No abstract

SUMMARYBrain waves are derived as multiple time-series signals from electrodes placed on various sites on the scalp. If the flow of information among these multiple sites can be visualized, it will be useful in obtaining a clue to the objective interpretation of brain activity, or in diagnosing the existence of failures of brain function. Directed information analysis has been proposed as a means of causality analysis in which the flow of information among the signals is investigated. The method notes an arbitrary two of a large number of time series, and examines the flow of information between these. When three or more signal series exist, they are in general related to each other. Consequently, even if the causality between two particular series is to be examined, the accurate flow of information cannot be determined unless the effect of another series is considered. Consequently, the authors have proposed multidimensional directed information analysis, which is an extension of directed information analysis. This paper tries to verify the effectiveness of multidimensional directed information analysis by simulation. Then, the method is applied to the brain waves of a healthy subject and a patient with a cerebral organic disorder, and it is verified that there is a difference, reflecting the presence of disease in the sense of information propagation in the cortex. © 2001 Scripta Technica, Electron Comm Jpn Pt 3, 85(4): 4555, 2002

Abstract-In psychiatry, the sleep stage is one of the most important evidence for diagnosing mental disease. However, doctors require much labor and skill for diagnosis, so a quantitative and objective method is required for more accurate diagnosis since it depends on the doctor's experience. For this reason, an automatic diagnosis system must be developed. In this paper, we propose a new type of neural network (NN) model referred to as a sleep electroencephalogram (EEG) recognition neural network (SRNN) which enables us to detect several kinds of important characteristic waves in sleep EEG which are necessary for diagnosing sleep stages. Experimental results indicate that the proposed NN model was much more capable than other conventional methods for detecting characteristic waves.Index Terms-Charachteristic wave, EEG, neural network, sleep stage. I. ITRODUCTION IN PSYCHIATRY, sleep staging is one of the most important means for diagnosis. The sleep staging of electroencephalogram (EEG), however, is liable to be subjective since it depends on the doctor's skill and requires much labor. An automatic diagnosis system must, therefore, be developed to reduce doctor's labor and realize quantitative diagnosis of sleep EEG.For sleep staging by EEG analysis, it is especially important to detect the characteristic waves from EEG. Most conventional methods of diagnosing the sleep stage, however, use long-term spectrum analysis [5], [6]. Such analysis is unable to detect transient and isolated characteristic waves such a hump wave from sleep EEG accurately. As a result, it is not possible to precisely diagnose the sleep stage based on characteristic waves as doctors do, though a roughly diagnosis is possible.Moreover, some methods are based on a kind of template matching. This makes it difficult to cope with the large variation of EEG, such as fluctuations of the frequency pattern and the differences between individuals. The method of detecting the characteristic waves in EEG must, therefore, be able to recognize the time transient of the frequency pattern and be robust to variation of patterns.Manuscript received July 3, 1998; revised September 2, 1999. Asterisk indicates corresponding author.*T. Shimada is with the Applied Superconductivity Research Laboratory, Tokyo Denki University 2-1200 Muzai-Gakuendai, Inzai-shi, Chiba 270-1382, Japan (e-mail: shimada@asrl.dendai.ac.jp).T. Shiina is with the Institute of Information Sciences and Electronics, University of Tsukuba 1-1-1 Tennoudai, Tsukuba-City, Ibaraki 305-8573, Japan.Y. Saito is with the Research Institute for EEG Analysis 1-9-1-101 Nakasato, Kitaku, Tokyo 114-0015, Japan.Publisher Item Identifier S 0018-9294(00)01774-2.Recently, neural networks (NN's) have been applied to various kinds of problems in many fields due to their ability to analyze complicated systems without accurate modeling in advance [4]. In this paper, we propose a new type of NN model to detect several kinds of important characteristic waves in sleep EEG that are needed to diagnose sleep stages (...

No abstract

No abstract