Clinical Electrophysiology -> Ventricular Arrhythmias -> Mapping & Imaging D-PO06 - Poster Session VI (ID 26) Poster

D-PO06-049 - Approximating Repolarization Time From Filtered Unipolar Electrograms (ID 681)


Background: The Wyatt method (WM) approximates repolarization time (RT) from a unipolar electrogram (UE) as the maximum slope during the T-wave. However, clinical UEs may be subjected to unavoidable high-pass (HP) filtering at acquisition, which distorts the repolarization phase and makes WM unreliable.
Objective: Accurately approximate RT from HP filtered UEs.
Methods: We computed UEs with morphologies corresponding to early and late activation and repolarization times. We measured RT error by calculating RT using the WM before and after applying a digital HP filter to the UEs, with each of the 8 CARTO cutoff frequencies of 0.05-30Hz.We trained a neural network (NN) with 100,000 filtered synthetic UEs (2Hz HP) to generate a probability distribution of RT (NN Prob). We applied the NN to 82 unfiltered human UEs acquired ex-vivo that we digitally filtered, and to 150 CARTO patient UEs.
Results: The mean (maximum) RT errors for the lower CARTO HP frequencies were 6.48 (9.31) ms for 0.5Hz, 15.1 (125) ms for 1Hz, and 59.7 (215) ms for 2Hz. For the ex-vivo data, the mean RT approximating error by the NN was 13.16 ± 8.95ms, and by the WM was 26.17 ± 54.94ms. 77% of the CARTO RT approximations matched those of the same synthetic morphologies.
Conclusion: Large errors occur using the WM on filtered UEs, and its use is not appropriate for even lightly filtered UEs.Our NN is validated with human UEs, and performs reliably on clinical UEs. This method provides RT from clinical UEs that would otherwise be inaccessible, offering access to reliable patient-specific in-vivo repolarization maps.