Respiratory Pattern Variability Analysis Based on Nonlinear Prediction Methods

Abstract

The traditional techniques of data analysis are often not sufficient to characterize the complex dynamics of respiration. In this study the respiratory pattern variability at different levels of pressure support ventilation (PSV) has been analyzed using nonlinear prediction methods. These methods use the volume signals generated by the respiratory system in order to construct a model of its dynamics, and then to estimate the deterministic level of the system from the quality of the predictions made with the model. Different methods of prediction evaluation and neighborhood definition have been considered. The incidence of different prediction depths and embedding dimensions have been analyzed. A group of 12 patients on weaning trials from mechanical ventilation have been studied at two different PSV levels. High statistically significant differences have been obtained when comparing the mean prediction error at two different PSV levels (p<O.002) with non-parametric analysis of variance test (Wilcoxon's signed rank test). The embedding dimension needed to model the system dynamics with low prediction error has also presented significant differences (p<O.005) between the complex dynamics of both PSV levels. Therefore, it may be concluded that the respiratory pattern variability depends on the level of pressure support ventilation.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Oct 25, 2001
Accession Number
ADA410559

Entities

People

  • B. F. Giraldo
  • D. Kaplan
  • L. Domingo
  • M. Vallverdu
  • P. Caminal
  • S. Benito

Tags

DTIC Thesaurus Topics

  • Algorithms
  • Computer Science
  • Data Acquisition
  • Data Analysis
  • Data Sets
  • Dynamics
  • Embedding
  • Errors
  • Frequency Domain
  • Information Science
  • Measurement
  • Parametric Analysis
  • Respiration
  • Respiratory System
  • Standards
  • Validation
  • Ventilation

Readers

  • Computational Modeling and Simulation
  • Neural Network Machine Learning.
  • Regression Analysis.