Imaging connectivity: MRI and the structural networks of the brain

Abstract

Magnetic resonance imaging (MRI) is a flexible and widely available neuroimaging technique. Structural MRI and diffusion MRI, in particular, provide information about connectivity between brain regions which may be combined to obtain a picture of entire neural networks, or the so-called connectome. In this review we outline the principles of MR-based connectivity analysis, discuss what relevant information it can provide for clinical and non-clinical neuroscience research, and outline some of the outstanding needs which future work will aim to meet.

Figures

Figure 1

Simulation of diffusion within an ideal impermeable cylinder, shown in cross-sections perpendicular (A) and parallel (B) to the axis of symmetry. Despite starting from the center of the figure in both cases, diffusing molecules progress further on average along the cylinder than across it.

Figure 2

Illustration of the principle of streamline tractography based on dMRI data. The principal direction of diffusion at each voxel location in the data is shown as a line, whose color corresponds to its orientation. The streamline, shown in white, is generated by beginning at a seed point (large white circle) and repeatedly stepping along the principal direction.

Figure 3

Key stages in the creation of a structural connectivity graph using magnetic resonance images: cortical parcellation (A), whole-brain tractography (B), and the final graph representing the pattern of connections between regions (C).

Figure 4

Two simple graphs with the same number of nodes and connections, but different topologies. Graph A has higher “efficiency”, because a maximum of three connections need to be traversed to get from any one node to any other, compared to up to seven for graph B. However, if either of the central “hub” nodes, 5 and 10, were to be destroyed and removed from the graph, a larger proportion of nodes in graph A would be disconnected. Graph B is therefore, in some senses, less vulnerable to attacks targeted at these nodes.