3.1 Patient and procedure characteristics

A total of 10 patients were enrolled with mean age 69 ± 10 years. One patient was female. Mean LA diameter was 4.6 ± 0.5 cm. Of the 10 patients, 5 presented in AF and were cardioverted into SR at first attempt; 3 presented in SR and were induced into AF after voltage mapping; and 2 presented in SR and could not be induced into AF. As one additional SR patient had inducible AF that then sporadically organized into AT, only 7 patients were EGF mapped biatrially in both rhythms per protocol.

All patients received PVI, and successful isolation was confirmed after at least a 20-min waiting period. Three patients received adjunctive anterior line ablations, one of whom also received a roof line; one patient received a cavotricuspid isthmus (CTI) ablation; and one patient received a lower posterior line ablation. One of the anterior lines and the lower posterior line were performed after PVI in AF, but the other therapeutic ablations were performed in SR after all other mapping was complete.

3.2 Characterization of EGF-identified source signatures in AF and SR

Sources were defined as origins of divergent flow with source activity (SAC) quantified as the percentage of 2 s segments within a recording during which the source was emanating flow. A total of 11 divergent sources with a SAC > 20% were detected during EGF mapping in AF, 7 in the LA and 4 in the RA. Mean SAC was 32 ± 9%. Eight of the 11 (73%) locations identified as active sources during EGF mapping in AF converted to passive (convergent) sources or “sinks” when mapped in SR (P < 0.0001). These passive sources, or “sinks,” were defined as points where EGF converged from 4 perpendicular orientations to one single point, forming a local minimum in EGFC at the precise point of convergence. Sink regions were always located within 1 electrode distance, or approximately 15 mm, of the active source. Example EGF maps with source-sink conversion are shown in Fig. 2.

Fig. 2

Example LA and RA EGF maps from a single patient in AF where active sources were present and corresponding LA and RA EGF maps in SR. In the LA map, the source at D2 in AF (SAC 27.4%) converted to a sink where flow vectors come together from four orthogonal directions at D2 in SR. In the RA map, the source near D4 in AF (SAC 32.9%) converted to a sink near D3 in SR. In both cases, EGFC – visualized as purple flow vectors v. blue – is higher in SR v. AF

Five of 7 (71%) sources in the LA converted into sinks, while 3 of 4 (75%) of the sources in the RA did so. Mean SAC was insignificantly higher, albeit with a small sample size, among sources that converted into sinks compared with those that did not (33 ± 10% v. 26 ± 7%, P = 0.30). No sinks were observed in SR at locations not linked to a source in AF. Among the 3 sources that did not convert to sinks, one still appeared as a source in SR, while all other sources in SR did not correlate with sources in AF. Moreover, the source that endured in SR was located in the anterior mid-roof, likely corresponding to Bachman’s bundle. Lastly, only 10 total sinks were identified in SR, 8 of which (80%) were therefore tied to active sources in AF (P < 0.0001). The remaining 2 sinks presented as sinks in both SR and AF.

For the 11 pairs of EGF maps on which sources in AF were identified, the EGFC was higher in SR than in AF (0.96 ± 0.30 v. 0.59 ± 0.14, P = 0.015). However, regions of high EGFC in AF were not correlated to regions of high EGFC in SR. The sources and sinks themselves were also not found to be consistently localized to anatomic regions of high or low EGFC or voltage within the maps.

3.3 EGFC increases globally but not with local correlates during rhythm conversion from AF to SR

Seven patients underwent full EGF mapping in all positions in both SR and AF, and their EGFCs were compared. Mean EGFC was higher in SR than AF (1.00 ± 0.11 v. 0.74 ± 0.14, P = 0.0097). In all but one of the patients, median EGFC was higher in SR than AF, as shown in Fig. 3. A linear mixed effects model of the data also demonstrated that rhythm had a significant influence on EGFC (log-likelihood − 0.235, P < 0.0001). Notably, the one patient with higher median EGFC in SR had junctional beats and PR interval fluctuations in SR.

Fig. 3

Box-and-whisker plot of EGFC from all paired recording locations of patients in AF v. SR. In all but one case, median EGFC was higher in SR. A linear mixed effect model of this data demonstrated that the EGFC in the two rhythms were significantly different (P < 0.0001)

Locally, the EGFC across all electrodes with high contact was also higher in SR than in AF (0.94 ± 0.48 v. 0.66 ± 0.34, P < 0.0001). The distribution of EGFC by electrode in AF and SR is shown in Fig. 4a. Heat maps were created to show the relative distribution of EGFC across the splines from all paired recordings in AF and SR (Fig. 4b); however, there was no overarching similarity in the distribution other than generally lower EGFC being observed at the beginning and end of the splines v. middle. All electrodes with good contact that could be positionally paired had EGFC plotted in AF v. SR, as shown in Fig. 4c. No meaningful correlations were observed between the locally aligned EGFCs.

Fig. 4

Electrode-level comparison of EGFC in AF v. SR. a shows the distribution of electrodes by EGFC in each rhythm from all positionally paired maps of patients with EGF maps in both SR and AF. AF is shown in orange v. SR in blue. Low-contact electrodes were excluded. EGFC was 38% higher on average in SR v. AF (P < 0.0001). b shows a heatmap of the EGFC by spline number and electrode number of all pooled electrodes from the maps in a. Darker regions had higher EGFC and lighter regions had lower EGFC, normalized per rhythm. Though electrodes 1 and 8 in all splines tended to localize near regions with relatively low EGFC in both rhythms, the EGFC distributions otherwise did not align. Results were comparable among all individual patients and mapping positions. c shows a scatter plot of all positionally-aligned electrodes with good contact from patients with maps in both rhythms. No notable trends were observed

3.4 Local and global EGFC approximates high-density bipolar voltage mapping

High-density bipolar voltage maps were compared with EGF maps’ calculation of EGFC. All 10 patients received voltage mapping in SR while only 5 received voltage mapping in AF due to non-inducibility or procedure time concerns. The percentage of endocardial surface area with high bipolar voltage (> 0.5 mV) was greater in SR than in AF (64.1% ± 7.7 v 57.1% ± 9.5%, P = 0.045). Additionally, there was a trend towards a higher percentage of endocardial tissue surface area containing voltage > 0.5 mV in the RA compared to the LA, although this was not statistically significant (65.3% ± 8.1% v. 60.1% ± 8.5%, P = 0.096).

Visually, bipolar voltages appeared to align with EGFC in both AF and SR. Following the splines of the basket catheter across the EGF maps and voltage maps revealed similar trends in EGFC and bipolar voltage at the same locations, as detailed in the example case in Fig. 5. In AF, areas of high EGFC can be seen at the first few electrodes of the A, B, G, and H splines and all but the last few electrodes of the C and D splines, aligning with where the splines run through high voltage regions of the corresponding voltage maps. In SR, splines A-D predominantly pass through high EGFC areas, and splines E–H predominantly pass through low EGFC areas, again correlating with the trends in the local voltages. Areas of EGF maps with low contact of basket electrodes generally corresponded to low-voltage areas or areas that also could not be mapped by bipolar voltage mapping.

Fig. 5

EGF maps in AF and SR with corresponding basket positions and overlaid voltage map taken from the same rhythm. Moving clockwise, A spline is cyan, B spline is red, C and D splines are white, E spline is yellow, F spline is green, G spline is blue and H spline is pink. In AF, the A and B splines (view 1) start in a high voltage region (purple) for their first two electrodes and then move to low voltage (gray). The C and D splines descend and stay in a high voltage region for their first 6 electrodes (view 1) before electrodes 7 and 8 ascend into an unmapped (tan) region (view 3). By contrast, the E and F splines (view 2) predominantly pass through low-voltage areas except locally at F3, and the G and H splines (view 1) start in high voltage regions before descending into low-voltage and unmapped areas (view 3). In SR, splines A–D (views 1–2) predominantly pass through high EGFC regions, even at A4–A5 where the spline passes through the left-sided PVI lesion (red spheres). At electrodes 7–8, the splines ascend into a low-voltage region (view 3). By contrast, the E and F splines (view 2–3) mainly pass through low voltage except for a high voltage patch at E3/F3-E4/F5 (view 3). Similarly, the G spline (view 3) passes through low voltage with possible exception near the aforementioned region extending to G3–G4, and the H spline (view 3) passes entirely through low voltage except at H1–H2 (view 1). For both rhythms, these trends generally align with the high EGFC (purple) v. low EGFC (blue) regions of the corresponding EGF map. It should also be noted that the source near G1 converted into a sink near G2 during the rhythm change

These local effects were seen universally across all voltage and EGF maps. Though no deviations were observed, high-resolution local deflections observed in EGF maps commonly appeared smoothed over in voltage maps such that the former were patchier. However, because the basket catheter electrodes could not be computationally aligned with the voltage map coordinates, quantitative analysis of this local relationship was not performed.

To further quantitatively assess the relationship between EGFC and voltage, EGFC was first examined locally and compared to unipolar voltage. The first recordings in each atrium were identified, and the median unipolar voltage measured from the f-wave amplitude at each basket electrode was plotted against the EGFC at the same electrode. As shown in Fig. 6a, EGFC increased linearly with increasing unipolar voltage (r = 0.389, P < 0.0001).

Fig. 6

EGFC versus Voltage. a Unipolar voltage was measured from the median f-wave amplitude at each electrode and plotted against local EGFC at the same electrode for the first recording in each atrium from each patient. SR is shown in blue v. AF in orange. An overall linear trendline was fit to the data (r = 0.389, P < 0.0001). b All voltage maps had their healthy voltage percent surface area plotted against the EGFC mean computed from the average of all concomitant EGFC maps in the same patient and rhythm. SR is shown in blue v. AF in orange; LA is shown as circles v. RA as diamonds. An overall linear trendline was fit to the data (r = 0.651, P < 0.0001)

Next, bipolar voltage from the voltage maps was compared to global EGFC. The percentage of high voltage surface area across each atrium in each rhythm was plotted against the mean EGFC of the concomitant EGF maps recorded from the same atria in the same rhythms. As shown in Fig. 6b, healthy voltage percent increased linearly with EGFC (r = 0.651, P < 0.0001) in the analyzed zone of healthy voltage percentages between 45 and 75% and EGFCs between 0.4 and 1.2. This relationship was maintained across SR v. AF and RA v. LA.