Analysis of P-wave Changes for Prediction of Atrial Fibrillation Episodes

Cristina Moreno1, Alba Martin2, Aleksei Savelev3, Pyotr Platonov4, Pablo Laguna1, Juan Pablo Martínez5
1Zaragoza University, 2Biomedical Signal Interpretation and Computational Simulation Group (BSICoS), Universidad de Zaragoza, 3St. Petersburg State University, 4Lund University,


Introduction: Non-invasive markers of predisposition to atrial fibrillation (AF) can help to early AF detection and treatment. We focused on the analysis of ECG signals in patients with paroxysmal AF. We hypothesized that, in sinus rhythm, the P wave widens, becomes more rugose and with higher complexity in the vicinity of an AF episode.

Methods: We applied principal and periodic component analysis (PCA and πCA) as lineal spatial transformations and beat averaging to characterize the P wave in 5-minute segments in sinus rhythm. These segments were selected at 60, 30 and 5 min before each AF episode, thus allowing to study the evolution of the P-wave features in the minutes previous to AF. The studied features were: power of the i-th transform component, PiΨ , relative power of the second-to-first component, P2Ψ,r , relative power of third-to-first-plus-second component, P3Ψ,r and a measure of rugosity: the power of the P wave after high-pass filtering with a 30 Hz cut-off frequency, PiΨ,HF, i∈{1, 2, 3}.Ψ∈{PCA, πCA}, as well as the P-wave duration in the first principal and periodic component DmnPCA.

Results: P2PCA,r increased as the AF episode approached. In fact it increased from 60 to 30 min (p= 0.009) and from 60 to 5 min (p= 0.002). Moreover DmnPCA increased from 60 to 5 min (from 112.8 to 116.8, p= 0.005) and DmnπCA also increased from 60 to 5 min (from 105.1 to 109.0, p= 0.011).

No significant differences were found in the power or relative power using πCA. In the high-frequency powers using either of the two methods no significant differences were found neither.

Conclusion: P-wave spatio-temporal features showed significant changes prior to the onset of AF episodes in PxAF, showing a more complex P-wave loop and enlarging its duration as AF episodes approached.