Análise não invasiva da complexidade da fibrilação atrial persistente durante a ablação por cateter: uma abordagem tensorial

• Ano de defesa: 2023
• Autor(a) principal: Abdalah, Lucas de Souza
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Palavras-chave em Português: Complexidade eletrofisiológica; Decomposição tensorial em termo de blocos; Eletrocardiograma; Fibrilação atrial; Tensor; Atrial fibrillation; Block term tensor decomposition; Electrocardiogram; Electrophysiological complexity; CNPQ::ENGENHARIAS::ENGENHARIA ELETRICA
• Link de acesso: http://repositorio.ufc.br/handle/riufc/78693

Atrial fibrillation (AF) is the most common cardiac arrhythmia encountered in clinical practice, and it is estimated to be responsible for one-fourth of cerebrovascular accidents. The persistent form of this arrhythmia is a complex case characterized by uncoordinated and irregular cardiac activation. Although catheter ablation is increasingly used due to its lower recurrence rate compared to other treatments, widely accepted intervention protocols by the cardiology community are still being established. Surface electrocardiogram (ECG)-based analysis is relevant due to its low cost and non-invasive nature. The development of new mathematical tools to characterize AF through ECG can improve intervention guidance and increase success rates, reducing the duration of arrhythmia and the risk of complications. However, despite the growing interest in these methodologies to evaluate the complexity of persistent AF signals, their performance is still limited. To overcome these limitations, this study proposes the application of tensor decomposition techniques to quantify the complexity of ECG signals during catheter ablation procedures for AF treatment. Tensor decompositions are powerful signal processing tools. However, their application in AF signal analysis is recent. The Constrained Alternating Group Lasso (CAGL) algorithm was developed to simultaneously calculate the block term tensor decomposition (BTD) into block terms and estimate its parameters (number and rank of the blocks), for the multilinear rank-(Lr,Lr,1) particular case. This algorithm showed promise in extracting atrial activity during AF and estimating its complexity. To evaluate its performance, we compare its results with the Nondipolar Component Index (NDI) method in the context of ablation, which demonstrates the multilinear advantages over matrix methods in non-invasively extracting and quantifying AF complexity. The tensor index correlates with the reduction in AF complexity throughout the ablation steps as instinctively expected. Additionally, it has presented a significant negative correlation with AF recurrence episodes, which presents clear clinical interest, since it can assist in the development of new medical protocols.