Performance of atrial conduction velocity algorithms with error-prone clinical measurements for the identification of atrial fibrosis

Introduction

Measuring conduction velocity, as a direct consequence of fibrosis, may provide a better method to localise fibrotic regions. This study aims to assess established cardiac conduction velocity calculation methods (Triangulation, Polynomial Surface Fitting, and Radial Basis Function) in identifying areas of conduction slowing caused by fibrosis, considering realistic measurement errors.

Method

Using a human left atrium computational model, atrial activation was simulated. Each conduction velocity calculation method's performance was evaluated under uncertainties in mapping point density, local activation time assignment and electrode locations by comparing calculated conduction velocity to ground truth conduction velocity derived from high-resolution simulated atrial activation.

Results

All methods agreed well with ground truth conduction velocity maps in noise-free, high-density sampling conditions. However, Triangulation and Polynomial Surface Fitting methods showed susceptibility to noise, exhibiting significant errors under moderate to high noise levels. Radial Basis Function method demonstrated greater robustness to noise and reduced sampling density. Fibrotic region identification accuracy was high under ideal conditions for all methods but declined with increasing noise, with the Radial Basis Function method maintaining superior performance.

Conclusion

While all methods accurately estimate conduction velocity under ideal conditions, the Radial Basis Function method shows robustness to a realistic clinical noise, hence making it the most suitable to identify fibrotic regions.