Traitement et analyse du signal pour les arythmies ventriculaires

Abstract : Signal processing tools are increasingly present in the electrophysiologist' s daily practice. These tools have the potential to enhance the detection of small electrical anomalies, and to enable the analysis of complex arrhythmia. Our work focuses on ventricular arrhythmia, and more specifically on how signal processing tech­ niques can help usbetter understand these diseases. lt is made up of three parts,focusing on different topics: uni-dimensional signal analysis (ECG and endocardial electrograms), non-invasive mapping, and invasive contact mapping. Concerning uni-dimensional signal analysis, we first propose a method to enhance the signal to noise ratio of ECG recordings. We use a combination of signal averaging and respiration gating to achieve this goal, and offer interesting perspectives for the detection of abnormal low amplitude potentials and non-invasive measurement of the HV interval. We then analyze the relationship between endocardial and ECG signais during ventricular fibrilla­ tion (VF) episodes. We demonstrate that endocardial and ECG dominant frequencies are similar,and that higher endocardial fragmentation results in a drop of the waveform amplitude on the surface ECG. Finally, we demonstrate through frequency domain analysis of 63 VF episodes a clear correlation between VF characteristics and clinical factors. Dominant frequency is particularly useful to dis­ criminate between different underlying causal substrates. We also demonstrate that VF characteris­ tics depend on the induction mode and induction site, in terms of dominant frequency, amount of fragmentation and ECG phase. Concerning non-invasive mapping, we compare different potential-based inverse problem resolution techniques. ECGi appears as one of the most reliable techniques. A clinical validation study of non-invasive ventricular activation mapping using ECGi isthen carried­ out. Fifty-five patients were included for whom non-invasive maps are compared to contact maps. We show a very poor overall correlation between non-invasive and invasive maps. Results are het­ erogeneous, with good correlation in patients with wide QRS activation patterns. Wethen propose different techniques to improve non-invasive activation mapping. A first study uses the surface laplacian and the gradient of the inverse-computed potential as inputs to activation map­ ping. A second study combines estimated delays between neighboring points and local activation time estimates to create a more globally coherent solution. Both studies demonstrate a significant improvement of activation maps. Concerning contact mapping, we first give an overview of interpolation domains and techniques that can be used to provide dense activation maps from sparse measures. We illustrate the influence of these techniques on the clinician's ability to make a correct diagnosis. We then use one of these interpolation techniques to create epicardial activation maps in Brugada patients. We show that these patients harbor epicardial electrical activity compatible with partial endo-epicardial conduction block. This phenomenon predominates in the right ventricular lateral wall and outflow tract. Finally,using an in silico model, we demonstrate the arrhythmogenic potential of such a dissociation. lnducibility peaks as the number of residual functional connections between endo and epicardium falls. ÜVERALL, our work uses signal processing techniques for different applications conceming ven­ tricular arrhythmia. We propose different methodological innovations that allow us to record and process cardiac electrical activity with increasing precision. Further progress is still required before non-invasive mapping can live up to its promises. The proposed methodological innovations can extend the use of electrocardiography and invasive mapping.
Document type :
Complete list of metadatas

Cited literature [121 references]  Display  Hide  Download
Contributor : Abes Star <>
Submitted on : Friday, July 12, 2019 - 10:13:09 AM
Last modification on : Monday, July 15, 2019 - 11:15:34 AM


Version validated by the jury (STAR)


  • HAL Id : tel-02181464, version 1



Josselin Duchateau. Traitement et analyse du signal pour les arythmies ventriculaires. Médecine humaine et pathologie. Université de Bordeaux, 2018. Français. ⟨NNT : 2018BORD0465⟩. ⟨tel-02181464⟩



Record views


Files downloads