Signal and variability extraction for online monitoring in intensive care [Elektronische Ressource] / vorgelegt von Karen Schettlinger

Fakultät StatistikSignal and Variability Extractionfor Online Monitoring in Intensive CareDissertationzur Erlangung des akademischen Grades’Doktor der Naturwissenschaften’– Dr. rer. nat. –der Technischen Universität DortmundDer Fakultät Statistikder Technischen Universität Dortmundvorgelegt vonKaren Schettlingergeboren inNeustadt am RübenbergeDortmund 2009Promotionsausschuss1. Gutachter: Prof. Dr. Ursula Gather2. Gutachter: Prof. Dr. Roland FriedVorsitzende der Prüfungskommission: Prof. Dr. Katja IckstadtTag der mündlichen Prüfung05. März 2009AbstractThis thesis proposes new methods for real-time signal and variabilityextraction, presents derivations of their robustness properties and dis-cusses their value for practical applications to physiological time series.Althoughtheproposedtechniquesaredevelopedagainstthebackgroundof online monitoring in intensive care, they are also applicable to anyother kind of time series.For Repeated Median regression on an equidistant grid, the distributionof the position and number of zero residuals is investigated, and thecorrelation structure between the residual signs is examined.For online signal extraction, an adaptive filter is proposed which essenti-allyreliesonagoodness-of-fittestbasedonresidualsignsfromRepeatedMedian regression. After deriving suitable settings for this filter in theunivariate case from a simulation study, the procedure is extended forapplication to multivariate time series.