Atrial fibrillation (AF) is the most common encountered cardiac rhythm disorder (arrhythmia) in clinical practice. It is responsible for about one third of arrhythmia-related hospitalizations. This arrhythmia, which increases in prevalence with age, leads to severe complications and subsequently decreases the quality of life for the affected patients. Lifetime risks for developing AF are ~25% in subjects older than 40 years old. Currently, this arrhythmia is considered as a major public health concern. AF is a progressive disease, starting by short and rare episodes which further develop into longer and more frequent occurrences. When the arrhythmia becomes sustained for more than one year, it is labelled as long-standing persistent. AF advancement gives rise to an electrical of the atria (the upper chambers of the heart) resulting from abnormal high frequency atrial activations. The main goals of therapeutic management for patients with AF are to prevent severe complications associated with this arrhythmia, and ultimately to restore a normal rhythm. Currently, the cornerstone of non-pharmacological therapy is the radiofrequency catheter ablation of AF, which consists in delivering at strategic locations within the atria high-frequency electrical impulses. However, catheter ablation for patients with long-standing persistent AF involves extensive ablation of the atria and the success rate reported in various publications is associated with conflicting results. Over the last twenty years, an important effort has been made by the scientific community to develop signal processing algorithms to quantify the complexity of temporal or spectral characteristics of AF dynamics in terms of organization. As such, multiple approaches have been proposed to quantify AF organization either based on time-domain or frequency-domain analysis. All these methods shared one common goal: the development of organization indices which are interpretable from an electrophyisiological viewpoint. In the context of catheter ablation of patients with long-standing persistent AF, the success rate appears limited as the "classical" organization indices are not performant in assessing the amount of ablation required to achieve AF termination. Thus, there is a strong interest in predicting the procedural outcome from the surface electrocardiogram (ECG) recorded at baseline, i.e., prior to ablation. The main objective of this thesis was to derive novel organization indices from surface ECG and intracardiac signals acquired at baseline which could discriminate patients in whom AF was terminated from patients in whom AF persisted during catheter ablation within the left atrium. As the standard surface ECG is not appropriate for measuring the atrial activity, we aimed at adapting the placement of at least one ECG lead such that additional electrical information from the atria was provided. In our ECG signals study, we hypothesized that a quantification of the harmonic structure of AF signals brings more insight into AF complexity. Time-invariant and time-varying approaches were used to derive the ECG organization indices, and their performance for predicting the acute outcome of catheter ablation were compared. In the first scheme, the harmonic components of AF waves were extracted using linear time-invariant filters. In the second one, the components were extracted using an adaptive harmonic frequency tracking algorithm. [...]