PASTA: Pointwise Assessment of Streamline Tractography Attributes Derek K. Jones, 1,2 * Adam R. Travis, 2,3 Greg Eden, 2 Carlo Pierpaoli, 2 and Peter J. Basser 2 Diffusion tensor MRI tractography aims to reconstruct nonin- vasively the 3D trajectories of white matter fasciculi within the brain, providing neuroscientists and clinicians with a potentially useful tool for mapping brain architecture. While this technique is widely used to visualize white matter pathways, the associ- ated uncertainty in fiber orientation and artifacts have, to date, not been visualized in conjunction with the trajectory data. In this work, the bootstrap method was used to determine the distributions of diffusion indices such as trace and anisotropy, together with the uncertainty in fiber orientation. A novel visual- ization scheme was developed to encode this information at each point along reconstructed trajectories. By integrating these schemes into a graphical user interface, a new tool which we call PASTA (Pointwise Assessment of Streamline Tractography At- tributes) was created to facilitate identification of artifacts in trac- tography that would otherwise go undetected. Magn Reson Med 53:1462–1467, 2005. Published 2005 Wiley-Liss, Inc. † Key words: DT-MRI; artifacts; distribution; visualization; hyper- streamline In diffusion tensor MRI (DT-MRI), a set of diffusion-weighted (DW) images is collected and used to form estimates of the self-diffusion tensor in each voxel of the imaged volume (1). From the diffusion tensor, one obtains estimates of mean diffusivity, diffusion anisotropy, and fiber orientation. This information has been used in various attempts to infer con- nectivity within the brain using a variety of algorithms that are generically referred to as tractography (e.g., 2– 6). To date, tractography research has focused predominantly on obtain- ing images that show the trajectories of individual white matter fasciculi. The reconstructed trajectories are repre- sented either as streamlines (e.g., 2,3) or illuminated stream- tubes (e.g., 7,8). However, in such representations, which we term here “trajectory-only” visualizations, there is no indica- tion of either the possible variability of diffusion quantities along the tract or the reliability of the tract reconstruction In this work, we describe an integrated approach not only to visualizing the trajectory of white matter fasciculi but also to indicating how diffusion indices, including mean diffusivity and anisotropy indices, vary along the tract. Furthermore, we show how to visualize uncertainty in fiber orientation at each point along the reconstructed tract and the distribution of diffusion indices at each ver- tex. This visualization proves to be extremely useful in identifying artifacts in the data that would otherwise go unnoticed in trajectory-only visualizations. As we effec- tively perform a pointwise assessment of streamline trac- tography attributes, we refer to this whole approach as PASTA. METHODS Acquisition Diffusion-weighted magnetic resonance imaging data were acquired from healthy volunteers on a 1.5-T GE Signa LX system (General Electric, Milwaukee, WI, USA) with 40 mT/m gradients. The acquisition was gated to the cardiac cycle using a peripheral gating device placed on the subjects’ forefinger. A multislice peripherally gated EPI acquisition sequence was used, providing nearly isotropic resolution (1.7  1.7  1.7 mm) and coverage of the whole head. Eighty-four contiguous axial slice locations with isotropic resolution were acquired using the dual-gradient scheme (9,10). The basic acquisition using this scheme consists of 1 image with no gradients applied and then 6 images in which gradients are applied along the vector directions [+1, +1, 0], [+1, -1, 0], [+1, 0, +1], [+1, 0, -1], [0, +1, +1], [0, +1, -1]. This acquisition sequence was repeated 16 times, resulting in 112 images per slice location. We refer to the data set con- sisting of all 112 DW images per slice location as the “super- set.” Following motion/distortion correction (11), the diffu- sion tensor was computed in each voxel and diagonalized to compute the eigenvectors and eigenvalues. The fractional anisotropy (12) was then computed in each voxel and, to- gether with the principal eigenvector, used to create a color- encoded fiber orientation map (13). Tractography To perform tractography, as well as to be able to obtain repeated estimates of the tensor field at subvoxel locations, a continuous representation of the diffusion tensor field was established using the B-spline approach described by Pajevic et al. (14). This approach provides an extremely rapid method for obtaining subvoxel estimates of the dif- fusion tensor field. Using the color-encoded maps, voxels lying within a fasciculus of interest were identified and fiber tracking was initiated from these “seedpoints” using a tractography algorithm akin to that of Basser et al. (4) with a step size of 0.5 mm. Tracking was terminated when the fractional anisotropy (FA) fell below 0.20. Note that, since we ultimately aimed to visualize uncertainty in fiber orientation, in contrast with the approach of Basser et al. 1 Centre for Neuroimaging Sciences, Institute of Psychiatry, London, United Kingdom. 2 Section on Tissue Biophysics and Biomimetics, Laboratory of Integrative Medicine and Biophysics, National Institute of Child Health and Human De- velopment, National Institutes of Health, Bethesda, Maryland, USA. 3 Department of Biomedical Engineering, Vanderbilt University School of En- gineering, Nashville, Tennessee, USA. Grant sponsor: Wellcome Trust; Grant sponsor: Whitaker Foundation. *Correspondence to: Derek K. Jones, Centre for Neuroimaging Sciences, Institute of Psychiatry, P089, De Crespigny Park, London SE5 8AF, United Kingdom. E-mail: d.jones@iop.kcl.ac.uk Received 15 September 2004; revised 15 December 2004; accepted 29 December 2004. DOI 10.1002/mrm.20484 Published online in Wiley InterScience (www.interscience.wiley.com). Magnetic Resonance in Medicine 53:1462–1467 (2005) Published 2005 Wiley-Liss, Inc. † This article is a US Government work and, as such, is in the public domain in the United States of America. 1462