Thèse soutenue

Extraction et analyse de l'activité auriculaire pendant épisodes de fibrillation auriculaire

FR  |  
EN
Auteur / Autrice : Pietro Bonizzi
Direction : Olivier MesteVicente Zarzoso
Type : Thèse de doctorat
Discipline(s) : Automatique, traitement du signal et des images
Date : Soutenance en 2010
Etablissement(s) : Nice

Résumé

FR  |  
EN

Atrial fibrillation (AF) as it affects up to 10% of people over 70 years of age is the most common cardiac arrhythmias encountered in clinical practice, in spite of its relevance and incidence, the mechanisms of initiation and maintenance of AF are still quite unknown. Different strategies for AF treatment are selected with respect to the duration of AF episodes, and their efficacy may also be influenced by the degree of organization in the atrial activity (AA). The degree of organization of the AA depends in turn on the chronification of AF, and on the consequent electro-structural remodeling concerning the myocardial substrate, affecting the functioning of the atrio-ventricular node in particular. Thus, proper signal processing tools are required in order to shed some light on the electrophysiological origins of AF and on its impairing influence on the cardiac system. Particularly, the signal processing interest relies in extracting as much information as possible from non invasive recordings, in line with the general tendency in the clinical domain, in order to reduce the risks to the patient and to make clinical analysis time and cost effective. In this sense, a certain knowledge of the degree of organization in the AA may be potentially relevant in clinical decision making, as this could guide the selection of the best treatment for AF for each patient. Classical methods proposed for the extraction of an AA signal from electrocardiogram (ECG) recordings and for the non invasive estimation of the degree of organization of the atrial activations during AF do not exploit completely the spatial diversity offered by multi-lead ECG recordings. They generally focus on the analysis of the spectral content of the AF on a single lead only, with the risk of underestimating the actual complexity of the inner atrial activations. In this doctoral thesis, we exploit the spatial diversity offered by multi-lead ECG recordings to accomplish two main tasks. First, we want to enhance the quality of an AA signal extracted from ECG recordings, necessary for further detailed analysis of AF. To this end, we exploit the spatial information of the ECG to generate suitable subspaces representing each of the main cardiac activities of interest, the ventricular and the atrial, respectively, by delineating the corresponding segments in the ECG recording. These subspaces are exploited as a priori information and inserted as additional constraints into the blind source extraction algorithm. Different possibilities to exploit these subspaces as a prior information are presented, underlining their versatility in satisfactorily focusing on different characteristics of the various cardiac activities and of their relationships. Second, we want to noninvasively quantify the degree of spatio-temporal organization of the atrial activations during AF from the analysis of multi-lead ECG recordings. This is achieved looking at the spatial complexity of the recorded atrial electrical activity, prop-1 erly segmented from the ECG recordings, and the temporal stationarity of its potential field spatial pattern. As for the extraction of the AA signal, spatial complexity and temporal stationarity of the AA are measured exploiting an estimate of its subspace. The results of our study confirm the interest of using the spatial information in the ECG in order to generate different subspaces suitably describing the ventricular and atrial components of the ECG. In turn, these components reveal to be useful both to define additional constraints into the blind source extraction algorithm for the AA signal extraction and to directly analyze the AF organization in surface recordings, strongly supporting the appropriateness of signal processing approaches exploiting spatial diver-sity in AF analysis. First applications of these techniques to study the effects of catheter ablation on the reorganization of the AF by exploiting standard 12-lead ECG recordings attest their potential clinical relevance in the selection of patients who may actually benefit from the ablation therapy and suggest their widespread use in future clinical applications.