Diffusion Tensor Imaging and Functional Tractography

The Science of Diffusion Tensor Imaging and Tractography

The concepts of diffusion-weighted imaging and DTI are based on measuring the random (Brownian) movement of water molecules. Water molecules in different environments (e.g., intracellular vs. extracellular water) have different diffusivities that may be influenced by structural barriers (e.g., cell membranes) and the molecular densities of tissue. To measure this diffusivity, magnetic resonance imaging (MRI) scanners exploit the loss of signal between two gradient pulses (usually 20 to 50 ms apart) as water moves; therefore areas with high diffusion have low signal and areas of restricted diffusion have higher signal. Diffusivity can be described as isotropic ( Fig. 3.1A ) or anisotropic ( Fig. 3.1B ). Isotropic diffusion is the same in all directions, as would be the case in areas of fluid without a microstructure, such as cerebrospinal fluid. On the other hand, tubular structures, such as neurons, have anisotropic diffusion, where water diffuses more freely along the axis of the neuron than in the other two perpendicular axes.

DTI sequences are designed to measure the diffusivity of water molecules in each voxel in at least six different gradients (directions), which are amalgamated to build up a tensor describing the magnitude of diffusivity, termed eigenvalues, in each of three orthogonal dimensions, termed eigenvectors . Using these three eigenvalues and eigenvectors, a number of descriptive properties of the tissue can be described including the fractional anisotropy (FA), axial diffusivity, radial diffusivity, and mean diffusivity ( Fig. 3.1C ), properties that may serve as useful markers of tissue characteristics such as fiber diameter, fiber density and myelination. Of these, FA is probably the widely used. FA ranges between 0 (isotropic diffusion) and 1 (anisotropic diffusion), with high FA values associated with white matter tracts.

Mathematical algorithms can then be used to map these white matter tracts. These mathematical algorithms can be broadly split into two different techniques. The original methods (including FACT and TEND ^, ) used deterministic modelling, which assigns each voxel with a directional orientation and uses the orientations of nearby voxels to map tracts. Newer probabilistic methods consider an orientation distribution function for each voxel, thereby generating a statistical likelihood map for the location of white matter tracts. ^{, ,} Although the deterministic method is more commonly used by commercial software and intraoperative neuronavigation systems, it is susceptible to areas of low FA, as can be the case with peritumoral edema or areas where there are intersecting fiber tracts in multiple directions. Probabilistic mapping is limited by its computational intensity, often taking days on traditional machines; the advent of graphic processing units has allowed probabilistic tractography to be sped up. Both methods require cautious interpretation as they may be susceptible to producing false-positive and false-negative tracts and do not have the ability to assign directionality (afferent vs. efferent) to the tracts.

Both types of mapping require a starting “seed” voxel that can either be performed by the user based upon anatomically preserved foci or be based on relevant areas of cortex identified through fMRI. For example, the corticospinal tract can be identified by selecting the cerebral peduncles as an anatomically preserved focus or by identifying areas of the primary motor cortex using an fMRI finger-tapping task.

The output of this processing, the estimation of the location of white matter tracts, can be presented via color-coded maps on two-dimensional slice MR projections or be integrated into three-dimensional reconstructions to portray fiber pathways ( Fig. 3.2 ).

Rationale for Diffusion Tensor Imaging in Surgical Neuro-oncology

The field of surgical neuro-oncology has placed significant focus on the so-called oncofunctional balance with the aim of maximizing the extent of resection (EoR) of intra-axial brain tumors while preserving neurologic function and quality of life. Although the importance of EoR has been established in numerous studies of high- and low-grade gliomas, it has also clearly been shown that causing a neurologic deficit affects outcome adversely. More recent developments have even advocated for supratotal resection of gliomas, switching from a paradigm of “image-guided” resection to “functionally guided” resection.

There are multiple techniques in a surgeon’s armamentarium that can be used in pursuit of this oncofunctional balance ( Table 3.1 ), but novel MRI modalities have undoubtedly been central to progress.

Modern intraoperative neuronavigation systems enable the coregistration of tractographic maps to traditional structural imaging sequences, thus facilitating pre- and intraoperative visualization. A basic understanding of the location and function of main white matter tract systems is a prerequisite for the use of DTI in surgical neuro-oncology ( Table 3.2 ).