1 A Novel Method for the Characterization of Synchronization and Coupling in Multichannel EEG and ECoG Manfred Hartmann, Andreas Graef, Hannes Perko, Christoph Baumgartner, and Tilmann Kluge Abstract—In this paper we introduce a novel method for the characterization of synchronziation and coupling effects in multivariate time series that can be used for the analysis of EEG or ECoG signals recorded during epileptic seizures. The method allows to visualize the spatio-temporal evolution of synchronization and coupling effects that are characteristic for epileptic seizures. Similar to other methods proposed for this purpose our method is based on a regression analysis. However, a more general definition of the regression together with an effective channel selection procedure allows to use the method even for time series that are highly correlated, which is commonly the case in EEG/ECoG recordings with large numbers of electrodes. The method was experimentally tested on ECoG recordings of epileptic seizures from patients with temporal lobe epilepsies. A comparision with the results from a independent visual inspection by clinical experts showed an excellent agreement with the patterns obtained with the proposed method. Keywords—EEG, epilepsy, regression analysis, seizure propagation. I. I NTRODUCTION A. Background Determination of electroencephalogram (EEG) and electro- corticogram (ECoG) activity propagation is important for the investigation of information processing in the human brain. It can be used for the analysis of epilepsy, a disease that is characterized by a sudded and recurrent malfunction of the brain that is termed seizure. Epileptic seizures reflect an excessive and hypersynchronous activity of neurons in the brain. They originate from a certain region in the brain and may spread out over other regions. Localization of the epileptic focus and brain regions affected by seizures is an important task for the clinical therapy, in particular in the course of pre- surgical clarification. This is usually done by visual inspection of raw EEG or ECoG. Numerous methods for the analysis of activity propagation have been published that may support clinicians with this task [1]. Important measures hereby are Manuscript received June 25, 2008 M. Hartmann, A. Graef, H. Perko and T. Kluge are with Austrian Research Centers GmbH - ARC, smart systems Division, Deparment of Neuroinfor- matics, Vienna, Austria C. Baumgartner is with General Hospital Hietzing with Neurological Center Rosenhügel, Neurological Department, Vienna, Austria the directed transfer function [2], [3], [4] and modifications of it [5], directed coherence and partial directed coherence [6], [7], and the ordinary coherence of multivariate spectral estimates [8]. These methods are based on autoregressive spectral estimation of multivariate signals [9] and an analysis of the resulting estimated parametric spectra. More precisely, multivariate spectra of the signals are estimated under the assumption of an autoregressive signal model. These are parametric models that are used in order to avoid overfitting, which is an important issue since due to short time stationarity the number of samples for this estimation typically is hardly limited. The measures mentioned above are calculated with regard to the obtained parametric spectra. B. Contributions In this article we introduce a novel approach for the characterization of synchronization and coupling effects in multichannel EEG and ECoG. The method is based on linear spatio-temporal regressions of individual signals that are gen- erated from a specific temporal neighborhood of each sample itself and of a selection of related signals. In contrast to most similar methods based on autoregressive spectral analysis, the temporal neighborhood for the regression may include succeeding samples in addition to preceeding samples. The regression parameters are obtained as the solution of a Wiener- Hopf equation using estimated correlation functions. For the subsequent analysis a linear decomposition of the regression value into terms related to the individual channels is used. The variance of these terms is derived, revealing measures of interactions and dependencies in the multivariate signal. Employed in a moving window scheme, temporal changes can be observed, and epileptic seizure evolution can be visualized. Based on the detailed spatio-temporal analysis, several measures characterizing synchronization and coupling effects from a more global perspective can be derived. A distinctive characteristic of our approach is the direct statistical analysis of the regression values and their linear terms, i.e., the regressands are linearly decomposed into terms associated to the involved signals. Based on the variances of these terms we introduce extrinsic-to-intrinsic power ratio (EIPR) and total extrinsic-to-intrinsic power ratio (TEIPR), which are physiologically meaningful and valuable measures. In contrast, numerous methods like directed transfer function, directed coherence, etc., are fundamentally different in that a multichannel autoregressive process is estimated and spectrally analyzed. World Academy of Science, Engineering and Technology 44 2008 6