Técnicas de processamento não linear de sinais aplicadas à investigação de limitações práticas do procedimento de ventilação mecânica não invasiva

AUTOR(ES)
FONTE

IBICT - Instituto Brasileiro de Informação em Ciência e Tecnologia

DATA DE PUBLICAÇÃO

06/05/2011

RESUMO

Individuals with respiratory system dysfunctions or who are recovering from complex surgeries may have difficulty to sustain spontaneous breathing. The mechanical ventilation is a clinical procedure that supports ventilation until the patient is recovered. Noninvasive mechanical ventilation is especially remarkable as it can be used outside the hospital, in home care. This work investigates the ventilatory dynamics during noninvasive ventilation administration using nonlinear signal processing techniques. Leakage is inherent in noninvasive ventilation when the mask used to supply ventilation has an opening to avoid CO2 reinhalation. This intentional kind of leakage may be taken into account after an experimental characterization of the mask leakage conductance. The face-mask interface and the mouth in nasal mask ventilation are alternative routes to air leakage. The non intentional leakage may be a hindrance to the ventilatory goals specified to the patient. It may also impact the ventilators performance as (i) trigger and cycling criteria are usually defined regarding the patient airflow; and (ii) it is harder to keep an appropriate pressurization at the mask. In this work, the extended Kalman filter integrates a model to the respiratory mechanics with measurements of mask pressure and airflow at the ventilatory circuit to estimate leakage. The Kalman filter based approach is compared with another method of leakage estimation considering a patient-ventilator interaction profile. System identification allows to derive mathematical models for dynamical systems described by data sequences. This technique is especially useful when the knowledge available about system behavior is limited. Pressure and airflow signals acquired during sessions of noninvasive ventilation are analyzed using system identification. The underlying dynamics is represented by NARMAX polynomial and RBF network models in both input-output and autonomous configurations. These models are initially validated in free-run simulation. After that, models are also evaluated by comparing static characteristics, performance indexes in free-run simulation and model output synchronization to validation data. An initial topological characterization for an autonomous RBF model is also presented. Data sequences derived from polysomnographic signals are processed using tools developed in the scope of nonlinear systems theory. The tools used in this work are the symbolic analysis, joint recurrence plots and discrete Markov models. These tools may be integrated to compose a wider description of the underlying dynamics. The employed analysis methodology focuses on pattern identification and on the organization of such patterns along time. So, it does not depend on specificities of patient clinical status. The joint recurrence plots are a promising tool to analyze signals from nocturnal noninvasive ventilation. Quantifiers for the diagonal and vertical lines distributions may discriminate opposite profiles of ventilation and sleep.

ASSUNTO(S)

engenharia elétrica teses.

Documentos Relacionados