Accelerated 3D Late Gadolinium Enhancement Pulse Sequence with Isotropic Spatial Resolution, Self-Navigation of Respiratory Motion, and Arrhythmia Rejection for Quantification of Left Atrial Fibrosis in Patients with Atrial Fibrillation

Background The left atrium (LA) is a thin structure (1.5–2.5 mm). Conventional 3D late gadolinium enhancement (LGE) pulse sequences for imaging the LA typically have limited through-plane resolution (2.5–5 mm) or prolonged scan times (10–15 min), resulting in poor image quality and limited clinical translatability. We sought to develop and test whether a 3D LGE pulse sequence featuring 1.5 mm³ isotropic spatial resolution and a predictable scan time of 600 heartbeats can produce diagnostically acceptable image quality at 1.5 T for quantification of LA structural remodeling in patients with atrial fibrillation (AF). Methods A 3D LGE pulse sequence was developed using gradient-recalled echo (GRE) readout, stack-of-stars k-space sampling with variable density along the partition direction, retrospective arrhythmia rejection, and XD-GRASP reconstruction with self-navigation of respiratory motion. Thirty patients with AF (twenty imaged prior to first-time ablation, ten prior to repeat ablation) and twenty healthy controls were scanned using this pulse sequence. The upper limit of normal image intensity ratio (IIR) was derived from the control group; specifically, 2 standard deviations (SDs) above the mean. Suspected LA fibrosis was quantified by one rater. Two cardiothoracic radiologists independently graded the image quality across three categories (conspicuity, artifacts, noise) using a 5-point Likert scale per category, with a score of 3 considered clinically acceptable. Results The proposed pulse sequence produced diagnostically acceptable image quality scores in all three categories for AF and control cohorts. Retrospective rejection of data acquired during ectopic heartbeats significantly (p<0.05) improved contrast-to-noise ratio (+3.7%) and sharpness (blur metric -3.2%) metrics, albeit at the expense of a significantly reduced signal-to-noise ratio (-4.4%). The upper limit of normal IIR was 1.43. Using this cutoff value to identify suspected LA fibrosis, compared with the control group (2.47 ± 2.24%), the first-time ablation group (7.45 ± 7.38%) was not significantly different (p = 0.14), whereas the repeat ablation group (14.94 ± 13.83%) was significantly different (p < 0.05). Conclusion The proposed pulse sequence produces diagnostically acceptable image quality at 1.5 T and provides evidence of LA structural remodeling in AF patients, including greater suspected LA fibrosis burden, relative to control subjects.