Adjustment of Synchronization Stability of Dynamic Brain-Networks Based on Feature Fusion

Abstract

When the brain is active, the neural activities of different regions are integrated on various spatial and temporal scales; this is termed the synchronization phenomenon in neurobiological theory. This synchronicity is also the main underlying mechanism for information integration and processing in the brain. Clinical medicine has found that some of the neurological diseases that are difficult to cure have deficiencies or abnormalities in the whole or local integration processes of the brain. By studying the synchronization capabilities of the brain-network, we can intensively describe and characterize both the state of the interactions between brain regions and their differences between people with a mental illness and a set of controls by measuring the rapid changes in brain activity in patients with psychiatric disorders and the strength and integrity of their entire brain network. This is significant for the study of mental illness. Because static brain network connection methods are unable to assess the dynamic interactions within the brain, we introduced the concepts of dynamics and variability in a constructed EEG brain functional network based on dynamic connections, and used it to analyze the variability in the time characteristics of the EEG functional network. We used the spectral features of the brain network to extract its synchronization features and used the synchronization features to describe the process of change and the differences in the brain network's synchronization ability between a group of patients and healthy controls during a working memory task. We propose a method based on the fusion of traditional features and spectral features to achieve an adjustment of the patient's brain network synchronization ability, so that its synchronization ability becomes consistent with that of healthy controls, theoretically achieving the purpose of the treatment of the diseases. Studying the stability of brain network synchronization can provide new insights into the pathogenic mechanism and cure of mental diseases and has a wide range of potential applications.

Keywords: EEG; EEG dynamic brain network; brain network synchronization adjustment and control; brain network synchronization stability; working memory.

Figure 1

Short-term memory scanning task (SMTS) paradigm. Subjects needed to remember the numbers appearing on the screen during the encoding stage and recall these during the maintenance stage. Finally, in the retrieval stage, the subjects were required to determine whether the number had appeared by searching their memory.

Figure 2

Brain-network. (A) Represents the healthy controls, (B) represents the patients.

Figure 3

EEG Working memory data Laplacian eigenvalue spectrum (descending order) diagram.

Figure 4

Time of initial synchronization in the healthy controls and the patients. (A) is the encoding stage, (B) is the maintenance stage, and (C) is the retrieval stage.

Figure 5

Brain network structure. (A) is a binary matrix constructed using the healthy controls; (B) is a binary matrix constructed using the patient subjects; (C) is the difference significant area S, a mathematical representation of (A,B,D) is a brain electrode position diagram corresponding to a (C) S region in the brain map.

Figure 6

R-value changes with the coupling of the S area in patients and healthy controls. The patient subject graphed here was #6 and the healthy control subject was #4 in (A), and the patient subject graphed in (B) was #4 and the healthy control subject was #6.

Figure 7

R-value changes with the strength of the S area in patients and healthy controls. The patient subject graphed here was #6 in (A) and the healthy control subject was #4 in (B).

Figure 8

R-value changes in a patient subject. Matrix G represents mean-clustering-coef and the data are from patient #6. (A) shows the parameters aa, bb are positive. (B) shows the parameter aa is negative, bb is positive. (C) shows the parameter aa, bb are negative.

Figure 9

R-value changes in a healthy subject. Matrix G represents mean-path-length and the data are from patient #6. (A) shows the parameters aa, bb are positive. (B) shows the parameter aa is negative, bb is positive. (C) shows the parameter aa, bb are negative.

Figure 10

Adjustment of patient's brain- network synchronization ability. The data are from patient #6 and healthy control subject number #4. The abscissa is the value of parameter bb, and the ordinate is the corresponding brain network synchronization feature value. P-A indicates the synchronization ability of the patient after the brain network adjustment; P-I indicates the synchronization ability of the patient before the brain network adjustment; N-I indicates normal synchronization ability.