Research On Classification Of Brain Signals And Construction Of Brain Functional Network Based On Mutual Information

As one of the most complex systems in nature, the human brain has become one of the major challenges faced by scientists today. By combining brain nerve technology with other disciplines, more and more researchers carry out a lot of multi-domain and multi-level research works on brain’s structure and function, in order to reveal the nature of brain’s activities and the operational mechanism of brain neurons deeply. Referring to previous works, this paper studies the epilepsy automatic detection algorithm and brain network of the schizophrenia based on mutual information(MI) respectively, and discusses the characteristics and the pathogenesis of these brain diseases that damage human health.This paper introduces the research background, significance and status quo of the epilepsy automatic detection algorithm and schizophrenia brain network, and then describes the basic principles and developments of the brain’s functional regions, the brain’s rhythm, MI, support vector machine(SVM) and complex network in detail. After these researches, this paper completes the following works.First, this paper improves the epilepsy automatic detection algorithm based MI. We use MI to quantify the relationship between the channels. Then automatically select channels which have better classification results and extract statistical values as feature vectors from MI sequence of the selected channel and import the feature vectors extracted from the channel to SVM learning and training. The experiments include two sets of data, which are electrocorticogram(ECoG) of 8 epileptic seizures and electroencephalogram(EEG) of 18 epileptic patients and 18 normal controls respectively. The experimental results show that the improved algorithm has not only higher accuracy but also faster processing speed. In addition, improved algorithm adds the adaptive selection of channel which has best classification based on individual cases, thus provides a useful reference for real-time clinical epilepsy monitoring.Second, this paper designs the construction algorithm of brain functional network based on MI and applies the algorithm to magnetoencephalography(MEG) of normal controls, schizophrenia patients and their uneffected relatives. We construct the brain functional networks and extract the complex networks’ measures of three categories of people to analyze the differences among their brain networks’ structure and information transmission. The results show that the normal have more obvious functional regions and less information transmission among different regions under resting state than the others, which offer us the strong reference for understanding the pathogenesis and genetic mechanisms of schizophrenia.Last, as the different rhythm reflecting different physiological state, this paper further analyzes the brain functional networks of the three people using alpha rhythm to construct. The results indicate that the results of alpha rhythm are better and easier to distinguish the three people than that of the whole band. Moreover, this paper analyzes the characteristics of the brain functional network of other rhythm such as delta rhythm, theta rhythm and beta rhythm and compares these results with alpha rhythm’s. The results indicates that the global characteristics of functional networks for these rhythms are similar, while local characteristics are different, which can be used to further explore the differences of three categories of people under specific physical state.