Limits...
Making Large-Scale Networks from fMRI Data.

Schmittmann VD, Jahfari S, Borsboom D, Savi AO, Waldorp LJ - PLoS ONE (2015)

Bottom Line: However, this approach generally results in a poor representation of the true underlying network.As a result, pairwise correlation networks can lead to fallacious conclusions; for example, one may conclude that a network is a small-world when it is not.We conclude that using partial correlations, informed by a sparseness penalty, results in more accurate networks and corresponding metrics than pairwise correlation networks.

View Article: PubMed Central - PubMed

Affiliation: Department of Methodology and Statistics/Social and Behavioral Sciences, Tilburg University, Tilburg, the Netherlands.

ABSTRACT
Pairwise correlations are currently a popular way to estimate a large-scale network (> 1000 nodes) from functional magnetic resonance imaging data. However, this approach generally results in a poor representation of the true underlying network. The reason is that pairwise correlations cannot distinguish between direct and indirect connectivity. As a result, pairwise correlation networks can lead to fallacious conclusions; for example, one may conclude that a network is a small-world when it is not. In a simulation study and an application to resting-state fMRI data, we compare the performance of pairwise correlations in large-scale networks (2000 nodes) against three other methods that are designed to filter out indirect connections. Recovery methods are evaluated in four simulated network topologies (small world or not, scale-free or not) in scenarios where the number of observations is very small compared to the number of nodes. Simulations clearly show that pairwise correlation networks are fragmented into separate unconnected components with excessive connectedness within components. This often leads to erroneous estimates of network metrics, like small-world structures or low betweenness centrality, and produces too many low-degree nodes. We conclude that using partial correlations, informed by a sparseness penalty, results in more accurate networks and corresponding metrics than pairwise correlation networks. However, even with these methods, the presence of hubs in the generating network can be problematic if the number of observations is too small. Additionally, we show for resting-state fMRI that partial correlations are more robust than correlations to different parcellation sets and to different lengths of time-series.

No MeSH data available.


Related in: MedlinePlus

Overlap between networks at different numbers of volumes (i.e., time-series lengths).Shown is the proportion of identical edges present in two respective networks. Black lines −− show overlap between the pairwise correlation network and the partial correlation network of a participant, based on a given number of volumes (i.e., time-series length). Separate lines for each participant (numbered 1 − 5). Red (or blue) lines indicate overlap between the pairwise correlation (red) (or partial correlation (blue)) network based on the full time-series of 240 volumes and the pairwise correlation (red) (or partial correlation (blue)) network based on smaller numbers of volumes (i.e., shorter time-series length.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4556712&req=5

pone.0129074.g018: Overlap between networks at different numbers of volumes (i.e., time-series lengths).Shown is the proportion of identical edges present in two respective networks. Black lines −− show overlap between the pairwise correlation network and the partial correlation network of a participant, based on a given number of volumes (i.e., time-series length). Separate lines for each participant (numbered 1 − 5). Red (or blue) lines indicate overlap between the pairwise correlation (red) (or partial correlation (blue)) network based on the full time-series of 240 volumes and the pairwise correlation (red) (or partial correlation (blue)) network based on smaller numbers of volumes (i.e., shorter time-series length.

Mentions: Fig 16 shows the obtained networks of the 3% strongest partial or pairwise correlations in the five participants. Both in the pairwise and in the partial correlation networks of all participants, those areas commonly reported as associated with resting-state activity (i.e., we considered precuneus, medialfrontal, inferior parietal, medial temporal lobe, primary sensorimotor, primary visual, extrastriate visual, bilateral temporal, insular, anterior cingulate cortex, superior parietal, superior frontal, posterior cingulate cortex, in line with [53–57]) had a larger average degree and a larger average betweenness than the remaining areas. However, the amount of overlap between pairwise and partial correlation networks was 62% at most, and decreased further with increasing number of ROIs or decreasing number of observations in each participant (see dashed black lines in Figs 17 and 18, respectively). As expected, network characteristics that depend on the inferred network topology differ substantially depending on the method used. Fig 19 shows network metrics of interest for the five participants over different parcellations and methods. As in the simulation study, the use of pairwise correlations results in more fragmented networks with a higher amount of clustering and a higher small-worldness index.


Making Large-Scale Networks from fMRI Data.

Schmittmann VD, Jahfari S, Borsboom D, Savi AO, Waldorp LJ - PLoS ONE (2015)

Overlap between networks at different numbers of volumes (i.e., time-series lengths).Shown is the proportion of identical edges present in two respective networks. Black lines −− show overlap between the pairwise correlation network and the partial correlation network of a participant, based on a given number of volumes (i.e., time-series length). Separate lines for each participant (numbered 1 − 5). Red (or blue) lines indicate overlap between the pairwise correlation (red) (or partial correlation (blue)) network based on the full time-series of 240 volumes and the pairwise correlation (red) (or partial correlation (blue)) network based on smaller numbers of volumes (i.e., shorter time-series length.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4556712&req=5

pone.0129074.g018: Overlap between networks at different numbers of volumes (i.e., time-series lengths).Shown is the proportion of identical edges present in two respective networks. Black lines −− show overlap between the pairwise correlation network and the partial correlation network of a participant, based on a given number of volumes (i.e., time-series length). Separate lines for each participant (numbered 1 − 5). Red (or blue) lines indicate overlap between the pairwise correlation (red) (or partial correlation (blue)) network based on the full time-series of 240 volumes and the pairwise correlation (red) (or partial correlation (blue)) network based on smaller numbers of volumes (i.e., shorter time-series length.
Mentions: Fig 16 shows the obtained networks of the 3% strongest partial or pairwise correlations in the five participants. Both in the pairwise and in the partial correlation networks of all participants, those areas commonly reported as associated with resting-state activity (i.e., we considered precuneus, medialfrontal, inferior parietal, medial temporal lobe, primary sensorimotor, primary visual, extrastriate visual, bilateral temporal, insular, anterior cingulate cortex, superior parietal, superior frontal, posterior cingulate cortex, in line with [53–57]) had a larger average degree and a larger average betweenness than the remaining areas. However, the amount of overlap between pairwise and partial correlation networks was 62% at most, and decreased further with increasing number of ROIs or decreasing number of observations in each participant (see dashed black lines in Figs 17 and 18, respectively). As expected, network characteristics that depend on the inferred network topology differ substantially depending on the method used. Fig 19 shows network metrics of interest for the five participants over different parcellations and methods. As in the simulation study, the use of pairwise correlations results in more fragmented networks with a higher amount of clustering and a higher small-worldness index.

Bottom Line: However, this approach generally results in a poor representation of the true underlying network.As a result, pairwise correlation networks can lead to fallacious conclusions; for example, one may conclude that a network is a small-world when it is not.We conclude that using partial correlations, informed by a sparseness penalty, results in more accurate networks and corresponding metrics than pairwise correlation networks.

View Article: PubMed Central - PubMed

Affiliation: Department of Methodology and Statistics/Social and Behavioral Sciences, Tilburg University, Tilburg, the Netherlands.

ABSTRACT
Pairwise correlations are currently a popular way to estimate a large-scale network (> 1000 nodes) from functional magnetic resonance imaging data. However, this approach generally results in a poor representation of the true underlying network. The reason is that pairwise correlations cannot distinguish between direct and indirect connectivity. As a result, pairwise correlation networks can lead to fallacious conclusions; for example, one may conclude that a network is a small-world when it is not. In a simulation study and an application to resting-state fMRI data, we compare the performance of pairwise correlations in large-scale networks (2000 nodes) against three other methods that are designed to filter out indirect connections. Recovery methods are evaluated in four simulated network topologies (small world or not, scale-free or not) in scenarios where the number of observations is very small compared to the number of nodes. Simulations clearly show that pairwise correlation networks are fragmented into separate unconnected components with excessive connectedness within components. This often leads to erroneous estimates of network metrics, like small-world structures or low betweenness centrality, and produces too many low-degree nodes. We conclude that using partial correlations, informed by a sparseness penalty, results in more accurate networks and corresponding metrics than pairwise correlation networks. However, even with these methods, the presence of hubs in the generating network can be problematic if the number of observations is too small. Additionally, we show for resting-state fMRI that partial correlations are more robust than correlations to different parcellation sets and to different lengths of time-series.

No MeSH data available.


Related in: MedlinePlus