Biomed Tech 2007; 52:83–89  2007 by Walter de Gruyter • Berlin • New York. DOI 10.1515/BMT.2007.016 2007/5216475 Article in press - uncorrected proof Coupled oscillators for modeling and analysis of EEG/MEG oscillations Lutz Leistritz 1, *, Peter Putsche 1 , Karin Schwab 1 , Wolfram Hesse 1 , Thomas Su ¨ ße 1 , Jens Haueisen 2,3 and Herbert Witte 1 1 Institute of Medical Statistics, Computer Sciences and Documentation, Medical Faculty, Friedrich Schiller University Jena, Jena, Germany 2 Biomagnetic Center, Department of Neurology, Medical Faculty, Friedrich Schiller University Jena, Jena, Germany 3 Institute of Biomedical Engineering and Informatics, Technical University Ilmenau, Ilmenau, Germany Abstract This study presents three EEG/MEG applications in which the modeling of oscillatory signal components offers complementary analysis and an improved expla- nation of the underlying generator structures. Coupled oscillator networks were used for modeling. Parameters of the corresponding ordinary coupled differential equa- tion (ODE) system are identified using EEG/MEG data and the resulting solution yields the modeled signals. This model-related analysis strategy provides information about the coupling quantity and quality between signal components (example 1, neonatal EEG during quiet sleep), allows identification of the possible contribution of hidden generator structures (example 2, 600-Hz MEG oscillations in somatosensory evoked magnetic fields), and can explain complex signal characteristics such as amplitude-frequency coupling and frequency entrain- ment (example 3, EEG burst patterns in sedated patients). Keywords: coupled oscillators; EEG; MEG; model-based signal analysis; parameter identification. Introduction EEG/MEG oscillations show characteristics identified in experiments using simulations with coupled oscillators, such as linear and non-linear phase synchronization/de- synchronization of the oscillations and frequency entrain- ment w 3x . As already demonstrated by other studies, simulations of physiological systems using coupled oscil- lator systems and corresponding model-based analysis methods not only contribute significantly to improve interpretation of the results, but also enhance the anal- ysis methods themselves w 4, 24x . In addition, coupled *Corresponding author: Lutz Leistritz, Institute of Medical Statistics, Computer Sciences and Documentation, Bachstr. 18, 07740 Jena, Germany Phone: q49-3641-93405 Fax: q49-3641-933200 E-mail: lutz.leistritz@mti.uni-jena.de oscillators (oscillator networks) can be directly used as an analysis method, whereby the parameters of the underlying ordinary coupled differential equation (ODE) system are identified using the measured EEG/MEG data w 13x . With this type of adapted ODE system, the solution yields the modeled signals. Using one ODE (oscillator), one EEG/MEG signal or signal component can be mod- eled. Interrelations between the signals or signal com- ponents are modeled by couplings between ODEs w 22x . The aim of this study is to demonstrate that model- based approaches can be used to analyze EEG patterns (example 1, neonatal EEG during quiet sleep). It can also be shown that models of coupled oscillators allow the identification of influences of hidden sources that cannot be measured and for which the signal characteristics are unknown. Adding hidden sources to the model leads to improved modeling and explanation of the EEG/MEG sig- nal results (example 2, 600-Hz MEG oscillations in soma- tosensory evoked magnetic fields). Furthermore, both approaches can be combined. A signal component relat- ed to a hidden source (e.g., known from experimental studies) can be used to model the contribution of such a hidden source to the whole signal pattern and to its signal characteristics. Parameters of a complete model might be utilized as analysis parameters (example 3, EEG burst patterns in sedated patients). Materials and methods Example 1: oscillatory networks for coupling analysis The first example is related to a group of six clinically and neurologically normal, full-term neonates (mean concep- tual age 39.3 weeks, range 38–41; mean birth weight 3152 g, range 2670–3420; 5-min APGAR score G8). The recordings were performed during sleep between 09:00 and 12:00 h. The EEG (unipolar recordings with linked ear reference; 8-channel EEG; international 10–20 sys- tems with electrodes Fp1, Fp2, C3, C4, T3, T4, O1, O2; sampling rate 128 Hz), heart rate, respiratory movements and EOG were recorded. Only the EEG recordings during quiet sleep were selected. The EEG was classified by a trained physician into the burst and interburst patterns. The data were filtered using two FFT filters (1.0–1.6 and 3.5–6.0 Hz) to extract rele- vant frequency ranges. Signal segments of 6 s in duration were used for the present examination, whereby one segment consists of a consecutive series of a 1-s inter- burst interval followed by a 4-s burst period and again a 1-s interburst. Example 2: modeling to explain 600-Hz oscillations Somatosensory evoked potentials and magnetic fields were simultaneously derived from 10 healthy volunteers Brought to you by | provisional account Unauthenticated | 141.35.67.128 Download Date | 7/31/14 12:11 PM