Integrating DTI and WMT with Function

New work is emerging that attempts to do more than simply identify differences in DTI measures as a function of some important variable such as age, disease, or performance. In these studies, the question is: what are the implications of local variations in FA and/or tract characteristics for behavior and brain activity?

Three recent studies examining correlations of local variations in FA with reaction time have found conflicting results. In an ROI analysis, FA was correlated with reaction time in a target-detection task in young and older adults. The results suggested higher FA in the splenium in younger adults and higher FA in the internal capsule in older adults were related to faster reaction times (Madden et al. 2004). Conversely, and somewhat counter intuitively, a voxel-based analysis in a different target detection task revealed primarily positive correlations: high FA was associated with longer reaction times (Tuch et al. 2005), with the strongest effects in the optic radiations. Finally, in traumatic brain injury patients, FA was not correlated with reaction time or cognitive measures, although mean diffusivity did correlate with learning and memory scores (Salmond et al. 2006). Clearly more work is required to understand these relationships.

A more integrative strategy is to examine interactions among FA, BOLD fMRI responses, and behavior or some other external variable, such as age. The few studies attempting to do this have taken a hierarchical approach (e.g., Olesen et al. 2003; Baird et al. 2005). In the first step behavior-FA and behavior-BOLD relations or BOLD activations are assessed separately, effectively reducing the analysis space by creating ROIs from significant clusters. The second step then examines BOLD-FA relations in the smaller subset of regions.

Alternatively, one could ask whether specific tracts are related to behavioural differences. Beaulieu, et al. (2005) used a voxel-based analysis to correlate FA with reading ability in a group of healthy children. The novel aspect to this work was that the authors then used the direction of the principal eigenvector in significant clusters as seeds for WMT. This allowed them to identify potential tracts passing through the significant clusters. They were able to demonstrate that the largest cluster was more likely associated with a tract not expected to be related to language processing (Fig. 18).

Finally, a number of studies have incorporated diffusion data with the results of fMRI activation studies.. The most common approach has focused on using activated clusters as starting points for tractography to identify anatomical connections. As in any tractography exercise, the choice of which activated voxels to use as seeds for tractography can result in substantially different tracts (Guye et al. 2003). The dependency of tract trajectory on the seed point chosen is compounded by the fact that significant BOLD responses are primarily measured in gray matter, which has generally has low anisotropy, and may be some number of voxels away from highly anisotropic white matter. Since regions of low anisotropy are typically excluded from fibre tracking algorithms, the user must select from nearby voxels with high FA for seeding the WMT. Because of this added uncertainty, it is even more critical to evaluate the robustness of identified tracts. Some progress in tracking between and through gray matter regions has been achieved through the use of probabilistic tractography methods that have been optimized to traverse regions of low anisotropy (e.g., Behrens and Johansen-Berg 2005).

That there is some correspondence between functional and anatomical regions has been recently shown by the Oxford group (Johansen-Berg et al. 2004). In this study, SMA (supplementary motor area) and preSMA were identified in each subject using tasks known to activate those areas independently. Probabilistic tractography was then applied to generate path probabilities from each of the two brain regions. The authors were able to show that separate groups of regions were connected to each of the BOLD regions, with little overlap, as would be expected based on known anatomy. They have recently expanded this analytical approach to show that the functional separation of

Fig. 18. (a) FA in the purple cluster of voxels (arrow) correlated with reading ability. Fibre tracking indicated this cluster was in the posterior limb of the internal capsule (b), and not in tracts more commonly associated with language (superior longitudinal fasiculus, in green; or superior fronto-occipital fasciculus, in yellow). (Reprinted from Beaulieu et al. 2005, with permission from Elsevier)

Fig. 18. (a) FA in the purple cluster of voxels (arrow) correlated with reading ability. Fibre tracking indicated this cluster was in the posterior limb of the internal capsule (b), and not in tracts more commonly associated with language (superior longitudinal fasiculus, in green; or superior fronto-occipital fasciculus, in yellow). (Reprinted from Beaulieu et al. 2005, with permission from Elsevier)

these two regions across subjects is more closely aligned to commonalities in local fibre architecture in adjacent white matter than to structural similarities based on conventional T1-weighted images (Behrens et al. 2006). As the authors point out, they do not yet know if similar relations will hold in other cortical regions. Additionally, the scan time needed to acquire the high resolution DTI dataset (45min) is not amenable for routine applications. However, the possibility for describing common patterns of functional activations based on common features in the properties of the underlying fibre architecture would be an important adjunct for understanding similarities and differences in brain connectivity.

It is important to keep in mind that DTI tractography is simply defining a model system for brain connectivity. The choice of a particular seed point will influence the derived tracts because of the inherent noise in the data acquisition and the sensitivity of the chosen algorithm to this noise. Tractog-raphy is blind to whether the seed point derives from a functional activation or from a well-placed ROI based on expert anatomical knowledge. Therefore, the tracts indicate only the possibility of an anatomical connection between a set of regions; tracts based on functional activations carry no additional "meaning" relative to those derived based on anatomical knowledge. Methods such as those being developed by the Oxford group (e.g., Behrens et al. 2006) will allow for refined anatomical models, but then the task will be to move beyond describing the possibility for information flow to describing how and when information is conveyed along the identified connections.

To fully understand brain function requires more than defining functional "blobs" correlated with some task or behavior. Methods for identifying neural systems and evaluating their interactions have been around for quite some time. Some of the earliest work examined functional connectivity using interregional correlation analyses (e.g., Clark et al. 1984; Horwitz et al. 1984); these were followed with more explicit systems-level analyses of functional and effective connectivity (e.g. Friston et al. 1993; Horwitz et al. 1999; McIntosh 2000), and more recently methods such as dynamic causal modeling (Friston et al. 2003). The importance of moving beyond identifying regions that correlate with some task or behavior has been reemphasized recently by Stephan (2004), who nicely illustrated how two brain regions can correlate independently with a task condition, but have no correlation between themselves (Fig. 19).

The possibilities for incorporating diffusion and other quantitative MRI data into analyses of functional and effective connectivity are many. However it is critical to recognize that simply demonstrating that a pathway exists between two regions that are separately related to some task or behavior does not imply nor guarantee that the identified path mediates the activity between those regions. A more fruitful strategy may be to concurrently determine the existence of pathways between functionally connected regions, forming the basis for models of effective connectivity. Regardless of how paths are identified, the information conveyed along those paths should be measured and assessed. Some common and readily available modeling techniques available

Fig. 19. A) Region Ai (red dotted line) and region A2 (green dashed line) are each correlated with the "task" (blue, solid line) at r = 0.73. B) Scatterplot showing that while the correlation of each voxel with the task is high (green, r = 0.73), the correlation between the two voxels is low (magenta, r = 0.07). Adapted from Stephan, 2004

Fig. 19. A) Region Ai (red dotted line) and region A2 (green dashed line) are each correlated with the "task" (blue, solid line) at r = 0.73. B) Scatterplot showing that while the correlation of each voxel with the task is high (green, r = 0.73), the correlation between the two voxels is low (magenta, r = 0.07). Adapted from Stephan, 2004

for assessing effective connectivity are reviewed in (Mcintosh 2000; Penny et al. 2004; Ramnani et al. 2004; Stephan et al. 2004; Stephan et al. 2005) See also chapters by Bressler and Mcintosh, Sporns and Tononi, and Stepan and Friston in this volume. Perhaps the most important contribution from diffusion and other qMRI techniques will come from their ability to provide additional anatomical and physiological constraints to the models. Thus, the confidence that a fibre exists, its length, diameter, "integrity", and myelin content are all important contributions to the regulation of information flow between two regions. Incorporating this information into systems-level analyses of functional imaging data will greatly enhance our understanding of brain function.

Essentials of Human Physiology

Essentials of Human Physiology

This ebook provides an introductory explanation of the workings of the human body, with an effort to draw connections between the body systems and explain their interdependencies. A framework for the book is homeostasis and how the body maintains balance within each system. This is intended as a first introduction to physiology for a college-level course.

Get My Free Ebook


Post a comment