BACKGROUND: Category fluency is impaired early in Alzheimer disease (AD). Graph theory is a technique to analyze complex relationships in networks. Features of interest in network analysis include the number of nodes and edges, and variables related to their interconnectedness. Other properties important in network analysis are "small world properties" and "scale-free" properties. The small world property (popularized as the so-called "6 degrees of separation") arises when the majority of connections are local, but a number of connections are to distant nodes. Scale-free networks are characterized by the presence of a few nodes with many connections, and many more nodes with fewer connections. OBJECTIVE: To determine if category fluency data can be analyzed using graph theory. To compare normal elderly, mild cognitive impairment (MCI) and AD network graphs, and characterize changes seen with increasing cognitive impairment. METHODS: Category fluency results ("animals" recorded over 60 s) from normals (n=38), MCI (n=33), and AD (n=40) completing uniform data set evaluations were converted to network graphs of all unique cooccurring neighbors, and compared for network variables. RESULTS: For Normal, MCI and AD, mean clustering coefficients were 0.21, 0.22, 0.30; characteristic path lengths were 3.27, 3.17, and 2.65; small world properties decreased with increasing cognitive impairment, and all graphs showed scale-free properties. Rank correlations of the 25 commonest items ranged from 0.75 to 0.83. Filtering of low-degree nodes in normal and MCI graphs resulted in properties similar to the AD network graph. CONCLUSIONS: Network graph analysis is a promising technique for analyzing changes in category fluency. Our technique results in nonrandom graphs consistent with well-characterized properties for these types of graphs.
BACKGROUND: Category fluency is impaired early in Alzheimer disease (AD). Graph theory is a technique to analyze complex relationships in networks. Features of interest in network analysis include the number of nodes and edges, and variables related to their interconnectedness. Other properties important in network analysis are "small world properties" and "scale-free" properties. The small world property (popularized as the so-called "6 degrees of separation") arises when the majority of connections are local, but a number of connections are to distant nodes. Scale-free networks are characterized by the presence of a few nodes with many connections, and many more nodes with fewer connections. OBJECTIVE: To determine if category fluency data can be analyzed using graph theory. To compare normal elderly, mild cognitive impairment (MCI) and AD network graphs, and characterize changes seen with increasing cognitive impairment. METHODS: Category fluency results ("animals" recorded over 60 s) from normals (n=38), MCI (n=33), and AD (n=40) completing uniform data set evaluations were converted to network graphs of all unique cooccurring neighbors, and compared for network variables. RESULTS: For Normal, MCI and AD, mean clustering coefficients were 0.21, 0.22, 0.30; characteristic path lengths were 3.27, 3.17, and 2.65; small world properties decreased with increasing cognitive impairment, and all graphs showed scale-free properties. Rank correlations of the 25 commonest items ranged from 0.75 to 0.83. Filtering of low-degree nodes in normal and MCI graphs resulted in properties similar to the AD network graph. CONCLUSIONS: Network graph analysis is a promising technique for analyzing changes in category fluency. Our technique results in nonrandom graphs consistent with well-characterized properties for these types of graphs.
Authors: Steven Lenio; Frances M Lissemore; Martha Sajatovic; Kathleen A Smyth; Curtis Tatsuoka; Wojbor A Woyczynski; Alan J Lerner Journal: Front Aging Neurosci Date: 2016-10-25 Impact factor: 5.750
Authors: Laura Serra; Matteo Mancini; Gabriella Silvestri; Antonio Petrucci; Marcella Masciullo; Barbara Spanò; Mario Torso; Chiara Mastropasqua; Manlio Giacanelli; Carlo Caltagirone; Mara Cercignani; Giovanni Meola; Marco Bozzali Journal: Neural Plast Date: 2016-05-25 Impact factor: 3.599