Limits...
Graph theoretical analysis of functional brain networks: test-retest evaluation on short- and long-term resting-state functional MRI data.

Wang JH, Zuo XN, Gohel S, Milham MP, Biswal BB, He Y - PLoS ONE (2011)

Bottom Line: We found that reliability of global network metrics was overall low, threshold-sensitive and dependent on several factors of scanning time interval (TI, long-term>short-term), network membership (NM, networks excluding negative correlations>networks including negative correlations) and network type (NT, binarized networks>weighted networks).Simulation analysis revealed that global network metrics were extremely sensitive (but varying degrees) to noise in functional connectivity and weighted networks generated numerically more reliable results in compared with binarized networks.For nodal network metrics, they showed high resistance to noise in functional connectivity and no NT related differences were found in the resistance.

View Article: PubMed Central - PubMed

Affiliation: State Key Laboratory of Cognitive Neuroscience and Learning, Beijing Normal University, Beijing, China.

ABSTRACT
Graph-based computational network analysis has proven a powerful tool to quantitatively characterize functional architectures of the brain. However, the test-retest (TRT) reliability of graph metrics of functional networks has not been systematically examined. Here, we investigated TRT reliability of topological metrics of functional brain networks derived from resting-state functional magnetic resonance imaging data. Specifically, we evaluated both short-term (<1 hour apart) and long-term (>5 months apart) TRT reliability for 12 global and 6 local nodal network metrics. We found that reliability of global network metrics was overall low, threshold-sensitive and dependent on several factors of scanning time interval (TI, long-term>short-term), network membership (NM, networks excluding negative correlations>networks including negative correlations) and network type (NT, binarized networks>weighted networks). The dependence was modulated by another factor of node definition (ND) strategy. The local nodal reliability exhibited large variability across nodal metrics and a spatially heterogeneous distribution. Nodal degree was the most reliable metric and varied the least across the factors above. Hub regions in association and limbic/paralimbic cortices showed moderate TRT reliability. Importantly, nodal reliability was robust to above-mentioned four factors. Simulation analysis revealed that global network metrics were extremely sensitive (but varying degrees) to noise in functional connectivity and weighted networks generated numerically more reliable results in compared with binarized networks. For nodal network metrics, they showed high resistance to noise in functional connectivity and no NT related differences were found in the resistance. These findings provide important implications on how to choose reliable analytical schemes and network metrics of interest.

Show MeSH

Related in: MedlinePlus

TRT reliability of summarized global network metrics (a) and metric-related differences in reliability (b).Areas under curves (AUCs) of each metrics were used to provide threshold-independent reliability estimation. Different metrics showed variable levels of reliability. Several of them were moderately reliable (e.g., lambda for S-AAL-based networks). Subsequent statistical analysis revealed significant differences in TRT reliability among the 12 global network metrics for S-AAL- but not for F-DOS-based networks. ICC values less than 0.25 were mapped to a single color of dark blue as well dark red color for ICC values greater than 0.75, respectively in (a). Network (+/-), networks constructed using absolute both positive and negative correlations; Network (+), networks constructed using only positive correlations; Binarized, binarized network analysis; Weighted, weighted network analysis; TRT, test-retest; S-AAL, structural ROIs from Anatomical Automatic Labeling atlas; F-DOS, functional ROIs from Dosenbach et al. (2006, 2010).
© Copyright Policy
Related In: Results  -  Collection


getmorefigures.php?uid=PMC3139595&req=5

pone-0021976-g006: TRT reliability of summarized global network metrics (a) and metric-related differences in reliability (b).Areas under curves (AUCs) of each metrics were used to provide threshold-independent reliability estimation. Different metrics showed variable levels of reliability. Several of them were moderately reliable (e.g., lambda for S-AAL-based networks). Subsequent statistical analysis revealed significant differences in TRT reliability among the 12 global network metrics for S-AAL- but not for F-DOS-based networks. ICC values less than 0.25 were mapped to a single color of dark blue as well dark red color for ICC values greater than 0.75, respectively in (a). Network (+/-), networks constructed using absolute both positive and negative correlations; Network (+), networks constructed using only positive correlations; Binarized, binarized network analysis; Weighted, weighted network analysis; TRT, test-retest; S-AAL, structural ROIs from Anatomical Automatic Labeling atlas; F-DOS, functional ROIs from Dosenbach et al. (2006, 2010).

Mentions: In the present study, individual networks were constructed at the same sparsity level by applying subject-specific correlation thresholds to individual correlation matrices (see Fig. S4 for the corresponding correlation thresholds under each sparsity level). Sparsity threshold ensures all resultant networks to have comparable topological structures of the same number of edges. Figure 5 shows the TRT reliability of 12 global network metrics over the whole sparisty range. Generally, most global network metrics exhibited poor to low reliability, irrespective of the factors of TI, NT and NM. For example, clustering coefficient was found to uniformly exhibit poor reliability (ICC<0.25) under all conditions. Nonetheless, we noted that some global metrics (e.g., lambda and assortativity ) exhibited modest long-term reliability when the networks were sparsely connected (sparsity<10%). Interestingly, we found that global network reliability appeared to depend on the factors of TI and NM but relatively insensitive to NT by qualitatively visual inspection. Specifically, long-term scans seemed to be associated with better reliability in compared with short-term scans and the exclusion of negative correlations enhanced network reliability (Fig. 5). These were reflected in both increased ICC values and the broadened threshold range of high ICC. Finally, a threshold-independent reliability scalar was obtained for each global network metric by using the AUC. Again, several specific global metrics (e.g., lambda) demonstrated moderate long-term reliability under certain analytical schemes (Fig. 6a, left). Of note, we found that assortativity showed moderated both short-term and long-term reliability for networks of positive correlations.


Graph theoretical analysis of functional brain networks: test-retest evaluation on short- and long-term resting-state functional MRI data.

Wang JH, Zuo XN, Gohel S, Milham MP, Biswal BB, He Y - PLoS ONE (2011)

TRT reliability of summarized global network metrics (a) and metric-related differences in reliability (b).Areas under curves (AUCs) of each metrics were used to provide threshold-independent reliability estimation. Different metrics showed variable levels of reliability. Several of them were moderately reliable (e.g., lambda for S-AAL-based networks). Subsequent statistical analysis revealed significant differences in TRT reliability among the 12 global network metrics for S-AAL- but not for F-DOS-based networks. ICC values less than 0.25 were mapped to a single color of dark blue as well dark red color for ICC values greater than 0.75, respectively in (a). Network (+/-), networks constructed using absolute both positive and negative correlations; Network (+), networks constructed using only positive correlations; Binarized, binarized network analysis; Weighted, weighted network analysis; TRT, test-retest; S-AAL, structural ROIs from Anatomical Automatic Labeling atlas; F-DOS, functional ROIs from Dosenbach et al. (2006, 2010).
© Copyright Policy
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC3139595&req=5

pone-0021976-g006: TRT reliability of summarized global network metrics (a) and metric-related differences in reliability (b).Areas under curves (AUCs) of each metrics were used to provide threshold-independent reliability estimation. Different metrics showed variable levels of reliability. Several of them were moderately reliable (e.g., lambda for S-AAL-based networks). Subsequent statistical analysis revealed significant differences in TRT reliability among the 12 global network metrics for S-AAL- but not for F-DOS-based networks. ICC values less than 0.25 were mapped to a single color of dark blue as well dark red color for ICC values greater than 0.75, respectively in (a). Network (+/-), networks constructed using absolute both positive and negative correlations; Network (+), networks constructed using only positive correlations; Binarized, binarized network analysis; Weighted, weighted network analysis; TRT, test-retest; S-AAL, structural ROIs from Anatomical Automatic Labeling atlas; F-DOS, functional ROIs from Dosenbach et al. (2006, 2010).
Mentions: In the present study, individual networks were constructed at the same sparsity level by applying subject-specific correlation thresholds to individual correlation matrices (see Fig. S4 for the corresponding correlation thresholds under each sparsity level). Sparsity threshold ensures all resultant networks to have comparable topological structures of the same number of edges. Figure 5 shows the TRT reliability of 12 global network metrics over the whole sparisty range. Generally, most global network metrics exhibited poor to low reliability, irrespective of the factors of TI, NT and NM. For example, clustering coefficient was found to uniformly exhibit poor reliability (ICC<0.25) under all conditions. Nonetheless, we noted that some global metrics (e.g., lambda and assortativity ) exhibited modest long-term reliability when the networks were sparsely connected (sparsity<10%). Interestingly, we found that global network reliability appeared to depend on the factors of TI and NM but relatively insensitive to NT by qualitatively visual inspection. Specifically, long-term scans seemed to be associated with better reliability in compared with short-term scans and the exclusion of negative correlations enhanced network reliability (Fig. 5). These were reflected in both increased ICC values and the broadened threshold range of high ICC. Finally, a threshold-independent reliability scalar was obtained for each global network metric by using the AUC. Again, several specific global metrics (e.g., lambda) demonstrated moderate long-term reliability under certain analytical schemes (Fig. 6a, left). Of note, we found that assortativity showed moderated both short-term and long-term reliability for networks of positive correlations.

Bottom Line: We found that reliability of global network metrics was overall low, threshold-sensitive and dependent on several factors of scanning time interval (TI, long-term>short-term), network membership (NM, networks excluding negative correlations>networks including negative correlations) and network type (NT, binarized networks>weighted networks).Simulation analysis revealed that global network metrics were extremely sensitive (but varying degrees) to noise in functional connectivity and weighted networks generated numerically more reliable results in compared with binarized networks.For nodal network metrics, they showed high resistance to noise in functional connectivity and no NT related differences were found in the resistance.

View Article: PubMed Central - PubMed

Affiliation: State Key Laboratory of Cognitive Neuroscience and Learning, Beijing Normal University, Beijing, China.

ABSTRACT
Graph-based computational network analysis has proven a powerful tool to quantitatively characterize functional architectures of the brain. However, the test-retest (TRT) reliability of graph metrics of functional networks has not been systematically examined. Here, we investigated TRT reliability of topological metrics of functional brain networks derived from resting-state functional magnetic resonance imaging data. Specifically, we evaluated both short-term (<1 hour apart) and long-term (>5 months apart) TRT reliability for 12 global and 6 local nodal network metrics. We found that reliability of global network metrics was overall low, threshold-sensitive and dependent on several factors of scanning time interval (TI, long-term>short-term), network membership (NM, networks excluding negative correlations>networks including negative correlations) and network type (NT, binarized networks>weighted networks). The dependence was modulated by another factor of node definition (ND) strategy. The local nodal reliability exhibited large variability across nodal metrics and a spatially heterogeneous distribution. Nodal degree was the most reliable metric and varied the least across the factors above. Hub regions in association and limbic/paralimbic cortices showed moderate TRT reliability. Importantly, nodal reliability was robust to above-mentioned four factors. Simulation analysis revealed that global network metrics were extremely sensitive (but varying degrees) to noise in functional connectivity and weighted networks generated numerically more reliable results in compared with binarized networks. For nodal network metrics, they showed high resistance to noise in functional connectivity and no NT related differences were found in the resistance. These findings provide important implications on how to choose reliable analytical schemes and network metrics of interest.

Show MeSH
Related in: MedlinePlus