Sanromán Junquera, Margarita2015-04-082015-04-082014-06http://hdl.handle.net/10115/12942Tesis Doctoral leída en la Universidad Rey Juan Carlos de Madrid en 2014. Directores de la Tesis: Dr. José Luis Rojo Álvarez y Dra. Inmaculada Mora JiménezThe present Thesis addressed the proposal of advanced methods for the analysis of intracardiac electrograms (EGMs). EGMs are a valuable source of clinical and diagnostic information about arrhythmias in electrophysiological (EP) studies. During EP procedures, the electrical activity of the heart is examined in order to diagnose the arrhythmia mechanism and, if appropriate, treat it. The treatments include the implantation of life support devices, such as pacemaker and implantable cardioverter defibrillator (ICD), and the application of ablation therapy, which sears the diseased tissue by means of radiofrequency or intense cold. Besides, cardiac navigation systems (CNSs) are used in order to build electrical and anatomical maps (EAMs) which help in the arrhythmia diagnosis and treatments. Nowadays, both the evaluation of the origin and activation sequence of an arrhythmia and the generation of EAMs are made by heuristically sampling the cardiac chamber. However, the number and spatial localization of EGMs during the sampling process have not been formally established. In this context, this Thesis aims to deal with this spatio-temporal analysis in two clinical applications of interest: (1) the regionalization of the left ventricular tachycardia (LVT) exit site by using EGMs from ICDs; and (2) the estimation of the spatial sampling rate (SSR) to build accurate EAMs from EGMs recovered in CNSs. For this aim, a method for the extraction of the temporal variations of the electrical signal (EGM) over time was first proposed based on digital image processing techniques, in order to create a digital database of EGMs stored in ICD and printed in paper. This method was tested using recording printed by devices of two different manufacturers. The gold-standard digital signal and the one recovered from printouts were compared by means of three time synchronization methods. The regionalization of LVT exit site was tackled by using machine learning techniques in a supervised scheme for classification and regression. Waveform and features (times and voltages) from EGMs were used as input spaces. The best discrimination between regions was obtained for the septal and lateral half, and for the basal-lateral-superior octant. The SSR estimation was dealt with a methodology based on manifold harmonic analysis. The methodology included the representation of the EAM spectrum as a spectral density in Fourier analysis, the estimation of the cut-off frequency, and then, the estimation of SSR. In addition, this methodology was extended to meshes with a scalar field (electrical feature) measured at vertices of the mesh representing the cardiac chamber. SSR was estimated for the anatomy and EAMs (anatomy and features) of ventricles and atria. Between 65 and 80 samples were enough to reconstruct the anatomy of the cardiac chambers, whereas the SSR of EAMs was dependent on the arrhythmia mechanism. The use of advanced processing techniques for spatio-temporal analysis tailored to specific applications can be useful for improving the technological support to electrophysiologists performing EP studies for arrhythmia ablation.engTelecomunicacionesMedicinainfo:eu-repo/semantics/doctoralThesisinfo:eu-repo/semantics/openAccess3325 Tecnología de las Telecomunicaciones