A computational based approach for non-invasive localization of atrial ectopic foci

• Autores: Eduardo Jorge Godoy
• Directores de la Tesis: Rafael Sebastián Aguilar (dir. tes.) , Miguel Lozano Ibáñez (codir. tes.)
• Lectura: En la Universitat de València ( España ) en 2020
• Idioma: español
• Tribunal Calificador de la Tesis: Pablo Laguna Lasaosa (presid.) , Francisco Martínez Gil (secret.) , Beatriz Ana Trenor Gomis (voc.)
• Texto completo no disponible (Saber más ...)
• Resumen

  • Atrial arrhythmias are the most common cardiac arrhythmia, affecting six million people in Europe and imposing a huge healthcare bur- den on society. New technologies are helping electrophysiologists to tailor the treatment to each patient in different ways. For instance, magnetic resonance imaging (MRI) allows to assess the spatial distribution of atrial fibrosis; electro-anatomical maps (EAM) permit to obtain an electrical char- acterization of tissue in real-time; electrocardiographic imaging (ECGI) al- lows to study cardiac electrical activity non-invasively; and radiofrequency ablation (RFA), allows to eliminate pathological tissue in the heart that is triggering or sustaining an arrhythmia. Despite the access to advanced technologies and well-developed clinical guidelines for the management of atrial arrhythmia, long-term treatment success rates remain low, due to the complexity of the disease. Therefore, there is a compelling need to improve clinical outcomes for the benefit of patients and the healthcare system.

    Detailed biophysical models of the atria and torso could be employed to integrate all the patient data into a single reference 3D model able to re- produce the complex electrical activation patterns observed in experiments and clinics. However, there are some limitations related to the difficulty of building such models for each patient, or performing a substantial number of simulations to plan the optimal RFA therapy. Considering all those lim- itations, we propose to use detailed biophysical models and simulations as a tool to train machine learning systems, for which we have all the ground- truth data which would be impossible to obtain in a real clinical setting.

    Therefore, we can perform hundreds of electrophysiology simulations, con- sidering a variety of common scenarios and pathologies, and train a system that should be able to recognize them from a limited set of non-invasive pa- tient data, such as an electrocardiogram (ECG), or a body surface potential map (BSPM).