Seminar Series

DTI, distortions and directions

Diffusion tensor imaging is a rapidly evolving technique that allows us to map white matter tracts in vivo. It uses a special type of MR sequence that is sensitized to diffusion of water in a particular direction. For example if the acquisition of the image has been preceded by a sensitizing gradient in the left-right direction it means that in each and every pixel there will be a loss of signal that is proportional to how easily water diffuses in the left-right direction. By performing multiple measurements with sensitizing gradients in different directions one can build up a 3D profile of how easily water diffuses. If that profile turns out to be a ~sphere it means water diffuses equally easily in all directions, and from our perspective that is not so interesting. If, on the other hand, that shape turns out to look more like a cigar, i.e. with a clear direction in which water diffuses more easily, that means there is a white matter tract going through that pixel in the direction of the cigar. The result (after some processing) is an image where each pixel contains either a sphere (boring) or a little cigar pointing in some direction. One can then start at one of the cigars, go along its direction to the next pixel (with cigar) it points towards, change direction (slightly) to go in the direction of that new cigar etc etc. This process is know as "tractography" and aims at deducing the paths of major white matter bundles in the brain thereby providing information about what bits of cortex is connected to what other bits of cortex. Or to what subcortical structures. In this talk I will briefly outline the principles (MR-physics) of diffusion tensor imaging. I will discuss some of the shortcomings of the imaging technique, in particular the distortions that are associated with diffusion weighted images. I will talk about ways in which these can be corrected. I will then give a brief introduction to probabilistic tractography and talk a little about recent developments at the FMRIB regarding e.g. the crossing fibres problem.