Limits...
Exploring the functional brain network of Alzheimer's disease: based on the computational experiment.

Li Y, Qin Y, Chen X, Li W - PLoS ONE (2013)

Bottom Line: We found there were differing degrees of decline in global efficiency, clustering coefficient, number of edges and transitivity in AD patients compared with healthy controls.The results of SVM classification of the evolution network show that the evolution network had a greater probability to be classified as an AD patients group.The use of SVM to classify the results of network evolution provides an objective criteria for determining evolution results.

View Article: PubMed Central - PubMed

Affiliation: Image Processing and Intelligent Control Key Laboratory of Education Ministry of China, Wuhan, China ; Department of Control Science and Engineering, Huazhong University of Science and Technology, Wuhan, China.

ABSTRACT
The purpose of this study is to explore the changes in functional brain networks of AD patients using complex network theory. In this study, resting-state fMRI datasets of 10 AD patients and 11 healthy controls were collected. Time series of 90 brain regions were extracted from the fMRI datasets after preprocessing. Pearson correlation method was used to calculate the correlation coefficient between any two time series. Then, a wide threshold range was selected to transform the adjacency matrix to a binary matrix under a different threshold. The topology parameters of each binary network were calculated, and all of them were then averaged within a group. During the evolution, node betweenness and the Euclidean distance between the nodes were set as control factors. Each binary network of healthy controls underwent evolution of 100 steps in accordance with the evolution rules. Then, the topology parameters of the evolution network were calculated. Finally, support vector machine (SVM) was used to classify the network topology parameters of the evolution network and to determine whether evolution results matched the datasets from AD patients. We found there were differing degrees of decline in global efficiency, clustering coefficient, number of edges and transitivity in AD patients compared with healthy controls. The topology parameters of the evolution network tended toward those of the AD group. The results of SVM classification of the evolution network show that the evolution network had a greater probability to be classified as an AD patients group. A new biological marker for diagnosis of AD was provided through comparison of topology parameters between AD patients and healthy controls. The study of network evolution strategies enriched the method of brain network evolution. The use of SVM to classify the results of network evolution provides an objective criteria for determining evolution results.

Show MeSH
The trend of the network topology parameters during the process of evolution.The black line represents transitivity, the red line indicates the global efficiency, the blue line shows the clustering coefficient. Excluding the impact of random factors, the topology parameters can be considered stable for evolution steps up to 100.
© Copyright Policy
Related In: Results  -  Collection


getmorefigures.php?uid=PMC3760893&req=5

pone-0073186-g001: The trend of the network topology parameters during the process of evolution.The black line represents transitivity, the red line indicates the global efficiency, the blue line shows the clustering coefficient. Excluding the impact of random factors, the topology parameters can be considered stable for evolution steps up to 100.

Mentions: For evolution processing from one network to another, connection weights (Equation 1) and disconnect weights (Equation 2) were formulated in this study:(1)(2)Where CP(i,j) and DP(i,j) denote the connection weights and disconnect weights. Ki, Kj represents the i-th and j-th node betweenness respectively, D(i, j) represents the anatomical distance between the node i and j. In order to coincide with the individual differences and uncertainties in the disease process, a random factor R(i, j) is added to the evolution process. R(i,j) is a 90 * 90 matrix of uniform distribution between 0 to 1. During the evolution process, there is no connection between the nodes i, j, and a new connection is established between the i-th and j-th node only if CP(i,j)≥0.5 and R(i,j)≤0.03. If a connection exists between the node i, j, and simultaneously satisfies DP(i, j)≥0.5 and R(i, j)≤0.03, then i, j are disconnected. Otherwise, the state between i and j remains. Our results show that the topology parameters of evolution network remain stable when the evolution does not exceed 100 steps (Figure 1). Therefore, the evolution process stops when the evolution reaches 100 steps.


Exploring the functional brain network of Alzheimer's disease: based on the computational experiment.

Li Y, Qin Y, Chen X, Li W - PLoS ONE (2013)

The trend of the network topology parameters during the process of evolution.The black line represents transitivity, the red line indicates the global efficiency, the blue line shows the clustering coefficient. Excluding the impact of random factors, the topology parameters can be considered stable for evolution steps up to 100.
© Copyright Policy
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC3760893&req=5

pone-0073186-g001: The trend of the network topology parameters during the process of evolution.The black line represents transitivity, the red line indicates the global efficiency, the blue line shows the clustering coefficient. Excluding the impact of random factors, the topology parameters can be considered stable for evolution steps up to 100.
Mentions: For evolution processing from one network to another, connection weights (Equation 1) and disconnect weights (Equation 2) were formulated in this study:(1)(2)Where CP(i,j) and DP(i,j) denote the connection weights and disconnect weights. Ki, Kj represents the i-th and j-th node betweenness respectively, D(i, j) represents the anatomical distance between the node i and j. In order to coincide with the individual differences and uncertainties in the disease process, a random factor R(i, j) is added to the evolution process. R(i,j) is a 90 * 90 matrix of uniform distribution between 0 to 1. During the evolution process, there is no connection between the nodes i, j, and a new connection is established between the i-th and j-th node only if CP(i,j)≥0.5 and R(i,j)≤0.03. If a connection exists between the node i, j, and simultaneously satisfies DP(i, j)≥0.5 and R(i, j)≤0.03, then i, j are disconnected. Otherwise, the state between i and j remains. Our results show that the topology parameters of evolution network remain stable when the evolution does not exceed 100 steps (Figure 1). Therefore, the evolution process stops when the evolution reaches 100 steps.

Bottom Line: We found there were differing degrees of decline in global efficiency, clustering coefficient, number of edges and transitivity in AD patients compared with healthy controls.The results of SVM classification of the evolution network show that the evolution network had a greater probability to be classified as an AD patients group.The use of SVM to classify the results of network evolution provides an objective criteria for determining evolution results.

View Article: PubMed Central - PubMed

Affiliation: Image Processing and Intelligent Control Key Laboratory of Education Ministry of China, Wuhan, China ; Department of Control Science and Engineering, Huazhong University of Science and Technology, Wuhan, China.

ABSTRACT
The purpose of this study is to explore the changes in functional brain networks of AD patients using complex network theory. In this study, resting-state fMRI datasets of 10 AD patients and 11 healthy controls were collected. Time series of 90 brain regions were extracted from the fMRI datasets after preprocessing. Pearson correlation method was used to calculate the correlation coefficient between any two time series. Then, a wide threshold range was selected to transform the adjacency matrix to a binary matrix under a different threshold. The topology parameters of each binary network were calculated, and all of them were then averaged within a group. During the evolution, node betweenness and the Euclidean distance between the nodes were set as control factors. Each binary network of healthy controls underwent evolution of 100 steps in accordance with the evolution rules. Then, the topology parameters of the evolution network were calculated. Finally, support vector machine (SVM) was used to classify the network topology parameters of the evolution network and to determine whether evolution results matched the datasets from AD patients. We found there were differing degrees of decline in global efficiency, clustering coefficient, number of edges and transitivity in AD patients compared with healthy controls. The topology parameters of the evolution network tended toward those of the AD group. The results of SVM classification of the evolution network show that the evolution network had a greater probability to be classified as an AD patients group. A new biological marker for diagnosis of AD was provided through comparison of topology parameters between AD patients and healthy controls. The study of network evolution strategies enriched the method of brain network evolution. The use of SVM to classify the results of network evolution provides an objective criteria for determining evolution results.

Show MeSH