Estimation of the number of EEG sources in the human brain using multidimensional directed information analysis

At present, various approaches are being used to analyze brain function. One of these is the analysis of EEGs (electroencephalogram), by which it is becoming possible to evaluate brain activity and abnormalities in the brain from the EEG observed on the scalp. There remain many unknown aspects of the source of EEGs and their mechanism of propagation. A clue to such an approach is causality analysis, whose objective is to analyze causality by quantifying correlation relations, including the directions of information flow, among measurement sites, based on multiple EEG series obtained from multiple sites on the scalp. The method should be applicable to the diagnosis of disease such as the detection of individual abnormalities and failure of information propagation. Consequently, we have proposed multidimensional directed information analysis as a means of causality analysis in which the flow of information among all signals is investigated. There are many multidimensional signals in nature, for which the number of information flow sources in unknown. It is very important to discover the number of sources. Therefore, we propose a method to estimate the number of information flow sources by multidimensional directed information analysis. In this paper, the effectiveness of the proposed method has been verified by a simulation, and the method has been applied to the EEG of a healthy subject and a patient with a cerebral organic disorder. As a result, we have found that the healthy subject had six information flow sources and the patient had nine.