Skip to Main content Skip to Navigation
Conference papers

Use of Normalized Correlation Function to Discriminate Outcome of Persistent Patients Undergoing Electrical Cardioversion

Abstract : Atrial activity (AA) during atrial fibrillation (AF) is a process characterized by different short-and long-term recurrent behaviors. In this work we hypothesize that the features derived from these behaviors contain an information on the lead locations that discriminate the most between persistent AF patients recurring after electrical car-dioversion from those who do not. Body surface potential maps (BSPMs, 184 electrodes) were recorded in 63 patients in persistent AF prior to electrical cardioversion (32 recurrences after 4-6 weeks). A correlation function (CF) was computed for each electrode, and normalized in order to make it independent of its magnitude. Finally, the first min and max values of the normalized CF from each electrode were used to discriminate patients outcome. The corresponding spatial maps showed that electrodes with the largest values of normalized CF are located on the front of the torso, centered around V1. A Wilcoxon rank-sum test was used to compare maps of recurrent and non-recurrent AF patients and find electrodes with significantly different magnitude. A significant difference was observed on the upper and lower parts of the torso, with higher values for the non-recurrent AF patients (higher recurrence of the underlying AA propagation patterns). When looking at the energy computed on each electrode, this parameter was not able to distinguish among the two groups. 1. Introduction Although under the scope of many studies and researches , the characterization of atrial fibrillation (AF) is still an open field, especially when the goal is the strati-fication for guiding AF therapy [1]. This guidance could be the targeting of the ablation sites by using information recorded non-invasively. In [3],[2], we intended to compute a recurrence AA signal based on a multidimensional autocorrelation function. Thus, novel insights in AA during AF have been proposed, in line with the underlying cardiac electrophysiology. However, these outcomes don't allow by nature an accurate spatial analysis, e.g., at the level of the recording electrode. In this study, we decompose the multidimensional correlation to introduce the role of individual correlation functions , i.e., for each electrode. Features are extracted from these functions and the corresponding spatial maps are compared. Results shows how relevant is this approach and paves the road to a non-invasive spatial discrimination of non-recurrent AF subjects from recurrent ones. Although this discrimination is not the ultimate goal, because it is based on spatial parameters it shows that relevant information can be extracted from each electrodes taken separately. 2. Methods 2.1. BSPM data and pre-processing BSPMs were recorded in 63 patients in persistent AF using a custom configuration of 184 leads with 120 anterior and 64 posterior leads (ActiveTwo BSM Panels Carbon Electrodes, Biosemi B.V., The Netherlands). All patients underwent electrical cardioversion. After a blank-ing period from 4 to 6 weeks after procedure, 32 patients showed AF recurrence (labeled R) and 31 remained under sinus rhythm (labeled NR). ECGs were sampled at 256Hz. A one-minute segment was selected for each subject, low-quality leads were excluded (low signal-to-noise ratio, poor electrode contact, motion artefacts), and Wilsons Central Terminal was subtracted in line with conventional ECG analysis. After band-pass filtering the signals between 1 and 100Hz (3rd order Chebyshev), QRST
Document type :
Conference papers
Complete list of metadata
Contributor : Olivier Meste <>
Submitted on : Saturday, November 14, 2020 - 5:18:55 PM
Last modification on : Thursday, November 26, 2020 - 3:41:33 PM
Long-term archiving on: : Monday, February 15, 2021 - 6:20:15 PM


Files produced by the author(s)


  • HAL Id : hal-02961805, version 1



Olivier Meste, Stef Zeemering, Joël Karel, Theo Lankveld, Ulrich Schotten, et al.. Use of Normalized Correlation Function to Discriminate Outcome of Persistent Patients Undergoing Electrical Cardioversion. Computing in Cardiology, Sep 2020, Rimini, Italy. ⟨hal-02961805⟩



Record views


Files downloads