Validation of Left Atrial 3D LGE with Isotropic Spatial Resolution against Pre-Ablation Electroanatomic Mapping in Patients with Atrial Fibrillation

Background: 3D left atrial (LA) late gadolinium enhancement (LGE) is a promising pulse sequence for imaging LA fibrosis, a substrate (i.e., ablation target) in atrial fibrillation (AF) (1). Standard 3D LGE, with poor through-plane resolution (2.5–5 mm), is susceptible to partial volume averaging effects, particularly because the LA wall is thin (average thickness 1.5 mm (2)). Our lab has recently developed a free-breathing 3D LGE featuring isotropic resolution (1.5 mm³) using self-navigation of respiratory motion and XD-GRASP reconstruction (3). This study aimed to validate the sequence against pre-ablation electroanatomic mapping (EAM) in AF patients.

Methods: In eighteen AF patients scheduled for ablation, 3D LGE with 1.5 mm³ isotropic resolution was performed 20-30 min after 0.20 mmol/kg gadobutrol administration, prior to ablation (Figure 1). During ablation, endocardial electroanatomic voltage mapping data were acquired in sinus rhythm or during atrial pacing before pulmonary vein isolation using either NavX (Abbott) or CARTO (Biosense Webster) mapping systems. Manual LA 3D surface segmentation from MRI was performed using ADAS 3D software. EAM and LGE data were registered using pulmonary vein landmarks with a 5-mm spatial constraint. The LA appendage, pulmonary veins beyond their ostia, and mitral annulus were excluded. The posterior wall region was annotated by interpolating across the surface mesh between midpoints of all pulmonary vein ostia. LA LGE signal was normalized to mean LA blood pool intensity to calculate image intensity ratio (IIR) values (4). 12,413 total posterior wall point-by-point bipolar voltage measurements were natural log-transformed to normalize the population distribution. These data were further segmented into 268 rectangular posterior wall quadrant sectors (16 per individual), defined along the principal axes extending outward from the posterior wall centroid, to investigate spatial agreement.

Results: Point-by-point analysis yielded a moderate negative correlation (r = -0.289, p < 0.001) between posterior wall IIR intensities and discrete ln-transformed EAM voltages (Figure 2). After data smoothing into 257 EAM-populated subregions, mean IIR and ln EAM voltages showed stronger negative correlation (r = -0.487, p < 0.001). Established IIR thresholding at 1.2 and 1.45 times mean blood pool intensity (5,6), respectively, identified proportional fibrotic scar density with low voltage (< 0.5 mV) and densely scarred (< 0.1 mV) EAM, with distinct ln bipolar voltage ranges (mean -0.424, IQR 1.461 vs. mean -2.264, IQR 1.137).

Conclusion: This study demonstrates 3D LGE with 1.5 mm³ isotropic resolution can detect LA fibrosis and discriminate between diffuse fibrosis and dense scarring. Although point-by-point IIR-EAM correlation was weak, segmental (patch-based) analysis demonstrated modest correlation. Future studies with larger AF cohorts are warranted to comprehensively validate 3D LGE as a tool for guiding AF ablation.

Baseline Characteristics of AF Cohort and Left Atrial EAM Data Collection

Characteristics	AF patients (n = 18)
Age (years)	64 ± 12
Men	13 (72%)
White	17 (94%)
Height (cm)	178.1 ± 10.9
Weight (kg)	85.3 ± 11.7
Hypertension	8 (44%)
Heart failure	1 (6%)
Diabetes mellitus	2 (11%)
Vascular disease	0
Stroke/TIA	1 (6%)
CHA2DS2-VASc Score	1 [0-2]
Prior cardioversion	9 (50%)
Prior ablation	1 (6%)
LA body volume (ml)	103.2 ± 38.1
LA sphericity (%)	79.5 ± 2.6
Posterior wall surface area (cm2)	14.8 ± 5.2
EAM vendor (Carto)	9 (50%)
# EAM points analyzed	690 ± 411

Values represent mean ± SD, n (%), or median [IQR]. Vascular diseases include coronary artery disease, aortic plaque, or any other structural heart disease; AF = atrial fibrillation; TIA = transient ischemic attack. CHA₂DS₂-VASc score quantifies AF stroke risk based on risk factors.Schematic of Late Gadolinium Enhancement MRI and Electroanatomic Mapping data registration and analysis. TOP: ADAS 3D renderings of left atrial LGE-MRI (left) and EAM (right). The posterior wall region is annotated on the 3D surface with discrete EAM points projected onto the rendered LGE-MRI layer and split into rectangular subregions, using the posterior wall centroid as the grid center.
Point-by-point and segmental IIR-bipolar voltage correlation approaches. TOP: Scatter plots of all posterior wall IIR-EAM pairs (left) and nonzero IIR-EAM segments (right). BOTTOM: Natural log-transformed bipolar voltage boxplot ranges sorted by Image Intensity Ratio. IIR < 1.2 is considered “healthy”, IIR 1.2-1.45 is considered “scarred”, and IIR > 1.45 “densely scarred”, using published language from Benito et al (5).