Spatial correlation of left atrial low voltage substrate in sinus rhythm versus atrial fibrillation: The rhythm specificity of atrial low voltage substrate

Improved sinus rhythm (SR) maintenance rates have been achieved in patients with persistent atrial fibrillation (AF) undergoing pulmonary vein isolation plus additional ablation of low voltage substrate (LVS) during SR. However, voltage mapping during SR may be hindered in persistent and long‐persistent AF patients by immediate AF recurrence after electrical cardioversion. We assess correlations between LVS extent and location during SR and AF, aiming to identify regional voltage thresholds for rhythm‐independent delineation/detection of LVS areas. (1) Identification of voltage dissimilarities between mapping in SR and AF. (2) Identification of regional voltage thresholds that improve cross‐rhythm substrate detection. (3) Comparison of LVS between SR and native versus induced AF.

Conclusion: Although the proposed region-specific voltage thresholds during AF improve the consistency of LVS identification as determined during SR, the concordance in LVS between SR and AF remains moderate, with larger LVS detection during AF. Voltage-based substrate ablation should preferentially be performed during SR to limit the amount of ablated atrial myocardium.

K E Y W O R D S
atrial fibrillation, induced AF, low voltage substrate, regional analysis, sinus rhythm

| INTRODUCTION
Atrial fibrillation (AF) is the most common sustained cardiac arrhythmia associated with an increased risk for stroke and heart failure. 1 The pulmonary veins are the primary trigger site of AF. Isolating the pulmonary veins can yield a high rate of arrhythmia freedom in paroxysmal AF patients. 2 However, the success rate is often lower in persistent AF patients due to atrial remodeling and additional pathological substrate contributing to arrhythmia maintenance. 3,4 Electroanatomical mapping to identify low bipolar voltages (peak-to-peak amplitudes) <0.5 or 1 mV during sinus rhythm (SR) has been shown to be a promising technique to identify the additional pathological substrate. 3,[5][6][7] However, mapping in SR is not always feasible: for example, when AF reoccurs shortly after electrical cardioversion due to recurrent/sustained fibrillatory trigger activity.
Moreover, electrophysiologists may choose to perform mapping during AF to identify both the potential arrhythmogenic rapid trigger sites and the underlying pathological substrate. 8,9 Two widely accepted voltage cutoff values have been reported in SR that allow the identification of potentially proarrhythmogenic tissue: <1 10 and <0.5 mV. 5,7 When mapping is done in AF, atrial areas displaying LVS < 0.5 mV have been reported as potential arrhythmogenic sites. 3 Uncertainties remain regarding which cutoff values should be applied when mapping during AF and how the voltages in both rhythms relate to one another. The current study aimed to compare LVS in SR and AF and identify regional voltage thresholds to improve cross-rhythm substrate detection.

| Patient cohort
Forty-one patients with persistent AF presenting for their first AF ablation procedure were included in the study. Six weeks before AF ablation procedure, all patients were electrically cardioverted to SR to enable favorable reverse electrical remodeling. 11  and all patients provided written informed consent before enrollment.
To avoid including information from points with poor contact, measurements were disregarded if the electrodes were located >6 mm from the atrial surface.
Bandpass filtering at 16-500 Hz was applied to the bipolar electrograms. To calculate the voltage in SR and AF, a window of interest was chosen restricted to the PR interval in the electrocardiogram.
Voltage values between the electrode positions were interpolated by the CARTO-3 system. Cutoff values of <0.5 and <1.0 mV were then applied to the bipolar SR voltage maps to define the low voltage substrate (LVS). 3,6 Areas demonstrating LVS were confirmed using a separate contact force-sensing mapping catheter with a contact threshold of >5 g.

| Analysis
Using the Scalismo statistical shape modeling software, 12 the geometries of each patient were aligned and registered to a mean left atrial (LA) geometry. This allowed the transfer of voltage information from each patient's geometry to a common geometry represented by the same number of surface points, which represent the same anatomical landmarks.
The voltage value for each point was calculated as the mean amplitude of all points within a 1.5 mm radius to compare local areas between the two rhythms. To investigate the correlation between SR and AF, receiver operation curves (ROCs) were created across the whole patient cohort, with the SR map being considered as the reference condition. The optimal AF thresholds for both SR cutoff values (0.5 and 1 mV) were identified and the sensitivity, specificity, and accuracy were computed for each The atrium was split into anatomical regions to examine the difference between SR and AF voltage mapping and identify optimal voltage thresholds for different atrial regions: inferior wall, lateral wall, posterior wall, anterior wall, and roof. The optimal threshold for each region was identified as the top left-hand corner of the ROC curve when comparing the identification of low voltage regions between rhythms in that region.
Finally, the patient cohort was split into two groups: (1) where induced AF was mapped (patients presenting in SR) and (2) where native AF was mapped (patients presenting in AF). The two patients where the coronary sinus (CS) was paced to maintain SR were removed from this part of the study due to the unknown effect of how different fast pacing from the CS relates to AF. A two-sample t test was performed to investigate if the voltage values between the two groups were significantly different. The two groups were then compared to their respective SR maps and ROC curves were computed. Since more patients were mapped with native AF (65%), a leave-p-out (p = 11) cross-validation was performed for the native ROC curve. A two-sample t test was additionally performed between the distance of the two ROC curves with respect to the top left-hand corner.

| Patient characteristics
Forty-one persistent AF patients (63 ± 11.1 years old, 43.9% female) presenting for their first AF ablation procedure were included. On the procedure date, 30 of 41 (73%) patients recurred with persistent AF. Table 1 describes further details regarding the patients' characteristics.

| Global AF thresholds for the detection of SR-LVS
ROC curve analysis provided the optimal threshold in AF for identifying LVS in SR. AF thresholds 0.34 and 0.45 mV for SR <0.5 and <1 mV provided the best balance between high sensitivity and specificity as identified by the top left-hand corner of the ROC curve ( Figure 3A). The percentage of concordance was moderate with a sensitivity of 67% (<0.5 mV) and 66% (<1 mV), specificity of 69% (<0.5 mV) and 68% (<1 mV), and accuracy of 69% (<0.5 mV) and 65% (<1 mV). The performance of the new AF voltage thresholds varied between patients with per-patient accuracy ranging between 53% and 94% (mean 69 ± 11% for SR < 0.5 mV, mean 67 ± 11% for SR < 1 mV) ( Figure 3B). In Supporting Information Material: Figure S2, the ROC curve identifying the optimal threshold in SR for AF <0.5 mV is shown.

| Regional AF thresholds for the detection of SR-LVS
The lowest concordance between SR and AF using the global optimal threshold for the entire atrium was found at the inferior wall (accuracy: 62%) Figure 5. By applying a regional AF threshold (0.3 mV), the accuracy on the inferior wall increased by 4%. For the posterior wall, an even lower AF threshold (0.27 mV) increased the accuracy from 69% to 76%. In both regions, the new regional thresholds decreased the sensitivity and increased the specificity.
From Figure 2, it can be seen that the voltage values are markedly higher on the posterior/inferior wall in SR than in AF. The entire atrium threshold is therefore too sensitive for these regions. On the other hand, on the anterior and lateral walls, a slightly higher threshold (0.36 and 0.39 mV, respectively) can optimally locate the regions of SR-LVS using the AF voltage map. The optimal AF regional thresholds, which correspond to SR < 1 mV are shown in the Supporting Information Material ( Figure S3). The ROC curves used to find the optimal regional thresholds are shown in Supporting Information: Figure S4. F I G U R E 4 Agreement map between SR and AF. Each map shows the percentage of patients in which both voltage maps in SR and AF agree that low or high voltage is located within a local neighborhood. The yellow box shows the agreement with SR < 0.5 mV, and the light blue box with SR < 1 mV. In each box, the left column uses the optimal AF voltage thresholds as identified by the ROC curve analysis ( Figure 3A). AF, atrial fibrillation; ROC, receiver operating characteristic; SR, sinus rhythm.

| Impact of inducing AF
The voltage was slightly higher in native AF patients than in patients in whom AF was induced ( Figure 6A, not significant). The ROC curves ( Figure 6B

| Main findings
This study investigated the differences in LVS identification for mapping during SR and AF. Three key findings can be reported: 1. The overall correspondence of LVS mapped in SR and in AF is moderate.
F I G U R E 5 Optimal AF threshold and the corresponding sensitivity, specificity, and accuracy for each anatomical region of the LA compared to SR < 0.5 mV. The previously defined global threshold for the entire atria is shown by the purple asterisk while the green asterisk represents the regional threshold. AF, atrial fibrillation; LA, left atrial; SR, sinus rhythm.

Discrepancies exist between mapping in SR and AF, specifically on
the posterior and inferior LA walls.
3. New regional AF cutoff values improve cross-rhythm substrate detection.

The concordance of SR and AF voltage maps is higher when AF
was induced compared to native AF.
This study addresses whether the same LVS sites can be identified irrespective of rhythm by adaptations of the thresholds.
However, to enable both a sensitive and specific detection of trigger sites for AF, additional markers for arrhythmogenesis, besides low voltage areas and late gadolinium-enhancement (LGE) areas (e.g., rapid repetitive activity in AF or atrial late potentials in SR 9 ) need to be considered.

| AF cutoff values for identifying LVS
A recent study using generalized additive models in a cohort of 31 patients found that a cut-off value of 0.31 mV in AF was best for predicting <0.5 mV SR-LVS. 13

| Differences in SR and AF voltage mapping
While the voltage is typically lower in AF than SR, this study identified that this difference is not uniform across the entire atrium.
The differences between rhythms were found to be much higher on the posterior and inferior LA wall (difference typically >0.55 mV) than on the anterior (difference typically <0.55 mV). Kurata et al. also reported higher voltages in the posterior region than the anterior region in both patients with and without low voltage areas. 15

| Influence of AF induction on detected LVS
The correlation between SR and AF voltage mapping was higher when AF was induced in the patients. One hypothesis is that native AF is on average more complex, with high levels of electrical remodeling, endoepicardial dissociations of wavelet activities, and more wavefronts approaching from multiple directions than in induced AF, which is mapped few minutes after its initiation. 19 Alternatively, patients who presented with AF on the procedure day

| LIMITATIONS
In some patients, AF was mapped first before cardioverting to SR, potentially affecting the results by undetected map shifts. To counteract this, all patients' voltage information was mapped to a joint geometry, and analysis points comprised of the mean voltage in a 1.5 mm radius. The size of the subcohort in which AF was induced is comparatively small (n = 11). Additionally, SR and AF LVS were not correlated to imaging modalities such as MRI, as the aim of this paper is not to provide a tool for AF-guided ablation but to identify if similar low-voltage sites can be found irrespective of rhythm. Further studies should be performed to identify the importance of the newly proposed thresholds for LVS ablation.

| CONCLUSION
The extent and distribution of LVS are different in SR and AF. The proposed AF thresholds improve the identification of SR-LVS when mapping is performed during AF. However, a global threshold for the entire atria can lead to over-or underestimation of LVS, which can be corrected only to some extent by applying the reported regional thresholds. When mapping in AF may be necessary for patients who cannot be cardioverted or maintained in SR or when AF mapping for detection of rapid activity sites is chosen, further electrogram or activation characteristics might be useful to localize the arrhythmogenic substrate.

ACKNOWLEDGMENTS
The authors gratefully acknowledge financial support by Deutsche

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by the Freiburg University Hospital ethics committee.
The patients provided written informed consent to participate in this study.