Dresden 2014 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
DY: Fachverband Dynamik und Statistische Physik
DY 13: Poster - Glasses / Stat. Phys. Bio. / Networks (joint session DY/BP/CPP/SOE)
DY 13.23: Poster
Dienstag, 1. April 2014, 09:30–12:30, P1
Estimation of sleep stages and sleep depth dynamics by neural clustering — •Stephan Volkland1 and Jens Christian Claussen2,1 — 1INB, Universität zu Lübeck, Germany — 2Computational Systems Biology Lab, Research II, Jacobs University Bremen, Germany
The quantitative analysis of sleep from polysomnographic date (i.e., simulataneous recording of EEG, EMG and EOG) is practically limited by the final step of sleep scoring, i.e. extensive manual inspection of data according to the Rechtschaffen and Kales rules or the recent AASM counterpart, leading to a manually classified time series of sleep stages on a discrete scale of six values (wake, REM, S1, S2, S3, S4). Starting from the observation that the stages S2, S3, and S4 are merely defined by spectral properties, namely by activity in the delta and sigma band, here we present a neural clustering approach to assess sleep stages automatically by unsupervised neural clustering and a posteriori assignment of sleep stages. One particular goal is to provide a finer resolution in time as well as a finer interpolation in sleep depth than obtainable from manual scoring. We find that EOG and EMG data are still needed to improve classification of wake and REM states, and still an interpolation of states at the borders of wake, REM and S1 is difficult. In the range between the NonREM stages S1–S4 an interpolation with higher resolution is feasible, as expected.