Comparison of Activation Times Estimation for Potential-Based ECG Imaging

Area of Science:

• Biomedical Engineering
• Computational Biology
• Cardiovascular Electrophysiology

Background:

• Cardiac activation times (ATs) are crucial for analyzing heart electrical activity but are difficult to estimate accurately.
• Noise, signal fractionation, and baseline wander complicate AT estimation.
• Electrocardiographic imaging (ECGI) introduces further challenges like over-smoothing and inverse problem ambiguities, leading to artifacts in AT maps.

Purpose of the Study:

• To compare different methods for estimating cardiac activation times (ATs) using a common dataset of simulated ventricular pacings.
• To evaluate the impact of various electrocardiographic imaging (ECGI) reconstruction methods and source models on AT estimation accuracy.
• To identify sources of error and artifacts in AT mapping derived from ECGI.

Main Methods:

• A community effort compared AT estimation methods on simulated ventricular pacing data.
• ECGI reconstructions utilized three surface source models: transmembrane voltages, epi-endo potentials, and pericardial potentials.
• All reconstructions employed 2nd-order Tikhonov regularization with six different parameters.

Main Results:

• Pacing site significantly affected AT correlation coefficients, with lateral pacings yielding better results than septal pacings.
• Differences in AT estimation accuracy between methods and source models were not always well-reflected by correlation coefficients.
• Purely temporal methods exhibited the most severe artificial lines of block artifact.

Conclusions:

• Transmembrane voltage models yield more precise ATs and are less susceptible to artifacts compared to epi-endo and pericardial potential models.
• Understanding the sources of error in AT estimation is critical for both clinical application and ECGI method evaluation.
• The choice of source model in ECGI significantly impacts the quality and reliability of estimated cardiac activation times.