PPGEBM PÓS-GRADUAÇÃO EM ENGENHARIA BIOMÉDICA FUNDAÇÃO UNIVERSIDADE FEDERAL DO ABC Telefone/Ramal: Não informado http://propg.ufabc.edu.br/ppgebm

Banca de DEFESA: ANDRÉ MORILHA DUARTE

Uma banca de DEFESA de MESTRADO foi cadastrada pelo programa.
DISCENTE : ANDRÉ MORILHA DUARTE
DATA : 11/01/2022
HORA: 09:00
LOCAL: Remota
TÍTULO:

Evaluation and Systematization of the Transfer Function Analysis and Autoregulation Index for Cerebral Autoregulation Assessment


PÁGINAS: 90
RESUMO:

Introduction: The cerebral autoregulation (CA) mechanism plays a crucial role in brain homeostasis, being responsible for maintaining constant cerebral blood flow (CBF) despite variations in arterial blood pressure (ABP) between the limits of 60 to 140 mmHg. Transfer function analysis (TFA) and Autoregulation index (ARI) are worldwide used techniques to characterize CA through interaction between ABP and CBF. TFA models the process of CA as a linear system in which spontaneous ABP variations (system input) reflect in the CBF speed (system output) to report gain, phase and coherence values. The TFA method uses spontaneous BP and CBF velocity (CBFv) response measurements to compute the transfer function of the system. The CA system in healthy subjects tends to attenuate large ABP variations to have stable CBFv values. ARI is a dimensionless index (0 to 9) for which a response of CBFv to a hypothetical impulse change in ABP is estimated using TFA of spontaneous fluctuations in ABP and CBFv. An ARI of 9 describes a system in which CBF returns quickly to baseline levels after step-changes in ABP, whereas an ARI value of 0 describes a system in which there is no compensatory change in CBF, indicating completely impaired CA. ARI estimation is somehow dependent on TFA. Previous studies have assessed how different parameter settings in the TFA analysis have distorted its outcome, in order to understand the influence of each parameter variation.  Currently, researchers at the Cerebral Autoregulation Network (CARNet) have concentrated efforts to standardize the TFA parameters used to quantify CA. In this context, this study aims to analyze the impact CA assessment through ARI by varying the TFA model parameters, such window length, type of tapered window and percentage of overlap. 

 

Methods: ABP and bilateral CBFv of 12 healthy participants and 12 ischemic stroke patients were recorded for 5 minutes on baseline (Fs: 100 Hz; CAPPesq no.126713 HCFMUSP). Signals were uploaded and analyzed in the Cerebral Autoregulation Open Source (CAAos) platform, a new free software research tool, created by the researchers.  Signals were pre-processed including calibration and filtering. For each beat, detection of systole and end-diastole is carried out in both signals. The mean value of ABP and CBFv of each cardiac cycle was computed and associated to the time instant of the central sample of the cardiac cycle. The averaged values become the samples of the beat-to-beat ABP and CBFv signals. It is followed by a 5 Hz resampling through a cubic spline interpolation. Welch’s method is applied to calculate the ABP auto spectrum (Sxx) and cross spectrum of ABP and CBFv (Sxy). Transfer function H(f) is obtained by the division between (Sxx(f)/Sxy(f))*. The CBFv response to a step change in ABP was estimated from the inverse FFT of the H(f). The subject’s ARI index, from left and right cerebral hemispheres, was obtained by the minimum root mean square error (RMSE) fitting error of the corresponding Tiecks model responses using the first 6 seconds of the step response. In summary, this model uses a second-order differential equation to predict CBFv response, V(t), corresponding to a relative ABP modulation, dP(t), by the formula: V(t) = 1 + dP(t) − K × x2(t), where K represents a gain parameter, an x2 (t) is a state variable obtained based in three parameters gain (K), time constant and dampening factor. Ten combinations of these three parameters were proposed to represent a different value of ARI. The template curve corresponding to the minimum RMSE determines the corresponding value of ARI of the subject.  TFA parameters and ARI were obtained for each parameter changed window type (Rectangular, Hanning, Hamming and Tukey), window length (25, 50, 75, 100, 120) and overlap percentage (25%, 50% and 75%). Results were compared with the recommended literature (length 100s, Hanning window and 50% overlap). 

 

Results: Gain values over the frequency range for healthy participants were low, specifically for lower frequency ranges, phase presented higher values for very low frequency (VLF) and LF, decreasing its values for high frequency (HF). Coherence values were not much higher than 0.5 n.u. for all frequency ranges, demonstrating then that both ABP and CBFv signals in the TFA were not related. For Stroke patients, gain showed slightly higher values in all frequency ranges when compared with healthy participants, phase also showed higher values in the VLF range as previous literature results. The coherence showed higher values in both VLF and HF, suggesting that for stroke patients the ABP and CBFv signals are more similar than in healthy participants. Varying the window length during the Welch Method affected differently the TFA results for healthy and stroke patients. For gain, a higher number of differences in the statistical analysis were found among the different window lengths in healthy participants than in stroke participants, suggesting that in a healthy cerebral autoregulatory system varying the window length results in a wider range of results. The same differences concerning the type of window applied in the Welch Method was observed for coherence results, in which rectangular x Hanning and rectangular x Haming presented significant differences. Both stroke patients and healthy participants presented differences for gain results by varying the overlap in the Welch Method. For all window lengths the ARI results from the affected side showed lower results when compared with the corresponding window length in the healthy participants group. For healthy participants, the higher ARI value was observed for 100 s and 120 s windows, whereas for stroke participants it was for 75 s and 100 s windows. The window type did not seem to interfere in the difference between healthy and stroke patients, lower ARI values continued to be observed in the stroke patients’ group, even by using different windows in both groups. The overlap also did not affect the lower ARI values in the stroke patients’ group. The statistical analysis performed by the ANOVA test indicated the influence of using different window lengths during the Welch Method for both healthy participants and stroke group. In both cases it was found that only the 25 s windows differ from the other windows, the other lengths used to estimate the ARI did not show statistical differences at all. Neither varying the window type nor the overlap affected the ARI estimation, in other words, varying window type and overlap during the Welch Method in both groups did not result in a different ARI estimation.

 

Conclusions: Concerning the importance of estimating the better clinical parameter to quantify the CA, the TFA and ARI seems to perform well trying to establish the difference between a healthy participant and stroke patient. By changing the parameters, the TFA presented more differences in among the parameters combination when compared with the ARI, nevertheless the TFA comprehension is wider by quantifying the CA in three parameters, while the ARI does it in one.


MEMBROS DA BANCA:
Presidente - Interno ao Programa - 2231661 - JOAO LOURES SALINET JUNIOR
Membro Titular - Examinador(a) Externo à Instituição - MÁRCIA APARECIDA SILVA BISSACO - UMC
Membro Titular - Examinador(a) Externo à Instituição - João Paulo do Vale Madeiro
Membro Suplente - Examinador(a) Interno ao Programa - 1671296 - HARKI TANAKA
Membro Suplente - Examinador(a) Externo ao Programa - 1188948 - JOAO LAMEU DA SILVA JUNIOR
Membro Suplente - Examinador(a) Externo à Instituição - MATHEUS CARDOSO MORAES - UNIFESP
Notícia cadastrada em: 23/11/2021 14:17
SIGAA | UFABC - Núcleo de Tecnologia da Informação - ||||| | Copyright © 2006-2024 - UFRN - sigaa-2.ufabc.int.br.sigaa-2-prod