Gaussian process manifold interpolation for probabilistic atrial activation maps and uncertain conduction velocity

Area of Science:

• Computational biology
• Medical imaging
• Cardiovascular research

Background:

• Local activation time (LAT) maps are crucial for understanding atrial fibrillation pathophysiology.
• Conduction velocity (CV) derived from LAT gradients offers insights into atrial substrate properties.
• Quantifying uncertainty in CV calculations is essential for reliable interpretation.

Purpose of the Study:

• To develop a method for probabilistic interpolation of uncertain LAT on human atrial manifolds.
• To enable the calculation of statistics for predicted CV, incorporating uncertainty.
• To assess the reliability of CV measurements in atrial fibrillation.

Main Methods:

• Utilized reduced-rank Gaussian processes (GPs) for probabilistic interpolation of uncertain LAT.
• Developed Gaussian process manifold interpolation (GPMI) accounting for atrial topology.
• Applied GPMI to clinical cases and validated against simulated data.

Main Results:

• Demonstrated GPMI's capability for probabilistic interpolation of uncertain LAT on atrial manifolds.
• Enabled calculation of CV statistics, reflecting uncertainty.
• Showcased that CV uncertainty is influenced by data density, wave propagation direction, and CV magnitude.

Conclusions:

• GPMI provides a robust framework for quantifying uncertainty in CV derived from LAT maps.
• The method enhances the interpretation of atrial substrate properties in patients with atrial fibrillation.
• GPMI is adaptable for probabilistic interpolation of other uncertain quantities on non-Euclidean manifolds.