Correlation of Automated vs. Manual Measurement of Left Atrial Low Voltage Burden in Patients with Atrial Fibrillation

Introduction:
Low voltage areas (LVA) in the left atrium (LA), identified through electroanatomic endocardial ultra-high-density (UHD) mapping, serve as markers for atrial fibrosis and correlate with higher recurrence rates post-ablation. Currently, LVA quantification is unstandardized and prone to high interoperator variability. Automated algorithms for LVA quantification are not yet widely implemented, but hold potential to streamline and standardize LVA analysis.

Methods:
This single-center study evaluated consecutive patients from the prospective Bernau Ablation Registry who underwent LA UHD mapping and repeat ablation for AF/AT recurrence between 2016 and 2022. Anatomically complete LA maps were included, manually cropped at the pulmonary vein (PV) ostia, mitral valve annulus, and left atrial appendage. Within the remaining LA body, the extent of LVA was manually quantified (defined as bipolar peak-to-peak voltage of <0.5 mV with an extent of >1cm²) and correlated with the LA body surface area by a trained operator. All cropped maps were then processed in MATLAB, where LA body surface area and LVA were analyzed automatically across multiple voltage thresholds (<0.1 to <0.5 mV). For each vertex on the triangulated surface mesh, associated surface area was computed (1/3 of the area of adjacent triangles). The manual (<0.5 mV) and automated LVA percentages across different voltage thresholds as well as total LA body surface were compared.

Results:
A total of 205 patients (mean age 68 ± 8.7 years; 46% female; mean 1.4 prior ablations) were included. Baseline characteristics are provided in Table 1. Mean EGMs of the analyzed maps were 9705 ± 5908 and mean LA volume 178 ± 39 ml with 135 maps in sinus rhythm and 70 maps during AT. The correlation between manual and automated measurements of total LA surface area was high (r = .964, p < .001). Intraclass correlations for LVA percentage also showed high concordance between manual and automated analysis across all voltage thresholds, indicating reproducibility of both measurements (Table 2). Mean deviation from the manual LVA measurement varied across the automated cutoff values with the strongest correlation observed at <0.35 mV (Figure 1).

Conclusion:
Automated analysis of LA LVA is feasible, providing rapid, accurate results across various voltage thresholds. Though, manual and automated measurements at the same cutoff value of <0.5 mV differ around 20%, most likely due to very small scattered areas of LVA or contact issues not considered in the manual analysis. Mean voltage burden best matches at <0.35 mV, indicating this setting in the automated analysis may best align with manual evaluations. Automated measurements may facilitate routine evaluation of LA LVA burden in the clinical and scientific setting.