Multilayer perceptron reveals functional connectivity structrure in thalamo-cortical brain network

Artificial neural networks (ANNs) are known to be a powerful tool for big data analysis. They are widely used in computer science, nonlinear dynamics, robotics, and neuroscience for solving tasks of classification, forecasting, pattern recognition, etc. In neuroscience ANNs allow recognizing specific forms of brain activity from multichannel electro- (EEG) or magnetoencephalographic (MEG) data and, therefore, widely used as a computational core in various brain-computer interfaces. Another challenging problem is the analysis of connectivity structures in big multivariate data. In neuroscience restoring the functional brain network using multichannel EEG/MEG signals uncovers mechanisms of neuronal interaction during various physiological or cognitive processes. In this report we use recent advances in the area of machine learning known as feed-forward artificial neuronal network to formulate a method for detecting functional dependence in unidirectionally and bidirectionally coupled systems without additional information about them. We apply our method for the first time to reveal functional connectivity structure in the thalamo-cortical network of epileptic brain based on a rodent electrocorticography (ECoG) data set.