Name
Abstract | ||
|---|---|---|
We present a physical phantom designed for fibre tractography validation and use it to evaluate tracking algorithms that employ (a) the classic diffusion tensor model of diffusion, (b) high angular res- olution reconstruction of the diffusion orientation distribution func- tion (ODF), and (c) a regularization algorithm capable of inferring complex subvoxel fibre configurations. This work addresses four is - sues in diffusion MRI tractography: validation of the tracking pro- cess using ground truth, evaluation of different approaches for dif- fusion ODF reconstruction, coping with imaging noise, and coping with confounding subvoxel fibre configurations. Fibre tractography concerns the reconstruction of possible white matter pathways which form the connections through which neu- ronal signals are conveyed in the brain. The delineation of these path- ways is useful in determining whether specific areas of the brain are connected, the course of these connections, and how these change in disease. Prior to the advent of tractography using diffusion MRI, our knowledge of human neuronal connectivity was derived primarily from postmortem studies and tracer studies in nonhuman primates (1). The first work on 3D fibre reconstruction using diffusion MRI used diffusion tensor (DT) data, so named because it is obtained by modeling the diffusion probability density function (pdf) as an anisotropic 3D Gaussian function, which can be described by a sec- ond order tensor (2). The reconstruction of tracts was done by line propagation using the principal eigenvector of the diffusion tensor (3-5). Such principal diffusion direction (PDD) techniques can be confounded when there is more than one fibre direction within a single imaging voxel. This partial volume averaging of fibre direc- tions happens when tracts have high curvature, cross, branch, splay, or come to a bottleneck within an imaging voxel. The tensor case is a low angular resolution estimate of the diffusion pdf. High an- gular resolution approaches for estimating the diffusion orientation distribution function (ODF), which is obtained by integrating out the radial information in the pdf, include diffusion spectrum imag- ing (DSI) (6), q-ball imaging (7), and a composite hindered and re- stricted model of diffusion (CHARMED) (8). Here, we present a phantom designed for fibre tractography val- idation and present comparisons between fibre tracking algorithms using diffusion tensor and high angular resolution diffusion ODF re- construction. The phantom has known fibre structures that simulate common tract geometry in the human brain. Additionally, we present a regularization method that can be used for denoising ODFs with arbitrary angular resolution. The algorithm, which uses information from neighbouring voxels, can infer multiple maxima in the ODF obtained from diffusion tensor reconstruction, and can differentiate between cases of subvoxel partial volume averaging of directions that can confound even high angular resolution reconstructions: for example, crossing versus branching versus fibre pathways forming a bottleneck. |
Year | DOI | Venue |
|---|---|---|
2006 | 10.1109/ISBI.2006.1624925 | Arlington, VA |
Keywords | Field | DocType |
biodiffusion,biomedical MRI,image reconstruction,image resolution,medical image processing,noise,phantoms,classic diffusion tensor diffusion model,complex subvoxel fibre configurations,diffusion MRI tractography,diffusion orientation distribution function,fibre tractography validation,high angular resolution reconstruction,imaging noise,phantom,regularization,tracking algorithms | Iterative reconstruction,Computer vision,Diffusion MRI,Pattern recognition,Computer science,Imaging phantom,Angular resolution,Ground truth,Regularization (mathematics),Artificial intelligence,Tractography,Image resolution | Conference |
ISSN | ISBN | Citations |
1945-7928 | 0-7803-9576-X | 9 |
PageRank | References | Authors |
0.89 | 5 | 4 |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Jennifer S W Campbell | 1 | 255 | 18.91 |
| Peter Savadjiev | 2 | 144 | 12.06 |
| Kaleem Siddiqi | 3 | 3259 | 242.07 |
| G Bruce Pike | 4 | 699 | 132.31 |