Noninvasive estimation of global activation sequence using the extended Kalman filter

Area of Science:

• Biomedical Engineering
• Computational Electrophysiology
• Medical Imaging

Background:

• Noninvasive estimation of cardiac electrical activity is crucial for diagnosing heart conditions.
• Existing methods for 3-D activation mapping often require invasive procedures.
• Body surface potential mapping offers a noninvasive alternative but faces challenges in accurately reconstructing 3-D activation sequences.

Purpose of the Study:

• To develop and validate a new algorithm for noninvasive 3-D imaging of cardiac activation sequences.
• To utilize the extended Kalman filter (EKF) for recursive estimation of the 3-D activation sequence.
• To incorporate a novel regularization scheme to address the ill-posed nature of the inverse problem.

Main Methods:

• Formulation of the nonlinear relationship between 3-D activation and body surface potentials.
• Application of the extended Kalman filter (EKF) for recursive state vector optimization.
• Integration of a new regularization scheme within the EKF's prediction step.
• Simulation studies under single-site and dual-site pacing conditions.

Main Results:

• High accuracy in simulations: average correlation coefficient (CC) of 0.95 and relative error (RE) of 0.13 for single-site pacing.
• Accurate localization of pacing sites with an average localization error (LE) of 3.0 mm.
• Robust performance under dual-site pacing (CC=0.93, RE=0.16, LE=4.3 mm) and resilience to noise.

Conclusions:

• The proposed EKF-based inverse algorithm accurately estimates 3-D cardiac activation sequences noninvasively.
• The algorithm demonstrates significant potential for clinical application in diagnosing cardiac arrhythmias.
• The integration of EKF and a novel regularization scheme enhances the reliability of body surface potential mapping.