Components of the Human Connectome Project
Acquisition of HARDI Diffusion Data
When acquiring diffusion data one typically acquires an image where each pixel has information about diffusion along one direction and for some range of diffusivities. The full data consists of a set of such measurements designed so that taken together they contain all the information needed to infer on the fiber bundles passing through each voxel.
Since a single image measures diffusion only along one direction one needs to collect many images, each of which measures diffusion along a different direction. The more different directions that are measured the higher the accuracy by which one can determine the precise direction in which an axon is running. This is also very important when trying to resolve multiple fiber bundles passing through the same voxel. Acquisition schemes with many directions are known as HARDI (High Angular Resolution Diffusion Imaging) schemes.
It is also possible to manipulate the acquisitions so that different populations of water molecules are gauged. Two acquisitions can be performed where both images measure diffusion along the same direction but where one measurement reflects water molecules that move quite freely and where the other measures those restricted to move only very short distances. This can be extended to more than two measurements, where each measurement is weighted towards slightly different populations of water in terms of their diffusivity. By combining this with different directions one can sample the space of direction and diffusivity on a Cartesian grid. Such a scheme is called a DSI (Diffusion Spectral Imaging) scheme.
How should we acquire our data?
There is a tradeoff between angular resolution and obtaining information about different pools of water. A "single measurement" takes a few seconds to acquire. A full acquisition typically consists of several hundreds of such measurements and takes tens of minutes. There is an upper limit to how long a subject can remain still in the scanner before becoming uncomfortable and it is therefore crucial that the "best" acquisitions are performed in that time.
There are additional parameters that affect acquisition time that are also considered. Acquiring images with higher resolution also increases the precision of the tractography, but means that each measurement will take longer to perform. Similarly there are acquisition schemes that deal with image distortions, thereby avoiding a source of bias in the tractography, but which increase acquisition time.
Hence there is a tradeoff between angular resolution, diffusion-sensitivity, image resolution and distortions. A major initial effort of the HCP is the performance of extensive experiments to optimize this trade-off to achieve the most accurate tractography possible. This effort is continuing in parallel with the effort to improve the low level image acquisition, to provide the HCP with the highest possible quality MR diffusion data.