Introduction: Synthetic atria expedite atrial fibrillation (AF) research but frequently rely on random fibrosis patterns, ignoring the relationship between geometric and fibrotic remodelling. Furthermore, many patients lack fibrosis, suggesting that conduction velocity (CV) may be a more sensitive indicator of the arrhythmogenic substrate.
Aim: To develop a statistical shape and appearance model (SSAM) of the left atrium that integrates anatomical and CV variability.
Methods: Data were collected from 54 persistent AF patients undergoing first-time catheter ablation. CV was derived from a 20 second charge-density mapping recording of AF. Left atrial geometries were reconstructed via intrachamber ultrasound, then processed in ShapeWorks 6.4 to establish 512 surface correspondences per subject. CV values were assigned to each correspondence by a k-nearest neighbours algorithm. We performed principal component analyses on (1) spatial coordinates (statistical shape model [SSM]), (2) conduction velocity distributions (statistical appearance model [SAM]), and (3) integrated spatial–CV data (SSAM).
Results: To explain 95% of cumulative variance, the SSM, SAM, and SSAM required 16, 22, and 26 modes, respectively. A moderate correlation (ρ=0.40) was observed between SSM mode 4 and SAM mode 2, linking regional atrial wall concavity to local CV slowing. SSAM, mode 1 primarily captured atrial volume and accompanying CV reduction on the anterior wall, while mode 2 reflected diffuse CV slowing, accentuated in larger atria. Higher modes (3–4) recapitulated the interplay between anatomical features (e.g., posterior and septal concavity) and regional conduction slowing.
Conclusion: Our SSAM framework more accurately reflects the complex interplay of anatomical and functional remodelling in AF, enabling generation of synthetic atria with realistic shape–function relationships. By incorporating CV as a sensitive substrate marker, it offers a more physiologically robust platform for large-scale AF simulations.