HIGH AND LOW FREQUENCY ELECTROCARDIOGRAPHIC IMAGING FOR DIFFERENTIATING ENDO VS EPICARDIAL EXIT SITES OF VENTRICULAR ARRHYTHMIAS

Abstract

Background Non-invasive assessment of the 3D activation pattern of ventricular arrhythmias (VA) along the endo-epicardial axis remains a challenge. Conventional ECGI reconstructs low frequency (LF) potentials on the epicardium. The high frequency (HF) signal amplitude of a body surface potential (BSP) reflects the sum of simultaneously depolarized myocardial cells close to the recording electrode. Derived HF-activation times (HFAT) reflect when activation has propagated halfway through the myocardial wall. Combining HF & LF components may allow to derive the help define transmural activation sequence of VAs. Objective Estimate epicardial activation times (EAT) and HFAT from BSPs recorded during VAs. Evaluate if the differences of low frequency epicardial AT & HFAT allow to distinguish endo vs. epicardial VA exit sites. Methodology Patients with 252-electrode-ECGI (sample rate 1kHz) and contact mapping for VA were retrospectively reviewed & included if >100 VE/ VT beats recorded on ECGI. Unipolar EGMs were reconstructed using commercial software and EATs annotated to max. -dV/dt with onset of surface QRS as reference. HFATs were computed offline using a custom-made Matlab app. BSP channels with peak-to-peak noise >120uV for >34% of the time were rejected & VA beats signal-averaged in time and frequency domain. The amplitude envelope of each BSP was computed in 5 frequency bands between 150-450Hz using Hilbert transform and then averaged per band. The HFQRS was constructed as the mean of the normalized average envelope of all frequency bands and projected onto the epicardium using the virtual points created for the inverse reconstruction. The HFAT was calculated from onset of surface-QRS to the centre of the mass of HFQRS. EAT & HFATs were subtracted and 'Diff-maps' generated. If EAT was earlier than HFAT at the exit site, an epicardial source was suspected, conversely if EAT later, an endo/ midmyocardial source. Results 30 patients were studied (57% male; 57+/-17years; LVEF 45+/-19%; 55% LGE+ in CMR). 17 (58%) were VEs, 13 (42%) re-entry-VTs. Mean number of averaged beats was 1029+/-718 for VTs, 258 +/-163 for VEs. Average number of channels included in HFAT estimation was 126 +/-56 (49.8 +/-22.2% of the vest). In 17 (57%) cases Diff-maps were in agreement with EAM for epi vs endo SoO. In 10 (33.3%) the DIFF-map failed to detect the correct SoO - of these, 7 (23%) were VAs associated with outflow tracts or basal septum. In 3 (10%) the diff-values were close to zero (+/-5ms) preventing a clear distinction of epi vs non-epi. At true epicardial origins, EAT was 47+/-20ms earlier than the HFAT, compared to 39 +/-19 ms at false positive epicardial origins. At sites of true non-epicardial origin EAT 12.5+/-9ms was later than HFAT, there were no false positive endocardial cases. Sensitivity for detecting an epicardial source was 92.8%, with specificity 28.5%. Conclusion HFECGI is a novel method to non-invasively derive the transmural activation sequence in VA and detected epicardial exit sites with high sensitivity but low specificity. Dedicated recording devices with higher sampling rates could improve the diagnostic accuracy as the lower frequency band of the current equipment limited the ability to spatially locate and identify the transmural transition below the electrode in DIFF-ECGI maps.