Comparison of Local Information Indices Applied in Resting State Functional Brain Network Connectivity Prediction

Abstract

Anatomical distance has been widely used to predict functional connectivity because of the potential relationship between structural connectivity and functional connectivity. The basic implicit assumption of this method is “distance penalization.” But studies have shown that one-parameter model (anatomical distance) cannot account for the small-worldness, modularity, and degree distribution of normal human brain functional networks. Two local information indices–common neighbor (CN) and preferential attachment index (PA), are introduced into the prediction model as another parameter to emulate many key topological of brain functional networks in the previous study. In addition to these two indices, many other local information indices can be chosen for investigation. Different indices evaluate local similarity from different perspectives. Currently, we still have no idea about how to select local information indices to achieve higher predicted accuracy of functional connectivity. Here, seven local information indices are chosen, including CN, hub depressed index (HDI), hub promoted index (HPI), Leicht-Holme-Newman index (LHN-I), Sørensen index (SI), PA, and resource allocation index (RA). Statistical analyses were performed on eight network topological properties to evaluate the predictions. Analysis shows that different prediction models have different performances in terms of simulating topological properties and most of the predicted network properties are close to the real data. There are four topological properties whose average relative error is less than 5%, including characteristic path length, clustering coefficient, global efficiency, and local efficiency. CN model shows the most accurate predictions. Statistical analysis reveals that five properties within the CN-predicted network do not differ significantly from the real data (P > 0.05, false-discovery rate method corrected for seven comparisons). PA model shows the worst prediction performance which was first applied in models of growth networks. Our results suggest that PA is not suitable for predicting connectivity in a small-world network. Furthermore, in order to evaluate the predictions rapidly, prediction power was proposed as an evaluation metric. The current study compares the predictions of functional connectivity with seven local information indices and provides a reference of method selection for construction of prediction models.

A Generative Network Model of the Human Brain Normal Aging Process

“…Prior studies [20,21] considered anatomical distance as a major determinant of the connection between two ROIs (regions of interest) during network modelling. Studies showed that network topological factor, as well as the anatomical factor, have significant effects on network formation and performance [12,22,23]. It is generally confirmed that link prediction algorithms provide various rules for link establishment and network formation by quantifying the topological similarity between two unconnected nodes [24,25].…”

“…If two nodes both have larger degrees, they score higher topological similarity and, therefore, their connecting probability is higher. At present, both network topological features and anatomical distance are treated as essential impact factors in modelling brain networks [12,22,23]. Vértes et al established a hemisphere-brain network by using an Economical Clustering Model (ECM) based on a link prediction algorithm of local topologies called common neighbours (CN) in order to simulate the formation mechanism of human brain networks [12,43].…”

“…It is particularly notable that artificial networks are generated by adding a certain amount of links at one time, and evaluations applied after all requested links are added [22,23]. In order for a clearer and more accurate observation on brain network formation, we apply a generative strategy as follows: For each given graph G (V, E), where V is the set of nodes, and E is the set of links.…”

“…Goodness-of-fit of each generative network model is theoretically evaluated by how well the generated networks fit the fMRI brain data by comparing the overall network performance of generated networks and fMRI brain networks. Inspired by previous studies on the evaluation of generative network models [12,22,23], the Likeness is calculated via Equation 18to measure the significance of the difference between a generated artificial brain network and the fMRI brain network, according to an overall condition. This function is defined based on the p-values associated with the two-sample t-test of the difference in the number of edges (edges), average clustering coefficient (C), modularity (Q), characteristic path length (L), global efficiency (Eglbl) and local efficiency (Elcl):…”

Resting-State Functional Network Scale Effects and Statistical Significance-Based Feature Selection in Machine Learning Classification

“…The topological features selected have usually included global properties [11], local properties [12], community structures [13], and connections [14]. In recent years, researchers have proposed new methods for network feature analysis, which have been applied in brain disease machine learning research, such as hypergraph [15], high-order network [16], minimum spanning tree [17], and frequent subgraphs [18] methods. Brain network topological features provide a new perspective for combined research using fMRI and machine learning.…”