Cardiac electrical dyssynchrony is accurately detected by noninvasive electrocardiographic imaging

Accepted Manuscript Cardiac Electrical Dyssynchrony is Accurately Detected by Noninvasive Electrocardiographic Imaging Laura R. Bear, PhD, Peter R. Huntjens, MSc, Richard Walton, PhD, Olivier Bernus, PhD, Ruben Coronel, PhD, MD, Rémi Dubois, PhD PII:

S1547-5271(18)30135-8

DOI:

10.1016/j.hrthm.2018.02.024

Reference:

HRTHM 7493

To appear in:

Heart Rhythm

Received Date: 6 September 2017

Please cite this article as: Bear LR, Huntjens PR, Walton R, Bernus O, Coronel R, Dubois R, Cardiac Electrical Dyssynchrony is Accurately Detected by Noninvasive Electrocardiographic Imaging, Heart Rhythm (2018), doi: 10.1016/j.hrthm.2018.02.024. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT

Cardiac Electrical Dyssynchrony is Accurately Detected by Noninvasive

2

Electrocardiographic Imaging

3

*Laura R Bear PhD1,2,3, Peter R Huntjens MSc1,2,3,4, Richard Walton PhD1,2,3, Olivier Bernus PhD1,2,3,

4

Ruben Coronel PhD, MD1,5, Rémi Dubois PhD1,2,3

RI PT

1

1

Electrophysiology and Heart Modelling Institute (IHU-LIRYC), Fondation Bordeaux Université, France;

6

2

Université de Bordeaux, Centre de Recherche Cardio-Thoracique de Bordeaux U1045, France;

7

3

Inserm, U1045, Centre de Recherche Cardio-Thoracique de Bordeaux, France;

8

4

CARIM School for Cardiovascular Diseases, Maastricht University Medical Center, The Netherlands

9

5

Department of Experimental Cardiology, Academic Medical Center, The Netherlands

M AN U

SC

5

Short Title: ECGi Detection of Electrical Dyssynchrony

11

Word count: 4989/5000

12

The authors declare that there is no conflict of interest

14

EP

13

TE D

10

*Dr Laura Bear

16

IHU-LIRYC, Avenue du Haut Lévêque, 33600 Pessac, France

17

[email protected]

18

+33(0) 535 38 1967

AC C

15

19 1

ACCEPTED MANUSCRIPT

2 3 4 5 6

Abstract Background Poor identification of electrical dyssynchrony is postulated to be a major factor contributing to the low success rate for cardiac resynchronization therapy (CRT). Objective To evaluate body surface mapping and electrocardiographic imaging (ECGi) to detect electrical

RI PT

1

dyssynchrony noninvasively.

Methods Langendorff-perfused pig hearts (n=11) were suspended in a human torso-shaped tank, with LBBB induced through ablation. Recordings were taken simultaneously from a 108-electrode epicardial sock and 128

8

electrodes embedded in the tank surface during sinus rhythm and ventricular pacing. CT provided electrode and heart

9

positions in the tank. Epicardial unipolar electrograms were reconstructed from torso potentials using ECGi.

SC

7

Dyssynchrony markers from torso potentials (e.g. QRS-duration) or ECGi (total activation time (TAT), interventricular

11

delay (D-LR) and intraventricular markers) were correlated with those recorded from the sock.

12

M AN U

10

Results LBBB was induced (n=8) and sock-derived activation maps demonstrated interventricular dyssynchrony (D-LR and TAT) in all cases (p<0.05), and intraventricular dyssynchrony for complete LBBB (p<0.05) compared to

14

normal. Only D-LR returned to normal with biventricular pacing (p=0.1). Torso markers increased with large degrees of

15

dyssynchrony, and no reduction was seen during biventricular pacing (p>0.05). Although ECGi markers were

16

significantly lower than recorded (p<0.05) there was a significant strong linear relationship between ECGi and recorded

17

values. ECGi correctly diagnosed electrical dyssynchrony, and interventricular resynchronization in all cases. The latest

18

site of activation was identified to 9.1±0.6 mm by ECGi.

21

EP

20

Conclusions ECGi reliably and accurately detects electrical dyssynchrony, resynchronization by biventricular pacing, and the site of latest activation, providing more information than body surface potentials.

AC C

19

TE D

13

Keywords: Non-invasive Electrocardiographic Imaging, Inverse Problem, Electrocardiography, CRT, LBBB

2

ACCEPTED MANUSCRIPT

1 2

1.

Introduction Cardiac Resynchronization Therapy (CRT) is as an established treatment in heart failure patients with

severely impaired ejection fraction and conduction abnormalities, significantly reducing mortality and

4

morbidity1,2. The recommended selection criteria are currently based on the 12-lead ECG: prolonged QRS-

5

duration and/or left bundle branch block (LBBB) morphology3. However, approximately one third of patients

6

show no response to CRT2 likely due to insufficient electrical dyssynchrony prior to CRT, or persistent

7

electrical dyssynchrony afterwards4. This points to an inadequate ability to detect and quantify electrical

8

dyssynchrony using QRS-duration and morphology.

SC

RI PT

3

Noninvasive electrocardiographic imaging (ECGi) has been developed to provide high-resolution imaging

10

of epicardial activation5. ECGi has been used previously to characterize conduction abnormalities in patients

11

amenable to CRT, and to optimize biventricular (BiV) pacing4,6–11. From these studies, intraventricular

12

electrical dyssynchrony, defined as inhomogeneous left ventricular (LV) activation, and interventricular

13

electrical dyssynchrony, defined as activation delay between the LV and right ventricular (RV), were

14

considered predictors for CRT response. Markers have been developed to quantify intraventricular and

15

interventricular electrical dyssynchrony and initial results suggest that they are better predictors for than

16

traditionally used QRS-duration and morphology6–11.

TE D

ECGi has been assumed to provide accurate detection of electrical dyssynchrony based on previous

EP

17

M AN U

9

validation studies in healthy hearts during pacing12,13. However, the reconstruction of activation maps in the

19

presence of electrical dyssynchrony, and the accuracy of electrical dyssynchrony markers have not been

20

evaluated experimentally. Furthermore, it is not clear whether the level of precision provided by ECGi is

21

required, or whether electrical dyssynchrony can be accurately detected directly from body surface

22

potentials14,15.

23 24

AC C

18

The objective of this study was to evaluate the sensitivity of body surface mapping or ECGi to detect electrical dyssynchrony of various degrees. 3

ACCEPTED MANUSCRIPT

1

2.

2

Detailed methods are available in the online supplement.

3

2.1.

Methods

RI PT

Ex-vivo Experimental Data

Hearts were excised from pigs (n=11; 30–40 kg) approved by Directive 2010/63/EU of the European

5

Parliament on the protection of animals used for scientific purposes and the local ethical committee. Hearts

6

were perfused in Langendorff mode with a 1:9 mixture of blood and Tyrode’s solution, oxygenated with

7

95%/5% O2/CO2(pH 7.4, 37°C). An epicardial electrode sock (108 electrodes) was attached to the ventricles

8

and bipolar pacing leads to the RV apex and postero-lateral LV epicardial free wall. An ablation catheter tip

9

was placed in the LV and stitched over the bundle branch in such a way to allow some manoeuvrability.

M AN U

SC

4

After instrumentation, perfusion was changed to 100% Tyrode’s solution and the heart was transferred to a

11

human-shaped torso tank (Figure 1a). LBBB was induced by 2-5 local radiofrequency ablations (25–30W; 30-

12

60 s; Figure 1b). Tank (128 electrodes) and sock potentials were recorded simultaneously (BioSemi, the

13

Netherlands) during sinus rhythm (SR) following ablation, and during LV, RV and BiV pacing in VOO-mode.

14

Afterwards, a CT (Artis, Siemens) was used to obtain the position of the epicardium and electrodes with

15

respect to the tank (Figure 1c).

16

2.2.

Definitions and Markers for Electrical Dyssynchrony

EP

17

TE D

10

LBBB was defined when SR recorded activation maps showed: 1) epicardial breakthrough on the RV with rapid RV activation, 2) no LV breakthrough and 3) late-activation of the LV from the septum. These

19

activation patterns are consistent with previous LBBB epicardial mapping studies in humans7,16,17. The degree

20

of LBBB was considered mild or incomplete if QRS-duration increase was <40%.

21

AC C

18

Epicardial activation times (AT) were derived from recorded electrograms at the moment of the

22

minimum dV/dt and for ECGi signals by fitting a global activation field to activation delays between

23

electrograms18. Time to RS-downslope was defined as the time from QRS-onset to the minimum dV/dt15. 4

ACCEPTED MANUSCRIPT

1

Electrical dyssynchrony markers were derived from epicardial electrograms (recorded or ECGi) and tank

2

potentials (defined in Table 1). In addition, the recorded and ECGi-derived sites of latest activation were

3

compared.

RI PT

4 5 6

2.3.

SC

7

ECGi reconstruction and comparison

Electrograms were reconstructed from tank potentials to experiment-specific epicardial surfaces derived from CT (Figure 1d) using ECGi methods used clinically19,20. Electrical dyssynchrony markers and

9

ATs computed from ECGi were compared to those recorded with the sock using root mean square error

M AN U

8

10

(RMSE), the RMSE of the linear regression residuals (
 ) and Pearson's correlation coefficient (R).

11

Localization error for the site of latest activation was defined as the Euclidean distance between sock

12

electrodes demonstrating the latest AT in recorded and ECGi-derived signals.

13

TE D

Statistical analysis was conducted using SAS (Statistical Analysis System). Normality was tested with a Shapiro-Wilk’s test. For each metric, the significance of differences between sequence types was tested

15

using a one-way ANOVA. A repeated-measures ANOVA was used to determine differences between recorded

16

and ECGi-derived metrics, accounting for possible interaction with sequence type. For p<0.05, the source of

17

the difference was sought using a matrix of mutually-orthogonal contrast vectors. Data are expressed as mean

18

± SD.

19

3.

AC C

20

EP

14

Results

Of the 11 experiments, normal SR was recorded prior to ablation in 4 hearts (defined as control).

21

LBBB was induced in 8 hearts (one heart had both control and LBBB states). For 5 of the 8 LBBB hearts,

22

electrical dyssynchrony was substantially augmented through further ablation and a second SR recording was

23

analyzed. Pacing sequences were analyzed for final-state LBBB hearts. 5

ACCEPTED MANUSCRIPT

1 2

3.1.

Markers for Dyssynchrony Derived from Recordings Figure 2a presents representative recorded SR activation maps in control (left), and with increasing

degrees of LBBB (middle and right). In control, multiple epicardial exit sites are seen with the majority of the

4

subepicardal myocardium activating simultaneously (short TAT, 2b). In mild LBBB (middle), the activation

5

map shows early breakthrough on the RV, and late activation over the LV or interventricular electrical

6

dyssynchrony (2b). Despite the increase in TAT, no intraventricular electrical dyssynchrony was present, with

7

LVTAT similar to control (compare 2b left to middle). With an increased degree of LBBB (right), the TAT,

8

D-LR, and LVTAT were increased.

SC

In Figure 2c, histograms of time to RS derived from torso electrodes are presented with the

M AN U

9

RI PT

3

10

corresponding lead I. Mild LBBB (middle) did not cause increased QRSd or LV-RS compared to control

11

(left), although both are increased with complete LBBB (right). The interventricular dissociation of ATs seen

12

in mild LBBB was not replicated by time to RS between the left and right torso (2b vs 2C, middle).

13

Figure 3a presents recorded activation maps from the complete LBBB heart (Figure 2, right) during RV (left), LV (middle) and BiV (right) pacing. In RV pacing, propagation was similar to complete LBBB-SR

15

with a large interventricular dissociation of activation (3b), although spread across the RV and LV freewall

16

was slower. LV pacing reversed the interventricular dyssynchrony with early activation on the LV, terminating

17

on the RV. During BiV pacing, activation spread from both pacing sites, with latest activation over the

18

epicardial aspect of the septum. While there was no interventricular AT dissociation (see overlapping red and

19

green bars, 3b), activation spread was very slow compared to control-SR (2c). The dispersion of time to RS

20

recorded from the body surface was similar to complete LBBB-SR during RV pacing (3c left), while for LV

21

and BiV pacing (3c middle and right), the time to RS was now later on the right torso.

22

AC C

EP

TE D

14

Figure 4 presents boxplots of torso and epicardially-derived electrical dyssynchrony markers during

23

control-SR, LBBB-SR and LBBB-BiV pacing. Suppl. Table 1 summarizes these markers for all sequence

24

type.Torso-derived markers (4a-c) were significantly increased with LBBB-SR compared to control-SR 6

ACCEPTED MANUSCRIPT

1

(p<0.03), although these markers were within the control range for 20, 13 and 13 % of cases respectfully (mild

2

LBBB).

3

Epicardial markers TAT and D-LR (4d and e) increased significantly in LBBB-SR (p<0.002). As with torso markers, intraventricular electrical dyssynchrony markers (LVTAT and SD-LVTAT; 4f and g) did not

5

increase in LBBB-SR except with larger levels of interventricular dyssynchrony (D-LR >25 ms and TAT > 65

6

ms).

RI PT

4

During LBBB-BiV pacing, QRSd, LV-RS, SD-RS were significantly larger than control-SR (p<0.04),

8

with no difference to LBBB-SR values (p>0.39). Interventricular electrical dyssynchrony present in LBBB-SR

9

was rectified with BiV pacing (p<0.0001), indicated by |D-LR|<20 ms and no difference compared to control

10

values (p=0.89). TAT showed no change (p=0.39), while LVTAT and SD-LVAT increased with BiV pacing

11

(p<0.02). All three markers were longer than control values (p<0.02).

12

3.2.

M AN U

ECGi Activation Maps

Figure 5a shows recorded (left) and ECGi-derived (right) activation maps for two LBBB hearts in SR.

TE D

13

SC

7

In both hearts, recorded maps showed typical LBBB patterns with early, rapid activation of the RV and late-

15

activated LV. ECGi correctly reproduced this here and in all other cases of mild and severe LBBB. ECGi in

16

Heart 2 (but not in Heart 1) presented a U-shaped activation around a line of block (very dense isochrones) on

17

the anterior epicardial aspect of the septum that was not recorded. This septal crowding of isochrones was

18

observed in 5 of 13 cases.

AC C

19

EP

14

Overall, the site of latest activation was identified to 9.1±0.6 mm (Figure 5b). That is, to the correct

20

electrode in 2 hearts (0 mm), within one electrode spacing in 7 hearts (5-14 mm), and within two electrodes for

21

the rest (14-17.8 mm).

22 23

Figure 6 shows the comparison between ECGi-derived and recorded ATs under various conditions. Overall, the overall correlation was high and residuals were low (R=0. 68±0.25, RMSE=13.4±5.3 ms and 7

ACCEPTED MANUSCRIPT

1


 =11.4±3.7 ms). R was lower in Control-SR and BiV pacing than in the other sequences (p<0.001).


2

was lower in SR than in any pacing modality (p<0.001). There was no difference in RMSE between any

3

sequence type (p=0.3).

4

3.3.

RI PT

5

ECGi-Derived Markers for Dyssynchrony

Figure 4 (d,e,f) shows that as with recorded values, ECGi showed a significant increase in TAT and DLR with LBBB-SR compared to control (p<0.03). Recorded LVTAT and SD-LVAT showed an increase in 60

7

and 80% of all LBBB-SR cases. ECGi captured this increase with a sensitivity of 83 and 100% and specificity

8

of 56 and 33% respectively.

During BiV pacing, as with recorded data, ECGi-derived TAT were not significantly different to in

M AN U

9

SC

6

LBBB-SR (p=0.72). ECGi-derived D-LR was significantly reduced (p<0.0001) and, like recorded values,

11

showed effective electrical interventricular resynchronization with no significant difference to control

12

(p=0.10). ECGi replicated the difference in LVTAT and SD-LVAT between control-SR and BiV pacing seen

13

with recorded values (p<0.04), though, unlike recorded data, no significant difference was seen compared to

14

equivalent LBBB-SR values (p>0.09).

15

TE D

10

Bland-Altman plots of recorded and ECGi-derived markers are presented in Figure 7. ECGi-derived TAT and SD-LVAT showed high RMSE (21 and 4.2 ms respectively), although there was a strong correlation

17

and low regression error (
 ) compared to recorded values. Overall, ECGi underestimated TAT and SD-

18

LVAT (p<0.04), as seen with the positive mean difference line (7a and d). For D-LR, there was significant

19

interaction with sequence type (p<0.0001), with a greater level of underestimation the larger the dyssynchrony.

20

This was seen with the slope in error between ECGi and recorded values (7b). Furthermore, the relationship

21

between ECGi and recorded values was very strong (R=0.95) with low
 . Although ECGi-derived LVTAT

22

were neither significantly longer nor shorter than recorded (p=0.56), the relationship was the weakest

23

(R=0.66), and
 similar to TAT.

AC C

EP

16

8

ACCEPTED MANUSCRIPT

1 2 3

4.

Discussion Several markers have been developed to quantify electrical dyssynchrony between the ventricles

(interventricular), or within the LV (intraventricular) using either body surface potentials14,15 or ECGi6–11. It

5

has largely been assumed that each of these metrics can accurately detect both small and large degrees of

6

electrical dyssynchrony. ECGi is further assumed to provide accurate reconstruction of activation patterns in

7

the presence of electrical dyssynchrony4,6,7,9–11, partially justified by previous studies evaluating the

8

performance of ECGi to detect activation and stimulus sites from healthy hearts during pacing12,13. In addition,

9

the accuracy of the detection of late (as opposed to early) activated sites with ECGi has never been evaluated.

M AN U

SC

RI PT

4

Our study has evaluated the accuracy of body surface potentials and ECGi derived markers to detect

11

electrical dyssynchrony with varying degrees of LBBB. The main results from our study are summarized as

12

follows. Body surface markers (LV-RS and SD-RS) provide no greater sensitivity to the degree of

13

dyssynchrony than QRSd in LBBB. Epicardial markers are very sensitive to electrical dyssynchrony, and the

14

combination of interventricular and intraventricular markers provide important information about electrical

15

activation not provided by QRSd. ECGi can be applied to reliably and noninvasively provide these markers,

16

activation patterns, and the site of latest activation.

EP

18

Detection of electrical dyssynchrony from body surface potentials

AC C

17

TE D

10

19

Recent studies have suggested that by expanding the 12-lead ECG to 53-electrode body surface

20

mapping, electrical dyssynchrony can be better diagnosed by LV-RS and SD-RS without the need for complex

21

ECGi14,15. Our study shows that LV-RS, SD-RS and QRSd were less sensitive than ECGi-derived markers.

22

Specifically, ECGi-derived D-LR showed an increase for all LBBB cases, while QRSd, SD-RS and LV-RS

23

were within the control range for several mild LBBB cases. Furthermore, none of the body surface derived 9

ACCEPTED MANUSCRIPT

1

marker showed an improvement with BiV pacing to give indication of resynchronization. Thus, ECGi adds

2

significantly to the diagnostic and therapeutic potential in cases of LBBB, although the technique is more time

3

consuming and requires cardiothoracic imaging and segmenting. The added value of LV-RS and SD-RS is explained from the fact that each torso lead contains

5

information for the entire heart, weighted heavily towards the closest ventricle. As expected, the time to RS on

6

the left and right torso gave a blurred reflection of their respective ventricular ATs (Figures 2 and 3). This

7

means the time to RS can give some indication of directionality of electrical dyssynchrony but that ECGi

8

would still provide a more precise image of the mechanisms involved.

SC

A theoretical limitation of LV-RS and SD-RS is ambiguity in marker placement on the steepest RS-

M AN U

9

RI PT

4

10

downslope (according to previous studies14,15) in LBBB and during BiV pacing, due to the tendency of ECGs

11

to have slurred, notched or irregular RS slopes21 (Figure 1d).

12

14

Epicardial electrical dyssynchrony markers

TE D

13

Our results have shown that epicardial electrical dyssynchrony markers can not only identify minor interventricular electrical dyssynchrony (D-LR as low as 15 ms), but also provides important information that

16

is not provided by QRSd. With mild LBBB, LV activation was rapid (LVTAT and SD-LVAT similar to

17

control), though delayed with respect to the RV (increase in D-LR and TAT). This is likely due to partially

18

blocked conduction through the main fascicles of the left bundle branches (possibly equivalent to incomplete

19

LBBB or intraventricular conduction delays in humans). LV activation was more slowed with an increased

20

degree of LBBB (increased LVTAT and SD-LVAT), attributed to failed conduction via all of the left Purkinje

21

branches (complete LBBB). Thus, the application of multiple epicardial electrical dyssynchrony markers helps

22

to delineate the various degrees of both inter- and intraventricular electrical dyssynchrony.

AC C

EP

15

10

ACCEPTED MANUSCRIPT

While the heart was electrically resynchronized with BiV pacing (D-LR substantially reduced),

2

myocardial activation was very slow compared to control SR, as shown by TAT, LVAT and SD-LVAT

3

remaining similar to their corresponding values in LBBB-SR. These markers paradoxically increased in some

4

cases. Our observations suggest that in the absence of complete Purkinje activation following pacing, the

5

myocardial conduction velocity is insufficient to provide normalization of synchronization, even in the

6

presence of BiV pacing. This supports the idea that CRT is only effective when considerable electrical

7

dyssynchrony is present during intrinsic activation4. Previous studies have demonstrated a hemodynamic CRT

8

response with both reductions 8,10 and increase4 in ECGi-derived dyssynchrony markers. Potential thus exists

9

to further improve resynchronization therapy, if methods can be developed to ensure resynchronization.

M AN U

SC

RI PT

1

10

Multisite and/or endocardial pacing may further improve not only interventricular, but also intraventricular

11

resynchronization, although the additional benefit on hemodynamic function remains questionable22.

12

14

Accuracy of ECGi for describing electrical dyssynchrony

With direct comparison of ECGi and recorded electrical dyssynchrony markers (Figure 7) we

TE D

13

demonstrated that ECGi successfully reconstructed the presence of electrical dyssynchrony and

16

resynchronization through pacing. Of all the electrical dyssynchrony markers, D-LR was most accurately

17

reconstructed. The variability seen in the correspondence of ECGi-derived to recorded markers is likely the

18

result of two known shortcomings of ECGi. First, ECGi tends to reduce the overall AT dispersion23. This is

19

seen directly with TAT results, and has likely influenced the accuracy of LVTAT. Second, a heterogeneous

20

spatial mismatch often exists between ECGi and recorded electrograms and thus ATs24. This spatial shift is

21

generally small but important for LVTAT, SD-LVAT and D-LR, which are derived from ATs within specific

22

regions of the heart.

AC C

EP

15

23

Inaccuracy in ECGi-derived ATs is not always a cause for electrical dyssynchrony marker inaccuracy.

24

Although ECGi derived activation patterns correlated poorly to those recorded during control-SR in this study 11

ACCEPTED MANUSCRIPT

1

(R=0.17 ±0.37), ECGi reproduced the relatively simultaneous activation over both ventricles, and short

2

activation dispersion, resulting in accurate electrical dyssynchrony markers. There is no doubt, however, that

3

an improvement in ECGi AT accuracy would improve electrical dyssynchrony marker accuracy. Previous studies suggests that targeting the LV lead to the site of latest electrical activation provides

RI PT

4

superior acute and long-term outcome for CRT25. This study demonstrates that ECGi can identify latest

6

activation to 9.1±0.6 mm. Thus, the coronary branch lying closest can be identified noninvasively with ECGi

7

prior to device implantation, markedly reducing procedure times and radiation exposure.

8

10

Comparison with previous clinical studies of ECGi and LBBB

M AN U

9

SC

5

This study corroborates the results of several previous clinical studies using ECGi in CRT candidates4,6,7,11,26. For example, in agreement with a study investigating 11 LBBB patients, we found in both

12

ECGi and contact measurements that LV free wall activation was slower during RV pacing compared to

13

intrinsic rhythm26 In contrast, these studies also frequently report the presence of functional U-shaped

14

activation and/or line of block in LBBB patients6,7,10. In this study, U-shaped activation was present in some

15

(but not all) ECGi derived activation maps (Figure 5) but not in any direct recordings. This supports the notion

16

that this is an artefact produced by ECGi. We speculate that in regions of change in electrogram morphology

17

or conduction velocity, such as the epicardial region over the septum in LBBB27 (Figure 5), or near the LV

18

pacing site28, ECGi compresses local activation creating an artefactual apparent “activation block”.

20 21

EP

AC C

19

TE D

11

Study Limitations

Validation was performed using a torso-tank model where electrical properties of the extra-cardiac

22

medium are homogeneous. Although the inclusion of other organs may alter propagation of electrical activity

23

from the heart to body surface29, previous studies have demonstrated the inclusion of inhomogeneous

24

structures produce no systematic improvement in ECGi potentials13. 12

ACCEPTED MANUSCRIPT

1

No endocardial recordings were made to assess transspetal conduction and the electrode sock did not entirely cover the posterior surface, meaning epicardial markers may have been slightly underestimated, i.e.

3

TAT appears to be 10-20 % shorter than QRSd (Figure 4). Despite this missing information, epicardial

4

markers still showed a greater sensitivity to dyssynchrony than body surface markers.

RI PT

2

5

As the ECGi methods used cannot assess endocardial or septal information, electrical activity and dyssynchrony was only assessed at the epicardial surface. Because endocardial activation differences may play

7

an important role both in CRT patient selection and therapy delivery, further investigation and development of

8

methods is required.

SC

6

Electrical dyssynchrony markers were assessed using acute models of LBBB, without cardiomyopathy

10

or infarction, in order to demonstrate their sensitivity to small degrees of electrical dyssynchrony. While we

11

expect these results are translatable to the large population of patients without structural abnormalities, the

12

incorporation of scar tissue into the model will be investigated in future studies.

M AN U

9

QRSd was computed using the 3 limb leads, not the 12-lead ECG as in clinical practice. Tank signals

14

were of very high quality, with signal averaging used to effectively eliminate noise. Indeed; QRSd computed

15

using 3 leads showed a non-substantial difference (-0.8±5.3 ms) compared to using all tank leads, and thus

16

would not change the conclusions drawn in this manuscript.

17 18

5.

19

We have demonstrated that ECGi is highly applicable to detect mild to severe electrical dyssynchrony and its

20

resynchronization during BiV pacing. ECGi provides more information than is available from the 12-lead

21

ECG or body surface potential mapping, particularly concerning quantitative measures of the spread of

22

epicardial activation and accurate detection of the latest site of activation. This non-invasive approach may

23

improve response rates to CRT by both improving patient selection, optimization of treatment delivery and

24

patient safety by reducing fluoroscopy times.

AC C

Conclusion

EP

TE D

13

13

ACCEPTED MANUSCRIPT

1 Acknowledgements

3

This work was supported by La fondation Coeur et Artères (FCA14T2), the French National Research Agency

4

(ANR-10-IAHU04-LIRYC), and the Leducq foundation transatlantic network of excellence RHYTHM. We

5

are grateful for the technical support of Charly Belterman and Jêrome Naulin.

RI PT

2

7

References

8

1.

10

Applications, Guidelines, Ongoing Trials, and Areas of Controversy. Circulation 2013;128:2407-2418. 2.

11 12

Prinzen FW, Vernooy K, Auricchio A. Cardiac Resynchronization Therapy: State-of-the-Art of Current

M AN U

9

SC

6

Linde C, Ellenbogen K, McAlister F. Cardiac resynchronization therapy (CRT): clinical trials, guidelines, and target populations. Heart Rhythm 2012;9:S3-S13.

3.

Brignole M, Auricchio A, Baron-Esquivias G, Bordachar P, Boriani G, Breithardt OA. Guidelines for cardiac pacing and cardiac resynchronization therapy: The Task Force for Cardiac Pacing and Cardiac

14

Resynchronization Therapy of the European Society of Cardiology. Developed in Collaboration with

15

the European Heart Rhythm Association. Eur Heart J 2013;34:2281-2329.

17 18

implications for patient selection and therapy improvement. Heart Rhythm 2015;12:782-791. 5.

19 20

Rudy Y. Noninvasive electrocardiographic imaging of arrhythmogenic substrates in humans. Circ Res 2013;112:863-874.

6.

21 22

Ploux S, Eschalier R, Whinnett ZI, et al. Electrical dyssynchrony induced by biventricular pacing:

EP

4.

AC C

16

TE D

13

Varma N, Jia P, Rudy Y. Electrocardiographic imaging of patients with heart failure with left bundle branch block and response to cardiac resynchronization therapy. J Electrocardiol 2007;40:S174-8.

7.

Jia P, Ramanathan C, Ghanem RN, Ryu K, Varma N, Rudy Y. Electrocardiographic imaging of cardiac

23

resynchronization therapy in heart failure: observation of variable electrophysiologic responses. Heart

24

Rhythm 2006;3:296-310. 14

ACCEPTED MANUSCRIPT

1

8.

2 3

Berger T, Pfeifer B, Hanser FF, et al. Single-beat noninvasive imaging of ventricular endocardial and epicardial activation in patients undergoing CRT. PloS one 2011;6:e16255.

9.

Silva JNA, Ghosh S, Bowman TM, Rhee EK, Woodard PK, Rudy Y. Cardiac resynchronization therapy in pediatric congenital heart disease: Insights from noninvasive electrocardiographic imaging.

5

Heart Rhythm 2009;6:1178-1185.

6

10.

RI PT

4

Ghosh S, Silva JNA, Canham RM, et al. Electrophysiologic substrate and intraventricular left

ventricular dyssynchrony in nonischemic heart failure patients undergoing cardiac resynchronization

8

therapy. Heart Rhythm 2011;8:692-699. 11.

10 11

Ploux S, Lumens J, Whinnett Z, et al. Noninvasive Electrocardiographic Mapping to Improve Patient

M AN U

9

SC

7

Selection for Cardiac Resynchronization Therapy. JACC 2013;61:2435-2443. 12.

Oster HS, Taccardi B, Lux RL, Ershler PR, Rudy Y. Electrocardiographic imaging: Noninvasive

12

characterization of intramural myocardial activation from inverse-reconstructed epicardial potentials

13

and electrograms. Circulation 1998;97:1496-1507. 13.

Ramanathan C, Rudy Y. Electrocardiographic imaging: II. Effect of torso inhomogeneities on

TE D

14 15

noninvasive reconstruction of epicardial potentials, electrograms, and isochrones. J. Cardiac

16

Electrophysiology 2001;12:241-252. 14.

Gage RM, Curtin AE, Burns KV, Ghosh S, Gillberg JM, Bank AJ. Changes in electrical dyssynchrony

EP

17

by body surface mapping predict left ventricular remodeling in patients with cardiac resynchronization

19

therapy. Heart Rhythm 2016:1-8.

20

15.

AC C

18

Johnson WB, Vatterott PJ, Peterson MA, et al. Body surface mapping using an ECG belt to

21

characterize electrical heterogeneity for different left ventricular pacing sites during cardiac

22

resynchronization - Relationship with acute hemodynamic improvement. Heart Rhythm 2016:1-7.

23

16.

24 25

Wyndham CR, Smith T, Meeran MK, Mammana R, Levitsky S, Rosen KM. Epicardial activation in patients with left bundle branch block. Circulation 1980;61:696-703.

17.

Derval N, Duchateau J, Mahida S, et al. Distinctive Left Ventricular Activations Associated With ECG 15

ACCEPTED MANUSCRIPT

1 18.

3

Imaging. IEEE Trans Biomed Eng 2016;64:1-8.

5 6

electrocardiography. Ann Biomed Eng 2006;34:1272-1288. 20.

7 8

Dubois R, Shah AJ, Hocini M, et al. Non-invasive cardiac mapping in clinical practice: Application to the ablation of cardiac arrhythmias. J Electrocardiol 2015;48:966-974.

21.

9 10

Wang Y, Rudy Y. Application of the method of fundamental solutions to potential-based inverse

RI PT

19.

SC

4

Duchateau J, Potse M, Dubois R. Spatially Coherent Activation Maps for Electrocardiographic

Surawicz R, Childers BJ, Deal LS, et al. AHA/ACCF/HRS recommendations for the standardization and interpretation of the electrocardiogram Part III. JACC 2009;53:976-981.

22.

M AN U

2

Pattern in Heart Failure Patients. Circulation: Arrhythm Electrophysiol 2017;10:e005073.

Ploux S, Strik M, Van Hunnik A, Van Middendorp L, Kuiper M, Prinzen FW. Acute electrical and

11

hemodynamic effects of multisite left ventricular pacing for cardiac resynchronization therapy in the

12

dyssynchronous canine heart. Heart Rhythm 2014;11:119-125.

13

23.

Han C, Liu Z, Zhang X, Pogwizd SM, He B. Noninvasive three-dimensional cardiac activation imaging from body surface potential maps: A computational and experimental study on a rabbit model. IEEE

15

Trans Med Im 2008;27:1622-1630. 24.

17 18

Cluitmans MJM, Bonizzi P, Karel JMH, et al. In-Vivo Validation of Electrocardiographic Imaging. JACC Clin Electrophysiol 2017;3:1-11.

25.

EP

16

TE D

14

Han Y, Liang Y, Yu H. Left Ventricular Lead Placement Targeted At the Latest Activated Site Guided By Electrophysiological Mapping in Coronary Sinus Branches Improves Response To Cardiac

20

Resynchronization Therapy. JACC 2016;67:1281.

21

26.

AC C

19

Varma N. Left ventricular electrical activation during right ventricular pacing in heart failure patients

22

with LBBB: Visualization by electrocardiographic imaging and implications for cardiac

23

resynchronization therapy. J Electrocardiol 2015;48:53-61.

24 25

27.

Strik M, Van Deursen CJM, Van Middendorp LB, et al. Transseptal conduction as an important determinant for cardiac resynchronization therapy, as revealed by extensive electrical mapping in the 16

ACCEPTED MANUSCRIPT

1 28.

3

Visualization by non-invasive electrocardiographic imaging. J Electrocardiol 2014;47:66-74. 29.

5 6 7

Bear LR, Cheng LK, LeGrice IJ, et al. The forward problem of electrocardiography: is it solved? Circ Arrhythm Electrophysiol 2015:CIRCEP--114.

30.

RI PT

4

Varma N. Variegated left ventricular electrical activation in response to a novel quadripolar electrode:

Sutherland DR, Ni Q, MacLeod RS, Lux RL, Punske BB. Experimental measures of ventricular activation and synchrony. PACE 2008;31:1560-1570.

SC

2

dyssynchronous canine heart. Circ Arrhythm Electrophysiol 2013;6:682-689.

8

AC C

EP

TE D

M AN U

9

17

ACCEPTED MANUSCRIPT

1

TABLE

2

Table 1. Electrical dyssynchrony markers derived from sock and torso potentials. DLR

(recorded and ECGi derived markers were compared)

Difference between maximum and minimum epicardial ATs

LVTAT

Difference between maximum and minimum ATs of the LV epicardium

SD-LVAT

Standard deviation of ATs on the LV epicardium

Torso Markers

QRSd

(compared to recorded sock markers and activation maps)

QRS duration estimated manually with calipers from the 3 limb leads as recommended by guidelines21

LV-RS14,15

M AN U

Average time from QRS-onset to RS-downslope for torso electrodes closer to the LV than the RV (defined by Euclidean distance) [as a measure of late LV activation] Standard deviation of the time to RS-downslope for all torso electrodes. [as a measure of interventricular electrical dyssynchrony]

AC C

EP

TE D

3

SC

TAT

SD-RS15

4

Difference in mean activation between the LV and RV free wall (positive value reflects LV delay relative to RV)

RI PT

Sock Markers6–11,30

18

ACCEPTED MANUSCRIPT

1 FIGURES

3

AC C

EP

TE D

M AN U

SC

RI PT

2

4

Figure 1. Experimental and post-processing workflow. (a) Torso-tank experimental set-up with Langendorff

5

perfused pig heart. (b) Representative ablation lesions (c) 3D fluoroscopy scan of (left) full torso and (right)

6

segmented tank (red) and sock (yellow) electrodes. (d) Epicardial potentials were reconstructed using ECGi.

7

Derived ATs and electrical dyssynchrony markers were compared to sock recordings.

19

1

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

Figure 2 (a) Representative recorded sock activation maps around the ventricular epicardium (gray) with

3

LAD (black). Activation maps during SR in control (left), and increasing degrees of LBBB (middle and right).

4

(b) Histogram of ATs over RV+LV (black), LV (red) and RV (green) electrodes. (c) Histogram of time to RS

5

recorded from left (red) and right (green) torso electrodes. Limb Lead I is overlain (black).

AC C

EP

2

20

1

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

Figure 3 (a) Representative recorded activation maps for RV (left), LV (middle) and BiV (right) pacing. (b)

3

Histogram of ATs over RV+LV (black), LV (red) and RV (green) electrodes. (c) Histogram of time to RS from

4

the left (red) and right (green) torso with Limb Lead I (black).

AC C

EP

2

21

1

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

Figure 4. Box plots of electrical dyssynchrony markers derived from (a-c) torso potentials and (d-g) recorded

3

(grey) and ECGi (white) sock electrograms during control-SR (n=4), LBBB-SR (n=13) and LBBB-BiV pacing

4

(n=7). Diamonds indicate means and red crosses outliers. Distributions significantly different to *control-SR

5

and ** LBBB-SR values (p<0.05).

AC C

2

22

1

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

Figure 5 (a) Activation maps for two LBBB-SR hearts from (left) recorded and (right) ECGi. (b) Scatter and

3

box plot for localization error of the site of latest activation (n=13; LBBB-SR).

AC C

2

23

1

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

Figure 6. Box plots of metrics comparing recorded and ECGi ATs for all and each sequence type. Diamonds

3

indicate means and red crosses outliers. *Significant difference to unmarked distributions (p<0.001).

4

AC C

2

24

1

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

Figure 7 Bland-Altman plots of recorded and ECGi-derived electrical dyssynchrony markers, with mean

3

difference (solid) and limits of agreement (dashed) lines overlain. Data collected from SR and pacing beats.

AC C

4

EP

2

25