Two Approaches for Inverse Localization from Clinical ECG Data Using Heart Surface Potentials

Jana Svehlikova1, Nika Rasoolzadeh2, Beata Ondrusova3, Peter Hlivak4, Yesim Serinagaoglu Dogrusoz2
1Institute of Measurement Science, SAS, 2Middle East Technical University, 3Institute of Measurement Science, 4National Institute for Cardiovascular Diseases


Abstract

The standard approach for the inverse problem of electrocardiography computes the extracellular potentials on the closed epicardial or endo-epicardial surface. In the case of localization of premature ventricular con-tractions (PVC), the point with the earliest activation time (AT) is considered as the PVC origin. The AT is estimated as the time with minimum derivative of the electrogram (EGM), which is a challenging task because the EGM is not smooth enough. Therefore, we studied a different approach assuming that at the beginning of the PVC (thus also EGM) signal, only a small area of the ventricles is activated. The inverse localization of the PVC origin using a homogeneous torso, and a standard method for epicardial potentials was performed on the measured body surface potential maps (BSPMs) of ten patients who underwent a successful radiofrequency ablation procedure. The true PVC location was known, as well as the patient-specific torso geometry from the CT scan. Then the inverse solution was computed for the first 30 ms of the PVC signal by the proposed method. The columns of the same forward matrix were used separately for every single point epicardial potential computation by the least squares method. The relative residual error between the measured and computed BSPMs from every point was evaluated, and its minimum indicated the presumed origin of PVC. The localization error (LE) between the estimated and true PVC origins for both methods was evaluated and compared. The average LE for the standard method was 37.2±15.9 mm for the epicardial and 37.6±17.8 mm for the endo-epicardial surface. For the second method, the average LE improved to 25.5±10.2 mm and 26.3±11.8 mm for epicardial and endo-epicardial surfaces, respectively. For a specific task of identifying local events on the heart, a simplified assumption of a single-point equivalent electrical generator can improve the LE.