillation byheart failure in dogs: Atrial remodeling of a different sort. Circulation1999;100:87-95.

14. St.-Georges D, Matthews C, Nattel S: Continuous maintenance infu-sion technique for stable anesthesia in the dog, using a-chloralose.Contemp Topics Lab Anim Sci 1997;36:62-65.

15. Woo M, Neider J, Davis T, Open GL: Evaluators and NURBS.Programming Guide. Second Edition. The Of� cial Guide to Learning,Version II. Addison-Wesley Developers Press, Boston, 1997, pp. 437-468.

16. Nattel S, Li D, Yue L: Basic mechanisms of atrial � brillation-very newinsights into very old ideas. Annu Rev Physiol 2000;62:51-77.

17. Li D, Benardeau A, Nattel S: Contrasting ef� cacy of dofetilide indiffering experimental models of atrial � brillation. Circulation 2000;102:104-112.

18. Roithinger FX, Cheng J, SippensGroenewegen A, Lee RJ, Saxon LA,Scheinman MM, Lesh MD: Use of electroanatomic mapping to delin-eate transseptal atrial conduction in humans. Circulation 1999;100:1791-1797.

19. Sun H, Velipasaoglu EO, Wu DE, Kopelen HA, Zoghbi WA, SpencerWH III, Khoury DS: Simultaneous multisite mapping of the right andthe left atrial septum in the canine intact beating heart. Circulation1999;100:312-319.

20. Kumagai K, Khrestian C, Waldo AL: Simultaneous multisite mappingstudies during induced atrial � brillation in the sterile pericarditismodel. Insights into the mechanism of its maintenance. Circulation1997;95:511-521.

21. Tondo C, Scherlag BJ, Otomo K, Antz M, Patterson E, Arruda M,Jackman WM, Lazzara R: Critical atrial site for ablation of pacing-induced atrial � brillation in the normal dog heart. J Cardiovasc Elec-trophysiol 1997;8:1255-1265.

22. Padeletti L, Porciani MC, Michelucci A, Colella A, Ticci P, Vena S,Costoli A, Ciapetti C, Pieragnoli P, Gensini GF: Interatrial septumpacing: A new approach to prevent recurrent atrial � brillation. J IntervCard Electrophysiol 1999;3:35-43.

23. Garrey WE: Auricular � brillation. Physiol Rev 1924;4:215-250.24. Moe GK: On the multiple wavelet hypothesis of atrial � brillation.

Arch Int Pharmacodyn Ther 1962;140:183-188.25. Jalife J, Berenfeld O, Skanes A, Mandapati R: Mechanisms of atrial

� brillation: Mother rotors or multiple daughter wavelets, or both?J Cardiovasc Electrophysiol 1998;9(Suppl 8):S2-S12.

26. Doshi RN, Wu T-J, Yashima M, Kim Y-H, Ong JJC, Cao J-M, HwangC, Yashar P, Fishbein MC, Karagueuzian HS, Chen P-S: Relationbetween ligament of Marshall and adrenergic atrial tachyarrhythmia.Circulation 1999;100:876-883.

27. Athill CA, Ikeda T, Kim YH, Wu TJ, Fishbein MC, Karagueuzian HS,Chen PS: Transmembrane potential properties at the core of functionalreentrant wave fronts in isolated canine right atria. Circulation 1998;98:1556-1567.

Derakhchan et al. Endocardial/Epicardial Mapping of AF 555