Avaliação da frequência cardíaca fetal baseada em métricas não-lineares

• Ano de defesa: 2015
• Autor(a) principal: Zacarias, Henriques Mateus Joaquim
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Palavras-chave em Português: Teleinformática; Cardiotocografia; Frequência cardíaca; Entropia
• Link de acesso: http://www.repositorio.ufc.br/handle/riufc/21128

Gestation is a period of countless physiologic modifications to adapt the maternal organism to the fetal formation demands and development. The continuous monitoring of maternal and fetal biological signs through Cardiotocography (CTG) allows their evaluation as a source of aid to medical diagnosis. The present work presents a group of researches accomplished over fetal signals, starting from the CTG tests, aiming at evaluating the behavior of the Fetal Heart Rate (FHR), using non-linear tools, seeking the subsequent addition of the proposed metrics already developed multparametric system for evaluation of fetal health condition. The used non-linear metrics are the approximated entropy (ApEn), sample entropy (SampEn), multi-scale entropy (MSE) and the DFA (Detrended Fluctuation Analysis). The results were obtained using a database classified previously by the medical team of the Germany company TriumGmBH. The database, identified as CTG-A, presents 39 CTG record in antepartum phase, with gestation periods varying from 28 to 34 weeks, divided in 5 different classes: Normal, High variation, self-similar, deceleration and High sample windows. The entropy values were analyzed for two different sizes of samples windows, being the first with 1200 samples and the second with 2400 samples. Concerning DFA, the window widths were varied for the computing of the scaling exponents alpha1 and alpha2. First k=1:3 and k =4:6 and second k=1:5 and k=5:10, respectively. When evaluating the behavior of the applied entropies, it was observed that ApEn, either used in an isolated way, either used as a tool for MSE, presents the highest sensibility to the signal time variations and long term variability. When ApEn and SampEn, are compared both present a similar behavior, just differing in the intensities of the variations registered for detection of changes in the original signal. The process of adjust window width, can provide relevant results for the detection of the signal complexity decreasing of the signal when repetitive patterns occur through FHR signal, which can indicate pathologies. Finally, the DFA analysis allows to characterize the degree of self-similarity of the analyzed signal. According to the detained results all of the analyzed records present different levels of self-similarity, which can characterize a classification of the evaluated signal. The results allow to conclude that combine different approaches more efficient than apply each one individually and all of them can be considered as viable metrics for a better characterization of the fetal signals for aid to medical diagnosis.