About this Research Topic
The focus of research in imaging neuroscience is changing from a functional specialization approach to a functional integration approach. That is, characterizations are sought that no longer describe function and behaviour in terms of specialized brain loci seen in response to specific task performance, but in terms of spatially distinct distributed systems of connected regions. Under this characterisation the strengths and directions of connections between regions are the defining characteristics of brain function. One of the fundamental issues in determining the connectivity within a network is in the definition of the nodes of the network. A second issue is how to characterise an entire network in terms of the strength of connections, efficiency, parsimony, etc. This Research Topic focuses on these two issues: (I) the effect of node selection on connectivity estimates and (II) methods to summarise the characteristics of an entire network.
Spatial analysis and node selection: Characterisation of distributed connectivity by means of networks typically requires the definition of nodes involved in a functional network. Only thereafter the strength and (possibly) the direction of connectivity between these nodes can be estimated. One common approach is to define nodes on the basis of prior anatomical or functional relevance. An alternative is to define nodes on the basis of initial analysis of connectivity from the same or auxiliary samples, e.g. by means of spatial clustering of time-series. From a methods development perspective significant amount of effort has been given to different approaches to assess interactions and connectivity between nodes (especially on the analysis of time-series from resting state data). A much more understudied area is the effect of node selection itself and the spatial reliability of these nodes. This spatial reliability of the nodes in a network comes into play at various stages of analysis, e.g. during initial node selection (using a-priori or data derived regions-of- interest to serve as nodes) or after node selection at the stage of summarising the node-characteristic activity over voxels. Spatial reliability here focuses on how different methods of characterisation may affect the connectivity estimates, e.g. the estimated activity of a region might depend on parameters such as the width of the smoothing kernel and might lead to biased connectivity estimates. For analysis approaches where the nodes are intrinsically defined from the data (e.g. on the basis of clustering time-series dynamics), the spatial reliability is the reliability of the resulting connected areas. Obvious methodological questions arise as to the ability to define ‘stable’ regions that exhibit validity across the population(s). Furthermore, do different approaches lead to different spatial configurations of the nodes?
Characteristics of networks: Recent advances in the field of neuroscience now allows analysis of increasingly complex networks of brain regions. In these networks (i.e with 30 nodes or more) it is no longer possible to describe the connections of each node separately. Graph theoretical approaches allow us to describe these networks in terms of their characteristics, for example which nodes have the most number of connections with other nodes (centrality), which nodes are highly connected to each other but not to other nodes (modularity). In analyses based on graph theory, there are many different methods to estimate network characteristics. Questions arise on which methods are best suited for brain connectivity data, and how pre-processing steps impact the outcome of these methods.
This Research Topic focuses on issues of spatial node definition for connectivity analysis with a focus on (i) eluding on the consequences of specific choices of preprocessing on connectivity estimation and network characteristics, (ii) and suggesting solutions to these problems. Submissions on any analysis method incorporating spatial modelling, or explicitly testing reliability in the light of spatial variability are welcome, in addition to submissions incorporating analyses based on graph theoretical methods.
Spatial analysis and node selection: Characterisation of distributed connectivity by means of networks typically requires the definition of nodes involved in a functional network. Only thereafter the strength and (possibly) the direction of connectivity between these nodes can be estimated. One common approach is to define nodes on the basis of prior anatomical or functional relevance. An alternative is to define nodes on the basis of initial analysis of connectivity from the same or auxiliary samples, e.g. by means of spatial clustering of time-series. From a methods development perspective significant amount of effort has been given to different approaches to assess interactions and connectivity between nodes (especially on the analysis of time-series from resting state data). A much more understudied area is the effect of node selection itself and the spatial reliability of these nodes. This spatial reliability of the nodes in a network comes into play at various stages of analysis, e.g. during initial node selection (using a-priori or data derived regions-of- interest to serve as nodes) or after node selection at the stage of summarising the node-characteristic activity over voxels. Spatial reliability here focuses on how different methods of characterisation may affect the connectivity estimates, e.g. the estimated activity of a region might depend on parameters such as the width of the smoothing kernel and might lead to biased connectivity estimates. For analysis approaches where the nodes are intrinsically defined from the data (e.g. on the basis of clustering time-series dynamics), the spatial reliability is the reliability of the resulting connected areas. Obvious methodological questions arise as to the ability to define ‘stable’ regions that exhibit validity across the population(s). Furthermore, do different approaches lead to different spatial configurations of the nodes?
Characteristics of networks: Recent advances in the field of neuroscience now allows analysis of increasingly complex networks of brain regions. In these networks (i.e with 30 nodes or more) it is no longer possible to describe the connections of each node separately. Graph theoretical approaches allow us to describe these networks in terms of their characteristics, for example which nodes have the most number of connections with other nodes (centrality), which nodes are highly connected to each other but not to other nodes (modularity). In analyses based on graph theory, there are many different methods to estimate network characteristics. Questions arise on which methods are best suited for brain connectivity data, and how pre-processing steps impact the outcome of these methods.
This Research Topic focuses on issues of spatial node definition for connectivity analysis with a focus on (i) eluding on the consequences of specific choices of preprocessing on connectivity estimation and network characteristics, (ii) and suggesting solutions to these problems. Submissions on any analysis method incorporating spatial modelling, or explicitly testing reliability in the light of spatial variability are welcome, in addition to submissions incorporating analyses based on graph theoretical methods.
Keywords: FMRI connectivity, FMRI network, ROI selection, Network analysis, Graph metrics, missing ROIs, Accuracy of Graph parameters
Important Note: All contributions to this Research Topic must be within the scope of the section and journal to which they are submitted, as defined in their mission statements. Frontiers reserves the right to guide an out-of-scope manuscript to a more suitable section or journal at any stage of peer review.
