The impact of the atrial wall thickness in normal/mild late‐gadolinium enhancement areas on atrial fibrillation rotors in persistent atrial fibrillation patients

Abstract Background Some of atrial fibrillation (AF) drivers are found in normal/mild late‐gadolinium enhancement (LGE) areas, as well as moderate ones. The atrial wall thickness (AWT) has been reported to be important as a possible AF substrate. However, the AWT and degree of LGEs as an AF substrate has not been fully validated in humans. Objective The purpose of this study was to evaluate the impact of the AWT in normal/mild LGE areas on AF drivers. Methods A total of 287 segments in 15 persistent AF patients were assessed. AF drivers were defined as non‐passively activated areas (NPAs), where rotational activation was frequently observed, and were detected by the novel real‐time phase mapping (ExTRa Mapping), mild LGE areas were defined as areas with a volume ratio of the enhancement voxel of 0% to <10%. The AWT was defined as the minimum distance from the manually determined endocardium to the epicardial border on the LGE‐MRI. Results NPAs were found in 20 (18.0%) of 131 normal/mild LGE areas where AWT was significantly thicker than that in the passively activated areas (PAs) (2.5 ± 0.3 vs. 2.2 ± 0.3 mm, p < .001). However, NPAs were found in 41 (26.3%) of 156 moderate LGE areas where AWT was thinner than that of PAs (2.1 ± 0.2 mm vs. 2.23 ± 0.3 mm, p = .02). An ROC curve analysis yielded an optimal cutoff value of 2.2 mm for predicting the presence of an NPA in normal/mild LGE areas. Conclusion The location of AF drivers in normal/mild LGE areas might be more accurately identified by evaluating AWT.


| INTRODUC TI ON
Pulmonary vein isolation (PVI) is a well-established ablation strategy for paroxysmal atrial fibrillation (AF), but it is much less effective in persistent AF patients. 1 Late-gadolinium enhanced magnetic resonance imaging (LGE-MRI) has been reported to detect myocardial fibrosis. Furthermore, the progression of atrial fibrosis after catheter ablation may be associated with AF recurrence. 2 It has been previously reported that AF rotors are observed in patchy LGE areas but not in dense LGE areas, in computer simulation models. 3 That shows the importance of a qualitative and quantitative analysis of the LGE areas. Recently, the modulation of the AF rotors has been proposed as one of the effective ablation strategies for persistent AF. 4 To evaluate the location of AF rotors precisely, a novel phase-mapping system (ExTRa Mapping TM ; Nihon Kohden) has been developed. It has been shown that reducing the number of rotors detected by ExTRa Mapping leads to a reduction in AF maintenance. 5

ExTRa
Mapping is a phase map based on myocardial action potentials, which has been validated by high-resolution optical membrane potential mapping in an animal study. ExTRa Mapping has some reliability for analyzing the activation pattern the region of interest. 6 We previously reported that AF rotors detected by the ExTRa Mapping were frequently found in moderate LGE areas assessed by LGE-MRI in persistent AF patients. However, some of them were also found in normal/mild LGE areas. 7 This has implied that there are other possible structural factors associated with AF rotors. A previous computer simulation study demonstrated the role of the atrial wall thickness (AWT) as a substrate for AF rotors and marker for the identification of AF rotor locations in patient-specific atria, and the AWT gradients acted as anchoring points for AF rotors in the absence of fibrosis. 8 However, such an effect of the AWT on AF rotors has not been fully verified in humans. The aim of this study was to evaluate the impact of the AWT in normal/mild LGE areas on AF rotors in persistent AF patients.

| MRI acquisition
Before the AF ablation, LGE-MRI was performed in all patients using a 1.5T MR system (Achieva; Philips Medical) equipped with a five-channel cardiac coil. This scan technique has been previously reported. 9 First, contrast-enhancement magnetic resonance angiography (CE-MRA) of the pulmonary vein (PV)-left atrial (LA) anatomy was obtained in the coronal plane using a breath-hold three-dimensional (3D) fast field echo sequence after the injection of 0.1 mmol/kg of a contrast agent (Gadobutrol, Gadovist, Bayer Yakuhin). 10 The purpose of the scanning in the coronal plane was to reduce the number of acquisition slices and shorten the breathhold time. Next, 15 min after the contrast injection, LGE-MRI of the LA including the PVs was performed using a lateral 3D inversion recovery, respiratory navigation, ECG gating, and T1-fast field echo sequence. 11 The CE-MRA and LGE-MRI images were transferred to customized software (MRI LADE Analysis; PixSpace Inc) for image post-processing and an image analysis.

| 3D Visualization and assessment of the tissue properties
To detect normal/mild LGE areas more sensitively, we used the same protocol as in our previous study. 12 The 3D visualization method for the LGE was as follows. First, the LA in the LGE-MRI was semimanually segmented by contouring the borders between the endocardium and epicardium of the atrium, including the PVs, with reference to the CE-MRA. Second, the mean value and standard deviation (SD) of the voxel intensity was measured on the "healthy" LA wall where no hyper-enhanced areas in LGE-MRA were involved.
Third, we identified LGEs with an intensity of >1 SD on the "healthy" LA wall by a voxel intensity histogram analysis of the LA wall.
Furthermore, the degree of the intensity was categorized by a colorcoded scaling (green: >1 SD: yellow: 2-3 SD; red: >3 SD). Finally, the 3D reconstruction, color-coded LGE, and volume-rendered LA and PV image generated from the CE-MRA were semi-automatically fused. In this study, atrial fibrosis was defined as an LGE site with a signal intensity of >1 SD. To evaluate the fibrotic tissue properties, the fibrotic density was measured as the LGE-volume. The fibrotic density was defined as the volume ratio of an LGE signal intensity >1 SD (LGE-volume ratio). The details of the measurement can be found in the previous publication. 7 In this study, the areas with an LGE-volume ratio of 0% were defined as normal areas, the areas with an LGE-volume ratio of 0%-10% were defined as mild LGE areas, and moderate LGE areas were defined as areas with an LGE-volume ratio of >10%.

K E Y W O R D S
atrial fibrillation, atrial wall thickness, fibrosis, late-gadolinium enhancement magnetic resonance imaging, rotor 2.4 | Thickness measurement of the LA As shown in Figure 1, the atrial wall thickness (AWT) was defined as the minimum distance from the manually determined endocardium to the epicardial border on the LGE-MRI. Regions of interest were manually drawn in specific atrial regions and the regions of interestbased AWT was estimated in the multiplanar reconstruction images perpendicular to the LA wall. To evaluate the AWT in the normal/ mild LGE areas associated with AF rotors, a receiver operating characteristic (ROC) curve analysis was performed for the optimal values of the AWT predicting AF rotors.

| Real-time phase mapping
After the integration of the anatomical 3D models of the LA and PVs obtained from the MRI, mapping was performed using the NavX system (Abbott) as a guide. A 20-pole circular mapping catheter (Optima TM or Reflexion HD TM , Abbott) and ablation catheterreconstructed LA posterior anatomy was aligned with the MRI. 13 To detect the distribution of the AF rotors, an online real-time phase mapping system (ExTRa Mapping) was used. The detail of this mapping system was previously described. 7 We evaluated all areas where the mapping catheter has reached. ExTRa Mapping was applied to persistent AF patients and as a result, each wave dynamics were classified into three patterns, meandering rotors (MRs), multiple wavelets (MWs), and planar wave. Planar wave propagation was defined as passive activation, whereas MR and MW were defined as non-passive activations. We described the example of ExTRa Mapping in Figure 2. Moreover, ExTRa Mapping system provides a "reliability signal" to monitor the quality of the signal. The reliability signals were colored as blue, green, orange, yellow, red, or gray according to the number of electrode pairs with impaired signal. Signals were recorded when a "reliability signal" was colored by blue, green, orange. When mapping catheter could not be adjusted, the signal sensing threshold was changed from 0.03 to 0.01 mV.

| Relationship between the AWT in the normal/ mild LGE areas and AF rotors
To clarify the relationship between the AWT in the normal/mild LGE areas and AF rotors, the following were assessed: (1) the distribution of the NPAs, (2) correlation between the AWT and LGE-volume ratio in the NPAs, and (3) optimal AWT in the normal/mild LGE areas for predicting the NPAs.

| Ablation strategy
Basically, the aim of our study was to evaluate the relationship between the AWT and AF rotors on the normal/mild LGE areas but not the impact on the NPA ablation. Concerning the ablation strategy, if an NPA was found at the PV antrum, we attempted to create a PVI line that included the NPAs. If an NPA was found on the LA posterior wall, we added a Box lesion including the NPAs.

| Statistical analysis
Data are expressed as percentages for the nominal variables, medians for the ordinal variables, and means for the continuous variables.
Discrete variables were compared using the chi-square or Fisher exact test as appropriate. The mean AWT was compared among the eight segments of the whole LA groups using a one-way ANOVA and post hoc analysis with a Tukey correction for multiple comparisons of data. ROC curves were used to determine the AWT that provided the best sensitivity and specificity for the NPAs. A value of p < .05 was considered statistically significant. The correlations between two parameters were assessed using Pearson or Spearman rank correlation tests. To assess the proportion of NPAs in each group, a correction for multiple comparisons was performed. All statistical analyses were performed using EZR on R commander, version 1.36 software.

| Patient and procedural characteristics
The patient and procedural characteristics are shown in Table 1. The mean age was 66 ± 12 years, mean left atrial dimension 43 ± 8 mm, and mean left ventricular ejection fraction 60 ± 8%. Ten (67%) out of 15 patients underwent an initial AF catheter ablation. The time from the MRI acquisition to the AF ablation was 95 ± 60 days. The mean AWT in 287 areas in the LA in 15 patients was 2.2 ± 0.3 mm.

| Relationship between the AWT and LGE volume-ratio
The distribution of the NPAs and passively activated areas (PAs) according to the AWT and LGE-volume ratio are shown in Figure 4.

| Optimal AWT of normal/mild LGE areas predicting AF rotors
An ROC curve analysis yielded an optimal cutoff value of 2.3 mm and the AUC was 0.77 (0.66-0.88) for predicting the presence of an NPA in normal/mild LGE areas ( Figure 5). As for the optimal AWT, the sensitivity, specificity, and positive and negative predictive values for the cutoff values were 65.0%, 78.4%, 35.1%, 92.6%, respectively. A representative case is shown in Figure 6. Five NPAs were found in normal/mild LGE areas where the AWT was thicker than 2.3 mm.

| Comparison of the proportion of MRs/MWs in the %NP between normal/mild and moderate LGE areas
The proportion of MRs in the %NP was significantly higher in normal/mild LGE areas than in moderate LGE areas (Normal/mild LGE: 65.7 ± 8% vs. Moderate LGE: 59.0 ± 10%, p = .01). On the other hand, the proportion of MWs in the %NP was significantly lower in normal/mild LGE areas than moderate LGE areas (Normal/mild LGE: 34.3 ± 8% vs. Moderate LGE: 41.0 ± 10%, p = .01) (Figure 7).

| AWT at normal/mild LGE area and AF recurrence
AF recurrence was observed in 5 (33%) of 15 patients. Of them, two had NPAs at the normal/mild LGE areas where no direct RF application was attempted. In the 2nd procedure, NPAs were still found at the same region. The remaining three had NPA at the normal/ mild LGE area around PV antrum, however no direct RF application was attempted, because it was included within PVI line. In the 2nd procedure, NPA recurred at the PV antrum owing to recondition of the PVI. Of interest, AWT was measured as 2.4, 2.6, and 2.7 mm, respectively.

| AWT on MRI images
Recently, computed tomography (CT) images have been used to estimate the AWT. 14

| Correlation between the AWT and fibrosis
A recent study showed that a thicker LA wall was associated with a stronger atrial maintenance substrate in patients with LA enlargement assessed by echocardiography. 16 17 In our study, the LA wall thickness was negatively correlated with the LGEvolume ratio and the correlation was significant at only the NPAs but not the PAs, which was consistent with these previous results.

| Impact of the AWT on AF rotors in normal/ mild LGE areas
Recent computational studies of patient-specific atrial models, LGE-volume ratio. The NPAs (red) and PAs (blue). The AWT correlated negatively with the LGE-volume ratio in the total areas (A). This correlation was stronger in the NPAs (B) than PAs (C). AWT, atrial wall thickness; LGE, late gadolinium enhancement; NPA, non-passively activated area; PA, passively area demonstrated that AF was sustained by re-entrant drivers persisting in fibrosis border zones. 3 We previously reported that the LGE properties in anchoring AF rotors predominantly consist of moderate LGE areas in persistent AF patients. However, AF rotors are also observed in normal/mild LGE areas. 7 Therefore, we considered that there might be other structural factors related to the AF rotor besides fibrosis. Roy et al. reported that AWT gradients or fibrosis and both played an important role in anchoring AF rotors. Of important, they also reported that AF reentrant driver initiated from the area with AWT gradients in absence of fibrosis. 8 In an optical mapping ex vivo study of perfused right atria from explanted diseased human hearts, activation delays between the endocardium and epicardium during atrial pacing were more prominent in areas with an increased wall thickness, transmural fiber orientation angle gradient, and interstitial fibrosis. 18 Therefore, thicker parts of the LA could be the 3D rotational substrate perpetuating AF due to long activation delays between the endocardium and epicardium. However, those have not been validated in humans.
In our study, there was a significant difference in the proportion

| Clinical implications
As we previously reported, the AF rotors were mainly be located in moderate LGE areas, which could be detected by LGE-MRI. 7 Preprocedural LGE-MRI could evaluate the LA wall thickness as well as LGE areas precisely and would be useful to predict the AF rotors in the normal/mild LGE area. This would help in planning where to ablate in addition to the PVI at a point before the ablation procedure and might reduce frequent electrophysiological mapping. We believed that this would make a significant contribution to the realization of an AF ablation with a higher specificity. Finally, we strongly recommended that thinner AWT areas in normal/mild LGE areas should be excluded from the ablation targets.

| Study limitations
Our study had several limitations. First, the sample size was relatively small. However, we focused on the association between F I G U R E 7 The proportion of MRs and MWs in the %NP of NPAs in normal/ mild LGE and heterogenous LGE areas. %NP, non-passively activated ratio; LGE, late gadolinium enhancement; MR, meandering rotors; MW, multiple wavelets; NPA, non-passively activated area the wall thickness and electrical properties of each segment.
Fortunately, statistical significance could be found even in a total of 287 segments in 15 patients. Second, some patients underwent a prior ablation. In such cases, we could not completely discriminate between the ablation lesions and pre-existing atrial fibrosis around the PVs. However, LGE was rarely observed on MRI before ablation. Furthermore, the LGE sites might have been overestimated on the posterior wall adjacent to the vertebrae and anterior wall adjacent to the aortic cusp because of wall compression by those organs. Moreover, it might have been difficult to measure the thickness of the posterior LA wall with consistency in all patients. Thirdly, the new phase-mapping system adopted in this study may have had unknown limitations because it is widely used in Japan but not in other countries. We expect that this system will be widely used worldwide in the future. Fourth, mapping was not performed in the right atrium (RA) because of the stability of the mapping catheter and the reproducibility of the LGE-MRI assessment in the RA. Additionally, as it is difficult to contact all electrodes of mapping catheter to LAA, mapping was not performed in the LAA. Fifth, considering the spatial resolution of our MRI, it was challenging to precisely measure the thin AWT. However, the wall thickness was only applied to the body of the LA and our results are consistent with that of the previous study. Furthermore, the inter-and intra-variability of measuring AWT were acceptable (inter-observer: r = .87, p < .001; in intra-observer: r = .94, p < .001). Finally, no histological validation was performed in the LGE areas.
LGE-MRI has a potential risk of over-and underestimating fibrosis.

| CON CLUS IONS
The AF rotors were likely to be located in thick AWT areas in normal/ mild LGE areas, which were possible ablation targets. Preprocedural LGE-MRI was considered to be useful for identifying such specific areas associated with AF rotors.

ACK N OWLED G M ENTS
We would like to thank Mr. John Martin for his linguistic assistance and Mr. Tsuyoshi Sakamoto for his development of the specially customized software (MRI LADE Analysis, PixSpace Inc.).

CO N FLI C T O F I NTE R E S T
The Section of Arrhythmia is supported by an endowment from Medtronic Japan and Abbott Japan.