Complexity Analysis of Physiological Time Series with Applications to Neonatal Sleep Electroencephalogram Signals

This thesis investigates the complexity in physiological time series with application to neonatal sleep electroencephalography (EEG) signals. Complexity analysis is applied to two clinical data sets of neonatal sleep Electroencephalography(EEG) time series, to uncover the evolution of signal dynamics and its relationship to neurodevelopment and maturation. A review of the advantages and disadvantages of various complexity measures is provided and it is determined that nonlinear dynamic analysis is complimentary to the traditional linear methods for EEG signal processing. Surrogate data analysis is used to test the nonlinearity structure in the signal. The complexity of the neonatal sleep EEG signals were further quantified by evaluating two complexity measures i.e. Approximate Entropy(ApEn) and Sample Entropy(SaEn). The suitability of ApEn and SaEn for moderate length data and their relative robustness to noise has made them the good candidate for analyzing EEG time series data. Parameter selection is of utmost importance in the computation of complexity measures, and this was addressed in the thesis by improving the process of determining the appropriate time delay. The time delay determination process was applied to both synthetic and real data; and incorporated into the computation of ApEn and SaEn. The two clinical data sets used in this study consist of both preterm and full-term neonates. The two data sets were collected with different cohorts, sampling rate and data collection hardware. The cohorts in one data set are all healthy while cohorts in the other one are either sick and healthy. Though the vast difference between the two data sets, the following conclusions are applicable to both cases: 1) Surrogate data test performed on both data sets shows evidence of non-linear structure;. 2) It further suggests the necessity of using nonunity time delay for the calculation of ApEn and SaEn; 3) ApEn and SaEn were shown to be effective in quantifying the temporal patterns in the dynamic process of neonatal sleep EEG signal.

Name thesis:degree_name

Doctor of Philosophy

Level thesis:degree_level

doctoral

Discipline thesis:degree_discipline

EECS - System and Control Engineering

Grantor dc:publisher

Case Western Reserve University School of Graduate Studies

Year dc:date

2013

Author dc:creator

• Li, Chang

Contributors dc:contributor

• Loparo, Kenneth

• Electrical Engineering
• Engineering
• Information Science
• Systems Science
• complexity analysis
• time series
• Approximate Entropy
• Sample Entropy
• nonlinear dynamic analysis

Statement dc:rights

• unrestricted
• This thesis or dissertation is protected by copyright: all rights reserved. It may not be copied or redistributed beyond the terms of applicable copyright laws.

Language dc:language

English

Repository record dc:identifier

http://rave.ohiolink.edu/etdc/view?acc_num=case1345657829

OAI identifier oai:identifier

oai:etd.ohiolink.edu:case1345657829