ISSN 1062-8738, Bulletin of the Russian Academy of Sciences. Physics, 2012, Vol. 76, No. 12, pp. 1361–1364. © Allerton Press, Inc., 2012. Original Russian Text © V.V. Grubov, E.Yu. Sitnikova, A.A. Koronovskii, A.N. Pavlov, A.E. Hramov, 2012, published in Izvestiya Rossiiskoi Akademii Nauk. Seriya Fizicheskaya, 2012, Vol. 76, No. 12, pp. 1520–1523. 1361 INTRODUCTION There are a great number of effective radiophysical methods for analyzing and diagnosing the behavior of complex oscillatory systems. These are widely used in different fields of natural science in studying and diag- nosing many systems, including those found in medi- cine and physiology [1–3]. The application of these methods in analyzing the rhythmic activity of the brain is particularly relevant. This activity is a result of the synchronized operation of a large number of neu- rons that make up the complex oscillating network of the brain [1]. Electroencephalograms (EEGs) are traditionally used in neurophysiological studies to analyze brain activities [4]. An EEG is the average sum of the cur- rents generated by the group of neurons in the area of the recording electrode. Several frequency bands (alpha, beta, gamma, etc.) are can usually be distin- guished in an EEG signal. It has been proven that there is a clear correlation between the nature of the rhyth- mic activity in an EEG in a specific frequency range (due to the presence of a rhythm or oscillatory pattern) and the functional state of the body [1, 4]. An impor- tant aspect of investigating the nervous system is thus the study of certain oscillatory patterns, along with the regularities of their appearance in EEGs in different states of the organism. One type of oscillatory activity in an EEG, which manifests itself during sleep, is sleep spindles, i.e., short (0.5–1.5 s) episodes of oscillations with frequen- cies of 10–16 Hz and a characteristic spindle shape [5]. It is known that sleep spindles are formed due to the synchronous operation of the neural network that consists of the cortex and the thalamus neurons. The interest in studying sleep spindles is due to their possible connection with epilepsy. [6] It is known that the neural network that normally generates sleep spindles can, under certain conditions, produce sei- zure activity (i.e., spike wave discharges) [7]. Spike wave discharges serve as a diagnostic sign of absence epilepsy, and their presence in an EEG is accompa- nied by characteristic clinical manifestations. There is a relationship between the neurophysiological mecha- nisms of spike-wave discharges and sleep spindles, but it is complex and far from obvious. The aim of this work was to study the time-fre- quency dynamics of sleep spindles (oscillatory pat- terns) in the EEGs of rats with a genetic predisposition to absence epilepsy (the WAG/Rij line). Both tradi- tional wavelet analysis [8] and a new method for decomposing signals into empirical modes (the Hil- bert–Huang transform) [9, 10] were used in our anal- ysis of EEGs. EXPERIMENTAL DATA AND TIME-FREQUENCY EEG ANALYSIS We used EEG records for the cortex and thalamus of six adult WAG/Rij rats. Recording was continuous over 24 hours and thus contains fragments of sleep with strong sleepy spindles and fragments of wakeful- ness. EEG signals were prefiltered in the range of 0.5– 100 Hz. Our experimental work was performed at the Institute of Higher Nervous Activity and Neurophysi- ology of the Russian Academy of Sciences. Continuous wavelet transform (CWT) [8] was used for the initial study of EEG signals. As applied to our Automatic Extraction and Analysis of Oscillatory Patterns on Nonstationary EEG Signals by Means of Wavelet Transform and the Empirical Modes Method V. V. Grubov a, c , E. Yu. Sitnikova b , A. A. Koronovskii a, c , A. N. Pavlov a , and A. E. Hramov a, c a Saratov State University, Saratov, 410012 Russia b Institute of Higher Nervous Activity and Neurophysiology, Russian Academy of Sciences, Moscow, 117485 Russia c Gagarin Saratov State Technical University, Saratov, 410054 Russia e-mail: vvgrubov@gmail.com Abstract—The time-frequency structure and dynamics of oscillatory patterns in electroencephalograms of rats is studied by means of continuous wavelet transform and the decomposition of the signal by empirical modes. A method for the automatic selection of patterns using the empirical modes is developed. The method is applied to the study of sleep spindles, and it is shown that their dynamics depends on the regularities of on– off intermittency. DOI: 10.3103/S1062873812120167