Resumen de Rotor detection in atrial fibrillation
Atrial fibrillation (AF) is one of the most common arrhythmias in the clinical practice. However the initiation and maintenance mechanisms responsible for this heart rhythm disorder remain poorly understood. In this sense, many theories have been proposed relying on experimental observations and theoretical assumptions. The ectopic firing from specific areas of the heart, multiple wavefronts propagating at random, or the spatio-temporal re-entrant waves (rotors) have been proposed as possible initiating and sustaining mechanisms. Most of the clinical approaches to palliate AF rely on the catheter ablation method, which was employed more than 20 years ago as an approach to terminate this rhythm disorder by isolating the pulmonary veins from the atrium. Since its outbreak, this technique obtained international acceptance among clinicians, and technological advances in this field increased its safety while reducing the procedure duration. However, there is no perfect AF treatment procedure described yet, since the understanding of the driving and sustaining AF mechanisms remains poor. Pulmonary vein isolation prevails as the most common ablation strategy given its proven clinical utility.
Focusing on the possible drivers, spatiotemporal stable sources (rotors) have been proposed as the maintenance mechanism for AF. The most representative characteristic of a rotor is the re-entry spiral-like propagation pattern that the electrical wavefront exhibits when it propagates in the atrial tissue. The assessment of its presence and posterior ablation of the sites where rotors anchor might improve the success of AF ablation.
Under these circumstances, technical solutions emerged focusing on the rotor assessment problem. They base their methods on the reconstruction of the atrial activity using multi-electrode catheters and phase maps, in which they detect singularity points, the sites where rotors spin. The ablation of these sites showed promising results, but the difficulty to reproduce the results by other authors increased the controversy on this technique. In this thesis we address the rotor detection problem in the time domain as opposed to current methods based on the phase domain of the signals.
We developed a new method to identify local activation times (LATs) in unipolar electrograms (EGMs) recorded with multi-electrode catheters. We propose a new filtering scheme to enhance the activation component of the EGM while considerably reducing the presence of noise in the signal. This signal processing method reflects the real activity of the tissue in contact with the electrode. It opposes the Hilbert transform (HT) used to extract the phase component of the signal, which does not correlate well with the temporal activations. With the EGM LAT we perform a spatial interpolation translating the electrode positions of the catheter into a regular 2D grid. This way we generate isochronal maps revealing the electrical wavefronts in the atrium. This step guarantees compatibility with multi-electrode catheters, not restricting the method to specific models or topologies.
For the local activation maps reconstructed from the electrode EGMs, we developed a new rotor detection algorithm based on the optical flow of the wavefront dynamics, and a rotation pattern match. The method calculates the propagation direction of every node in the 2D grid by computing the vectorial gradient of the direction of the wavefront using consecutive isochronal map frames. Using a novel approach based on the comparison of the obtained vectorial map with respect to a predefined rotor mask, we quantify the rotational degree of the wavefront. Then, we use a temporal sliding window to capture the dynamics of the rotational activity, and we detect its presence with a double threshold which discerns the flowing direction of the gyre, i.e., clockwise if the gyre positively matches the predefined mask, or counter-clockwise if the wavefront negatively matches the mask.
Additionally we developed a new method based on Granger's causality to estimate the directionality of the wavefronts, which serves as an additional indicator for rotational patterns. This approach does not need so many filtering and pre-processing steps as other methods in the literature do, and can be also adapted to any kind of catheter topology. We validate all the new methods using in silico and real AF signals, and we also compare the outcomes of other existing approaches to offer a fair comparison while analyzing their main differences.
As for the technical contribution of this thesis, we implemented these methods into a system that assesses the presence of rotational activity in the atrium. Our system is able to operate in real-time with multi-electrode catheters of different topologies in contact with the atrial wall. We integrate signal acquisition and processing in our system, which allows direct acquisition of the signals without requiring signal exportation from a recording device and additional post-processing, which delay the clinical procedure. We address the computational time handicap by designing parallelizable signal processing steps. We employ multi-core processors and GPU based code to distribute the computations and minimize the processing times, achieving near real-time results. We use the system's detected rotor sites information to spatially characterize their distribution and voltage in the left atrium.
The results presented in this thesis provide a new technical solution to detect the presence of rotational activity (rotors) in AF patients in real-time. Although the presence of rotational activity is itself controversial, we individually validate each of the steps of the procedure and obtain evidence of the presence of rotational activity in AF patients. The system has been also found useful to characterize the atrial sites where rotational activity was found in terms of spatial and voltage distribution. The results of this thesis provide a new alternative to existing methods based on phase analysis and open a new research line in the detection of the mechanisms sustaining AF.
