Original Article
Reliability evaluation on weighted graph metrics of fNIRS brain networks
Abstract
Background: Resting-state fNIRS (R-fNIRS) imaging data has proven to be a valuable technique to quantitatively characterize functional architectures of human brain network. However, whether the brain network metrics derived using weighted brain network model is test-retest (TRT) reliable remains largely unknown.
Methods: Here, we firstly constructed weighted brain networks on a group of 18 participants, and then applied graph-theory approach to quantify topological parameters of each weighted network. The intraclass correlation coefficient (ICC) was further applied to quantify the TRT reliability of network metrics.
Results: We found that the reliability of the weighted network metrics is threshold-sensitive, and most of these network metrics showed fair to excellent reliability. Specifically, the global network metrics, e.g., clustering coefficient, path length, local efficiency and global efficiency were of excellent level reliability (ICC >0.75) on both HbO and HbR signals. The nodal network metrics, e.g., degree and efficiency, generally also showed excellent level reliability on both HbO and HbR signals, and the reliability of these two metrics was better than that of nodal betweenness.
Conclusions: Overall, these findings demonstrated that most weighted network metrics derived from fNIRS are TRT reliable and can be used for brain network research.
Methods: Here, we firstly constructed weighted brain networks on a group of 18 participants, and then applied graph-theory approach to quantify topological parameters of each weighted network. The intraclass correlation coefficient (ICC) was further applied to quantify the TRT reliability of network metrics.
Results: We found that the reliability of the weighted network metrics is threshold-sensitive, and most of these network metrics showed fair to excellent reliability. Specifically, the global network metrics, e.g., clustering coefficient, path length, local efficiency and global efficiency were of excellent level reliability (ICC >0.75) on both HbO and HbR signals. The nodal network metrics, e.g., degree and efficiency, generally also showed excellent level reliability on both HbO and HbR signals, and the reliability of these two metrics was better than that of nodal betweenness.
Conclusions: Overall, these findings demonstrated that most weighted network metrics derived from fNIRS are TRT reliable and can be used for brain network research.
