Estimation of the PVC Origin from Simulations Using Cellular Automaton

The ability of the patient-specific heart model with activation propagation simulated using a cellular automaton principle to mimic premature ventricular contraction (PVC) and corresponding surface ECGs was examined on ten patients.Body surface mapping was performed for 5-20 minutes on patients with PVC.Their specific torso geometry models were created from CT scans.The PVC simulations were computed from 100-300 starting points in the whole ventricular myocardium model, and 12 leads of standard ECG were simulated/calculated.The measured and simulated signals were compared and evaluated by the Pearson correlation coefficient (PCC) in five intervals of the QRS duration, and the mean PCC for all 12 leads was computed. The distance between the stimulation point with the best mean PCC and the true ablation point was computed as a localization error (LE).The best mean PCC (0.95) was obtained for the shortest evaluated interval of 20 % of QRS duration.However, the LE for this interval was the worst one (42.9mm).The best mean LE (27.8 mm) was obtained for 80% and 100% of the QRS duration, with a mean PCC of 0.80.The obtained LE was comparable with the results of the inverse solution for the same patients' cohort.