Can sequentially collected electrograms be effectively used for dominant frequency mapping during persis-tent AF?

Xin Li1, Charlie Hugill1, Gavin Chu1, Mahmoud Ehnesh2, Tiago Paggi de Almeida1, Bharat Sidhu1, Ibrahim Anton1, Ahmed Kotb1, Peter Stafford3, G. Andre Ng1, Fernando Schlindwein1
1University of Leicester, 2University of Leicester. Department of Engineering, 3University Hospitals of Leicester NHS Trust


Sequential electrogram collection using multipolar catheters (Pantaray, HD Grid, etc.) is increasingly used in persistent AF (persAF) mapping. How-ever, it is unknow whether sequential mapping is suitable for dominant fre-quency (DF) mapping, as DF tends to be spatiotemporally unstable. We aim to model and compare simultaneously- and sequentially- collected EGMs for DF mapping. 10 persAF patients undergoing left atrial (LA) ablation were enrolled. 2048-channel virtual EGMs (EnSite Array, Abbott; 5 mins) were analysed. After QRST subtraction, fast Fourier transform was used to estimate DF with 4-s sliding windows (2-s overlap). LA meshes were automatically segmented into 20 random-selected captures (Fig. A) to model multipolar catheter cap-tures. Sequential maps were generated altering time delay (gap) between cap-tures and compared with the simultaneous map. Correlation coefficients (CCs) and absolute difference between maps were calculated (Fig C i-ii). Av-erage DF map over 5 mins were generated as ‘gold standard‘(Fig. D) , aver-age DF map using shorter time duration ranging from 0 to 5 mins for all pa-tients were compared with the ‘gold standard‘ map (Fig. C iii). Similarity between simultaneous and sequential maps was low (CC: 0.13 ± 0.12) with significant differences between capture delays (p<0.05, Fig. C i). A considerable DF difference at each node (0.61 ± 0.22 Hz) was found (Fig. C ii). As expected, the CC was correlated with increased time for DF map calculation, while DF difference were inverse-correlated (Fig. C iii). Using data for 84 s, CC of 0.91 ±0.06 and DF error of 0.065±0.023 Hz were achieved compared with full-length data (5 mins). This was confirmed visually in Fig. D with fully reproducible 84-s maps. Sequentially collected DF maps with short duration (4 seconds) captures generate distinct results from simultaneously collected maps. A duration of 84 seconds per capture is required to achieve reproducible mapping.