Author + information
- Received March 24, 2017
- Revision received July 3, 2017
- Accepted July 6, 2017
- Published online January 15, 2018.
- Koji Higuchi, MD,
- Joshua Cates, PhD,
- Gregory Gardner, BSc,
- Alan Morris, MSc,
- Nathan S. Burgon, BSc,
- Nazem Akoum, MD and
- Nassir F. Marrouche, MD∗ ()
- Comprehensive Arrhythmia and Research Management (CARMA) Center, University of Utah School of Medicine, Salt Lake City, Utah
- ↵∗Address for correspondence:
Dr. Nassir F. Marrouche, CARMA Center, University of Utah Health Sciences Center, 30 North 1900 East, Room 4A100, Salt Lake City, Utah 84132.
Objectives The purpose of this study is to evaluate the spatial distribution of late gadolinium enhancement (LGE) of the left atrium (LA) by LGE-magnetic resonance imaging in an atrial fibrillation (AF) population.
Background LGE of the LA can be a surrogate of pre-existing structural remodeling of LA.
Methods LGE-magnetic resonance imaging scans were used for 160 patients with AF (mean age 66 ± 11 years) before AF ablation. To know the spatial distribution of LGE, the extent of LGE in 6 LA subregions was examined. Overall LGE distribution was also summarized as a spatial frequency histogram using an atlas of LA shape. These data were also compared between paroxysmal AF (87 patients) and persistent AF (73 patients).
Results LGE coverage (%) in each subregion was as follows: 41.8 ± 18.9% in the left pulmonary vein (PV) antrum, 27.1 ± 16.7% in the left lateral wall, 25.8 ± 15.3% in the posterior wall, 19.7 ± 15.3% in the anterior wall, 17.1 ± 15.0% in the right PV antrum, and 12.0 ± 13.2% in the septum wall. LGE was heterogeneously distributed in the LA and was found with the highest frequency in the posterior wall near the inferior left PV antrum by the LGE histogram. A comparison of paroxysmal AF with persistent AF suggests that LGE was more expected in persistent AF compared with paroxysmal AF, particularly with a spread on the posterior and the anterior wall.
Conclusions LGE in the LA was heterogeneously distributed. LGE was highly distributed in the inferior left PV antrum near the posterior wall side, and spread on the posterior and anterior wall with AF progression.
Pulmonary vein (PV) isolation is the cornerstone procedure of atrial fibrillation (AF) ablation, especially for patients with paroxysmal atrial fibrillation (PAF) (1,2). However, in cases of persistent atrial fibrillation (PeAF), structural remodeling in the left atrium (LA) perpetuates AF, and the perpetuation of AF contributes to further developments of remodeling (3–5) (i.e., AF begets AF). Because structural remodeling was already developed and conduction heterogeneity in the LA exists in most cases of PeAF, simple PV isolation is usually not sufficient (3,4). Therefore, remodeled areas in the LA are now the additional target of ablation in patients with PeAF for better outcome by means of LA posterior wall debulking (6) and/or low voltage zone (LVZ) ablation (7–9). Knowing the existence and the spatial distribution of remodeled area in the LA before AF ablation is an increasing concern.
Structural remodeling in the LA can be visualized and quantified using late gadolinium enhancement magnetic resonance imaging (LGE-MRI), and the quantity of LGE in the LA determines the prognosis after AF ablation (10–13). Using image intensity ratio technique, another group also visualized structural remodeling by LGE-MRI (14). However, previous studies have described LGE only as a total percentage of LGE in the LA, and have not yet examined its spatial distribution. In this study, we examined the spatial distribution of LGE by means of a subsegmentation of the LA and by mapping LGE distributions to a statistical shape atlas of the LA for summary and quantitative comparison (15–17).
We retrospectively identified consecutive 160 patients (103 male; mean age 67 ± 11 years) who underwent AF ablation at the University of Utah between September 2009 and June 2011. All patients underwent an interpretable LGE-MRI within 1 month before AF ablation to assess LGE in the LA. Patients were a part of the AF database approved by our institutional review board and were compliant with the Health Insurance Portability and Accountability Act. Patients were classified as PAF or PeAF by the American College of Cardiology, American Heart Association, and European Society of Cardiology guidelines (18).
LGE-MRI was acquired on either a 1.5- or 3.0-T clinical MR scanner (Siemens Medical Solutions, Erlangen, Germany) within 1 month before ablation. The image was acquired about 15 min after contrast agent injection (0.1 mmol/kg, Multihance [Bracco Diagnostic Inc., Princeton, New Jersey]) using a 3-dimensional inversion recovery, respiration navigated, electrocardiography-gated, gradient echo pulse sequence. Typical image acquisition parameters included the following: free-breathing using navigator gating, a transverse imaging volume with voxel size = 1.25 × 1.25 × 2.5 mm (reconstructed to 0.625 × 0.625 × 1.25 mm), and an inversion time of 270 to 320 ms. The inversion time value for the LGE-MRI scan was identified using an inversion time scout scan. Repetition time, echo time, and flip angle were optimized to 5.4 ms, 2.3 ms, and 20°, respectively, for the 1.5-T scans, and 3.1 ms, 1.4 ms, and 14°, respectively, for the 3.0-T scans. Electrocardiography gating was used to acquire a small subset of phase encoding views during the diastolic phase of the LA cardiac cycle. The time interval between the R-peak of the electrocardiography and the start of data acquisition was defined using the cine images of the LA. Fat saturation was used to suppress the fat signal. These parameters were previously described (12).
LGE-MRI processing and analysis for LGE
MRI scans were evaluated for LGE using the Corview image analysis software (MARREK, Inc., Salt Lake City, Utah). The LA wall was segmented manually from LGE-MRI scans by careful tracing of the LA contour without PVs. Dark blood MRI scans, which can visualize LA endocardium, helped this process to distinguish the LA wall boundaries, especially for the boundary between LA wall and aortic root (Online Figure 1). The PV–LA junction was defined by narrowing diameter of PV insertion area compared with the LA diameter area (Online Figure 2). LGE was distinguished from normal myocardium using an interactive tool for selecting intensity thresholds that correspond with LGE in the LA wall, as previously described (10–13). Figure 1A illustrates the process for segmenting the LA wall and determining LGE area. Figure 1B shows example images of LGE distributions in the LA (trivial, mild, moderate, and extensive).
Subsegmentation of LA
The LA wall was subsegmented into 6 regions: left PV antrum, right PV antrum, posterior wall, septum wall, anterior wall, and left lateral wall. We defined each LA region as follows: (1) the left PV antrum is the LA wall extending 10 mm from the left PV–LA junction, (2) the right PV antrum is the LA wall extending 10 mm from the right PV–LA junction, (3) the posterior wall is the posterior LA extending from the LA floor to the LA roof and bordered by both PV antra, (4) the septum wall is the wall between LA and right atrium, (5) the anterior wall is the anterior part of the LA, and (6) the left lateral wall is the left side of the LA that is not covered by other areas. Figure 2 shows an example of subsegmentation. After subsegmentation, the LGE area (mm2) and LGE coverage (%) in each LA subregion were calculated using the Corview software.
Shape atlas for comparison of LGE across LA surfaces
To analyze the spatial distribution of LGE, we summarized LGE appearance data from all 160 LA surfaces by mapping each subject’s LGE appearance data to a reference shape taken from an LA atlas. We computed the LA atlas using the ShapeWorks software (Scientific Computing and Imaging Institute, University of Utah, Salt Lake City, Utah), which implements a statistical shape modeling approach described by Cates et al. (15–17). For this analysis, we used MRI scans of the LA surface after excluding PVs from the PV–LA junction. Using the LA image without PVs, the shape modeling approach places 512 landmark points on all LA surfaces, except for the PV ostium, to define a 1-to-1 mapping between surface positions on a different LA surface. Collectively, these mappings form an atlas that allows us to compare LGE appearance at corresponding LA locations across all patients.
After constructing the LA atlas, we computed the likelihood of LGE at each of the 512 landmarks. The likelihood at each landmark was estimated as the number of samples in which LGE was detected at that landmark, divided by the total number of samples. Collectively, this set of likelihoods forms an estimate of the spatial frequency distribution of LGE across all patients. This concept is depicted in Figure 3: there are 4 example sets of LGE appearance (left) and histogram of LGE distributions (right). Green spheres on examples indicate landmark points where LGE were detected. Blue spheres were landmark points where no LGE is detected. PVs were shown in gray. In the histogram, red indicates a high frequency of LGE appearance and blue indicates low frequency of LGE appearance. Gray PVs were put on the LA atlas as a reference. We also performed same procedures for PAF and patients with PeAF to compare LGE distributions. When comparing PAF and PeAF, we used different cohorts and different LA atlases, which represents each patient group.
To evaluate the error in registration or correspondence, 2 expert observers manually placed landmarks at key anatomical positions. These landmarks were compared with automatically placed landmarks.
AF-free survival after AF ablation according to the total LGE coverage (%) in the LA
We also analyzed the AF-free survival rates after ablation. In this analysis, patients were divided into 4 stages according to the total LGE coverage (%) in the LA as follows; stage 1, <10%; stage 2, ≥10% to 20%; stage 3, ≥20% to 30%; and stage 4, ≥30%.
Statistics and analysis
SPSS version 16.0 (SPSS Inc., Chicago, Illinois) was used for all analyses. Continuous data were described as mean ± SD. Continuous data of the same subject were analyzed by the paired Student’s t-test and analysis of variance with the post hoc Tukey test for significant differences. An unpaired Student’s t-test was used for the comparison of data between PAF and PeAF. The chi-square or Fisher exact tests were used to test for differences in categorical measurements such as the number of patients with certain comorbidity. A probability value of p < 0.05 was considered statistically significant. Kaplan-Meier analysis was used for graphical assessment of time-related recurrences after single AF ablation.
Our patient population includes 87 patients with PAF and 73 patients with PeAF. PeAF group is significantly older (p = 0.029), and includes significantly more individuals with hypertension (p = 0.021), diabetes (p = 0.047), and cardiomyopathy (p = 0.002) than PAF group (Table 1). In this patient cohort, AF-free survival rates 1 year after single AF ablation in each stage of total LGE coverage (%) in the LA were as follows: 80.4% in stage 1 (n = 63), 67.8% in stage 2 (n = 57), 55.8% in stage 3 (n = 31), and 22.2% in stage 4 (n = 9). The Kaplan-Meier survival curve is shown in Online Figure 3.
LGE in each LA area: Entire population
The LGE area in each subregion was 427.7 ± 238.1 mm2 in the left PV antrum, 241.9 ± 225.0 mm2 in the right PV antrum, 822.5 ± 565.2 mm2 in the posterior wall, 248.3 ± 313.3 mm2 in the septum wall, 557.9 ± 487.4 mm2 in the anterior wall, and 566.7 ± 385.5 mm2 in the left lateral wall (Table 2). The LGE coverage (%) in each subregion was 41.8 ± 18.9% in the left PV antrum, 17.1 ± 15.0% in the right PV antrum, 25.8 ± 15.3% in the posterior wall, 12.0 ± 13.2% in the septum wall, 19.7 ± 15.3% in the anterior wall, and 27.1 ± 16.7% in the left lateral wall (Table 2, Figure 4). The left PV antrum was significantly occupied by LGE compared with all other regions (p < 0.001). The left lateral wall had the second highest and the posterior wall had the third highest coverage by LGE. There was no difference between the lateral wall and the posterior wall region (p = 0.997). The posterior wall had higher LGE coverage compared with anterior wall (p = 0.008). Over all, these data suggest that LGE exhibits a heterogeneous distribution, and is not uniformly distributed (Figure 4) (p < 0.001 analyzed by analysis of variance with the post hoc Turkey honest significant difference).
LGE in each LA area: PAF versus PeAF
In addition to an overall assessment of LGE for the AF population, we also compared the LGE area (mm2) and coverage (%) in each LA subregion between patients with PAF and patients with PeAF (Table 3). Comparisons of LGE areas (PAF vs. PeAF) were as follows: 379.9 ± 201.9 mm2 versus 484.7 ± 265.4 mm2 in the left PV antrum (p = 0.005), 189.7 ± 193.7 mm2 versus 304.1 ± 244.5 mm2 in the right PV antrum (p = 0.001), 671.2 ± 510.1 mm2 versus 1,002.7 ± 578.0 mm2 in the posterior wall (p < 0.001), 172.5 ± 214.5 mm2 versus 338.7 ± 382.8 mm2 in the septum wall (p < 0.001), 469.6 ± 428.9 mm2 versus 663.1 ± 533.2 mm2 in the anterior wall (p = 0.012), and 544.0 ± 392.7 mm2 versus 593.8 ± 377.6 mm2 in the left lateral wall (p = 0.417). The LGE area in each LA subregion was significantly greater in patients with PeAF than in patients with PAF, except for the left lateral wall. The total LGE area was also significantly greater in patients with PeAF than in patients with PAF (2,426.9 ± 1,501.5 mm2 vs. 3,387.1 ± 1,828.5 mm2; p < 0.001).
The LGE coverage (%) was also higher in PeAF than PAF in the posterior wall (22.9 ± 14.8% vs. 29.3 ± 15.3%; p = 0.008), the right PV antrum (14.8 ± 14.6% vs. 19.9 ± 15.1%; p = 0.032), and the septum wall (10.1 ± 12.0% vs. 14.4 ± 14.2%; p = 0.039). In the anterior wall, the data suggest a trend of difference (17.6 ± 14.1% vs. 22.3 ± 16.3%; p = 0.051). No difference was seen for the left PV antrum (40.3 ± 17.7% vs. 43.7 ± 20.2%; p = 0.260) and the left lateral wall (27.9 ± 17.5% vs. 26.2 ± 15.9%; p = 0.544). Total LGE coverage (%) was also significantly higher in patients with PeAF compared with PAF (21.2 ± 11.6% vs. 24.6 ± 11.9%; p = 0.043) (Table 3).
Distribution of LGE in the entire AF population
Four different views of the LGE frequency histogram are shown in Figure 5. The red areas indicate regions of the LA wall in which >60% of patients showed appearance of LGE. The orange areas indicate regions in which 40% to 50% of patients exhibited LGE. From this histogram, our AF patient population had the highest probability of LGE in the inferior left PV antrum region, especially on the posterior wall side. A moderate probability of LGE was in the posterior wall and the left lateral wall. LGE was more or less uniformly distributed on the anterior LA wall. Our histogram analysis of LGE distribution agrees well with the subregion analysis given in Figure 4 and Table 2.
Distribution of LGE: Comparison between PAF and PeAF from the frequency histogram
We generated separate histograms of LGE distributions for PAF and PeAF to examine the difference in LGE distribution (Figure 6). From the comparison of these histograms in 4 different views, LGE is more widely distributed in patients with PeAF compared with patients with PAF, especially on the posterior and anterior wall. These distributions underscore the statistical results regarding the comparison of regional concentrations of LGE in these 2 groups that are shown in Table 3.
Error in registration or correspondence of point correspondence approach
The distances from the manually placed landmarks to automatically placed landmarks were evaluated to establish the variation in observer landmark placement. For each observer, the distance between the manual position mapped to the reference atlas and the atlas based landmark position, chosen independently, was measured. For observer 1, the mean distance was 4.3 ± 2.2 mm. For observer 2, the mean distance was 5.3 ± 2.6 mm.
Major findings of this study can be summarized as follows. 1) LGE was heterogeneously distributed in the LA when analyzed from LGE-MRI and the histogram of LGE distribution. 2) Regions with high LGE coverage were the left PV antrum, the left lateral wall, and the posterior wall, which can be seen on the histogram of LGE distribution, even from PAF stage. 3) LGE spreads on the posterior LA wall and occurs on the anterior LA wall in PeAF, which can be seen on the histogram of LGE distribution. 4) Overall LGE coverage was significantly greater in PeAF than PAF.
Heterogeneity of LGE distribution in the LA: Histology, anatomy, and electrophysiology
In a histological study of autopsied human hearts by Platonov et al. (4), the LA posterior wall at the inferior PV level was the only time-dependent area of significant fibrosis. In an experimental study using a sheep AF model with heart failure, LA fibrosis in posterior wall and the PV antrum played a significant role in maintaining AF (19). Using patients with chronic AF who underwent mitral valve surgery and autopsied control, Corradi et al. (20) also demonstrated that fibrosis occurs more in the LA posterior wall than in the left atrial appendage. These histological studies correlate well with our present study.
We suppose that the heterogeneous LGE distribution in particular LA regions is derived from the external structures around the LA. More LA remodeling could occur due to a mechanical pressure or a friction with the aorta, esophagus, and vertebral body (21). The proximity of the LA posterior wall and the descending aorta can be recognized at the left inferior PV ostium level where mechanical friction may occur (22). This finding could explain why we found more LGE on the left PV ostia (especially around left inferior PV ostia) compared with the right side. The vertebral body and ascending aorta attach more on the distended LA wall as AF progress, and increase remodeling on the attached area (posterior and anterior LA wall). That can be an explanation why LGE spreads on the posterior wall and occurs on the anterior wall when progressing from PAF to PeAF in this study.
The pressure stress other than the mechanical friction to the LA wall also varies due to different pericardial structures, resulting in heterogeneity of LGE (23). In a study by Hunter et al. (24), 3-dimensional computed tomography scans were used to estimate the wall pressure stress in patients with PeAF. Wall stress was particularly common around left PV ostia and the left atrial appendage ridge (24).
Nakahara et al. (25) demonstrated that areas of pivot activation during AF (which were cores of localized activation in AF) were found at the anterior wall, left PV antrum, and posterior wall where external structures such as aorta and vertebral body attached. They found LVZ at same areas during sinus rhythm in the LA (25). Our findings correlated well with findings of these previous reports.
Emergence and progression of LGE: From PAF to PeAF
When comparing LGE distribution of PAF and PeAF, no difference was seen on the left PV antrum and the left lateral wall, although a significantly greater LGE distribution was recognized in PeAF in other areas (especially on the posterior wall and anterior wall). This result suggests that LGE was common on the left PV antrum or left lateral wall, even in patients with PAF with rather healthy LA myocardium, and then occurs in other areas as disease progresses to PeAF. The LGE frequency histogram in this study also suggests this finding.
Clinical implications: The importance of modifying structural remodeling
A recent large, multicenter, randomized trial (STAR AF II trial [Substrate and Trigger Ablation for Reduction of Atrial Fibrillation Trial Part II]) showed that empiric linear ablation or complex fractionated electrogram ablation did not have a benefit compared with simple PV isolation in patients with PeAF (26). However, we have reported the importance of LGE in the LA for predicting the AF recurrence after ablation (10–13). Recent studies also showed the importance of LVZ in AF ablation. By additional modification of LVZ in the LA, the rate of AF recurrence was decreased than only by PV isolation (7–9,25,27). Therefore, knowing the sites of LVZ appearance is now an increasing concern. Our study showed that LGE (which can be a surrogate of remodeled area in the LA) was quite heterogeneous. Evaluating the site-specific LGE appearance before AF ablation might be useful, in additional modification at remodeled site of LA together with PV isolation.
A major limitation is that we did not evaluate the LGE in the roof of LA, because we could not display the top of the LA enough because of the partial volume effect as LGE-MRI were evaluated axially.
A second limitation is that the area of LGE seems broad compared with electro-anatomical mapping of usual voltage settings (healthy tissue, >0.5 mV; mild illness, <0.5 and >0.1 mV; and scar, <0.1 mV). However, in our previous paper that compared LGE area and voltage mapping, the voltage setting was as follows: 1) healthy tissue, >1.0 mV; 2) mild illness, <1.0 mV and >0.5 mV; 3) moderate illness, <0.5 mV and >0.1 mV; and 4) scar, <0.1 mV (10).
The heterogeneity of LGE in LA in patients with AF was demonstrated in this study using LGE-MRI. The areas with the greatest LGE coverage were in the LA posterior wall around left inferior PV ostium. When progressing from PAF to PeAF, LGE seems to become more extensive, especially on the posterior wall and anterior wall.
COMPETENCY IN MEDICAL KNOWLEDGE: The amount of LGE in the before ablation is highly important in determining the AF recurrence after ablation. This study revealed LA subregions where LGE frequently occurs and how LGE spreads as AF progresses from PAF to PeAF.
TRANSLATIONAL OUTLOOK: Knowing the site-specific LGE appearance using LGE-MRI before AF ablation might be quite useful in additional modification at LA remodeling site together with PV isolation.
Dr. Marrouche is a stock holder of MARREK, Inc., which is a main developer of Corview software. Dr. Morris has equity interest in MARREK, Inc. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
All authors attest they are in compliance with human studies committees and animal welfare regulations of the authors’ institutions and Food and Drug Administration guidelines, including patient consent where appropriate. For more information, visit the JACC: Clinical Electrophysiology author instructions page.
- Abbreviations and Acronyms
- atrial fibrillation
- left atrium
- late gadolinium enhancement magnetic resonance imaging
- low voltage zone
- paroxysmal atrial fibrillation
- persistent atrial fibrillation
- pulmonary vein
- Received March 24, 2017.
- Revision received July 3, 2017.
- Accepted July 6, 2017.
- 2018 American College of Cardiology Foundation
- Ouyang F.,
- Bansch D.,
- Ernst S.,
- et al.
- Burstein B.,
- Nattel S.
- Platonov P.G.,
- Mitrofanova L.B.,
- Orshanskaya V.,
- Ho S.Y.
- Segerson N.M.,
- Daccarett M.,
- Badger T.J.,
- et al.
- Rolf S.,
- Kircher S.,
- Arya A.,
- et al.
- Yang G.,
- Yang B.,
- Wei Y.,
- et al.
- Yamaguchi T.,
- Tsuchiya T.,
- Nakahara S.,
- et al.
- Oakes R.S.,
- Badger T.J.,
- Kholmovski E.G.,
- et al.
- McGann C.,
- Akoum N.,
- Patel A.,
- et al.
- Cates J M.M.,
- Fletcher P.T.,
- Whitaker R.
- Cates J F.P.,
- Styner M.,
- Shenton M.,
- Whitaker R.
- Oguz I C.J.,
- Fletcher P.T.,
- Whitaker R.,
- Cool D.,
- Aylward S.,
- Styner M.
- Fuster V.,
- Ryden L.E.,
- Cannom D.S.,
- et al.
- Tanaka K.,
- Zlochiver S.,
- Vikstrom K.L.,
- et al.
- Asirvatham S.J.,
- Gard J.J.
- Hunter R.J.,
- Liu Y.,
- Lu Y.,
- Wang W.,
- Schilling R.J.
- Nakahara S.,
- Yamaguchi T.,
- Hori Y.,
- et al.
- Jadidi A.S.,
- Lehrmann H.,
- Keyl C.,
- et al.