Fig. 15. Left: Optic radiation variability as a function of threshold used to define connectivity (n=22). Right: Mean FA decreased as optic radiation ROI size became larger and more dispersed, but the relation to BOLD response in visual cortex was similar. (Reprinted from Toosy et al. 2004, with permission from Elsevier)
line with random-field theory (Kiebel et al. 1999). It is not yet clear whether smoothing is appropriate for analysis of DTI data, but the size of the smoothing filter can dramatically affect residual errors and the sensitivity to detect group-wise differences (Jones et al. 2005b). In the latter study, significant FA differences between schizophrenic patients and controls were either not found, or were localized to superior temporal sulcus (STS), STS and cerebellum, or cerebellum only. This variability was due only to the size of the smoothing filter, and indicates the reasons for the choice of a specific smoothing filter should be specified.
Alternative methods for voxel-based studies have focused on registering the tensor directly (Xu et al. 2003) or tensor components (Park et al. 2003). Another approach is to use iterative registrations of FA maps to create study-specific templates (Toosy et al. 2004), as is frequently done with voxel-based-morphometry analyses (Good et al. 2001). Finally, a new method has been suggested where non-linear registration is used as the first step in aligning all subjects' FA images together; peak FA "ridges" on are found on the group-averaged FA template, creating a skeleton of the dominant WM tracts. Subject-specific FA values are then derived by finding the location in each subject's data that most closely matches the spatial location of the ridge (Smith et al. 2006, Fig. 16). This approach appears to be robust against residual misregistration since only peak FA values (corresponding to probable tract centers) are analyzed. The use of approaches that attempt to ensure better alignment of tracts across subjects or provide more robust estimates of tract-specific DTI parameters such as FA are critical to furthering our understanding of how alterations in brain connectivity affect brain function and behavior.
Tractography. Obviously, the ability of white matter tractography to estimate patterns of brain connections in vivo has piqued the interest of the
Fig. 16. (A) Example of an FA skeleton on a coronal FA map. The outlined region includes the cingulum bundle, corpus callosum, fornix, ventricles and thalamus and is shown in B-E. (B) FA skeleton is shown in blue, and significant differences between a group of controls and schizophrenics are in red. (C) Voxel-based analysis found additional differences at the lower edge of the ventricles (arrow). (D,E) Examination of the separate group-mean FA maps indicates this spurious finding was produced because the larger ventricles in the patient group (E) were not in register with the controls (D). Note that the corpus callosum was well-registered, and the location of FA differences more closely matched the skeletonized FA results. Images courtesy of S. Smith
Fig. 16. (A) Example of an FA skeleton on a coronal FA map. The outlined region includes the cingulum bundle, corpus callosum, fornix, ventricles and thalamus and is shown in B-E. (B) FA skeleton is shown in blue, and significant differences between a group of controls and schizophrenics are in red. (C) Voxel-based analysis found additional differences at the lower edge of the ventricles (arrow). (D,E) Examination of the separate group-mean FA maps indicates this spurious finding was produced because the larger ventricles in the patient group (E) were not in register with the controls (D). Note that the corpus callosum was well-registered, and the location of FA differences more closely matched the skeletonized FA results. Images courtesy of S. Smith neuroscience and neuroimaging communities. It is currently the only non-invasive method for reconstructing white matter trajectories in the human brain. Detailed and careful studies using white matter tractography will potentially reveal important information about brain connectivity. However, the links between tractography results, which provide information about anatomical connectivity, and measures of functional and/or effective connectivity (see below) have not yet been clearly established. Several potential anatomical measures that could influence connectivity may be derived from tractography, including the volume, length and/or cross-sectional area of the reconstructed tracts, but these are not routinely applied.
WMT has several potential applications. (1) WMT offers the unique ability to non-invasively visualize the organization of specific WM pathways in individual subjects (e.g., Fig. 11). To date, most studies of white matter neu-roanatomy have been conducted using either anatomic dissection methods or axonal tracer studies in animals. The majority of tractography studies have focused on well-known and readily identifiable WM pathways such as the cortico-spinal tract, the corpus callosum and optic radiations. Many of these studies have demonstrated that WMT can generate tract reconstructions that are consistent with known neuroanatomy (e.g., Mori et al. 1999; Stieltjes et al. 2001; Catani et al. 2002; Jellison et al. 2004; Wakana et al. 2004). Recent WMT studies have moved beyond tracking prominent bundles and have attempted to determine the utility of WMT to distinguish between direct and indirect connections (Catani et al. 2003) and whether highly curved pathways near CSF can be mapped with confidence (Concha et al. 2005b). A common criticism of WMT is that the validation of these results are missing. Two approaches have been applied to address this concern - histopathological measurements and WMT have been compared in animal models (e.g., Bürgel et al. 2005; Ronen et al. 2005); and measures of WMT confidence have been developed and applied to provide an estimate of the reliability of specific tractography results. It should also be noted that most neuroimaging results must be interpreted without validation. Thus it is critical to establish the reliability and repeatability of any new WMT method (e.g., Ciccarelli et al. 2003a; Ding et al. 2003; Heiervang et al. 2006). (2) WMT may be used to parcellate specific WM pathways or portions of WM pathways (see Fig. 17). This will enable tract-specific measurements such as tract volume, cross-sectional dimensions, and the statistics of quantitative measurements within the pathways such as mean diffusivity and FA. Several studies have used WMT to perform measurements in specific WM pathways: e.g., fronto-temporal connections in schizophrenia (Jones et al. 2005a; Jones et al. 2006); pyramidal tract development in newborns (Berman et al. 2005), and the pyramidal tracts and corpus callosum in multiple sclerosis (Vaithianathar et al. 2002). Concurrently, progress has been made in the development of tract-specific group templates, which may be useful for voxel-based analyses (Ciccarelli et al. 2003b; Burgel et al. 2005; Johansen-Berg et al. 2005; Thottakara et al. 2006). (3) WMT may be used to visualize specific white matter patterns relative to pathology including brain tumors, M.S. lesions, and vascular malformations. The increased
Fig. 17. Parcellation of major white matter pathways using white matter tractography in a single subject. Superior longitudinal fasciculus (red); corpus callosum (purple ); inferior occipital fasciculus (light blue); inferior longitudinal fasciculus (yellow); uncinate fasciculus (orange); fornix/stria terminalis. (dark orange); corona radiata (green)
specificity of WM trajectories may ultimately be useful for planning surgeries (Holodny et al. 2001; Henry et al. 2004) as well as following the patterns of brain reorganization after surgery (Lazar et al. 2006). However, it should be noted that WMT reconstructions still need further validation before advocating their use as a tool for surgical guidance on a widespread basis. Indeed one recent study demonstrated that their WMT method underestimated the dimensions of the specific tract of interest (Kinoshita et al. 2005). Other studies have started to examine the relationship between specific white matter tracts affected by multiple sclerosis lesions and specific clinical impairments (Lin et al. 2005).
Was this article helpful?