Diffusion Tractography: Methods, Validation and Applications in Patients with Neurosurgical Lesions

Diffusion tensor imaging (DTI) tractography is increasingly used in presurgical mapping in tumors located in eloquent areas since it is the only non invasive technique that permits in vivo dissection of white matter tracts. Concordance between the DTI tracts and subcortical electrical intraoperative mapping is high, and DTI tractography has proven useful to guide surgery. However, it presents limitations due to the technique and the tumor, which must be known before using the images in the operative room. This review focuses on the possibilities and limits of DTI imaging in intraoperative tumoral mapping and presents an overview of current knowledge.

In patients with surgical lesions in functional areas, the purpose of surgery is to achieve maximal tumor removal while preserving essential brain functions. Although intraoperative cortical and subcortical direct electrical stimulations (DESs) remain the gold standard to map the functional boundaries of the resection cavity, the use of DESs can be limited by time constraint. DESs in an awake patient are time consuming and the progressive patient’s tiredness explains that only a limited number of tasks can be performed during surgery. In this context, preoperative localization of cortical and subcortical functional areas is crucial. Preoperative evaluation based on anatomic knowledge is insufficient because white matter fiber tracts can be displaced by the tumor and language pathways present high interindividual variability.

Diffusion tensor imaging (DTI) tractography is the only noninvasive technique that permits in vivo dissection of white matter tracts. Tractography can now be easily performed using clinical magnets and therefore is increasingly used to map fiber tracts for surgical planning. Tractography provides a unique anatomic information by reconstructing and visualizing chosen fiber tracts in the 3-dimensional (3D) anatomy of a patient. Motor, language, and visual tracts and their relationship with the tumor can then be visualized preoperatively and integrated with a surgical neuronavigation system to guide surgery. Yet, DTI tractography still presents limitations because of the technique and the tumor. This review focuses on the possibilities and limits of DTI imaging in preoperative tumoral mapping and provides an overview of the current knowledge.

Possibilities and limits of diffusion tensor brain mapping

DTI

Physical basis

Diffusion imaging is based on the self-diffusion of water molecules in tissues.

In a biologic environment, water molecules present with random thermally driven motions that are so-called Brownian motion. When water molecules diffuse equally in all directions, as in the ventricular cerebrospinal fluid (CSF), the diffusion is called isotropic. In brain white matter, the micrometric movements of water molecules are hindered to a greater extent in a direction perpendicular to the fiber orientation than parallel to it. Therefore, diffusion is higher following the direction of fiber bundles. Hindrance is attributed to multiple factors, including myelination, axon density and diameter, and axonal membrane integrity. Diffusion is then not equal in all 3 orthogonal directions, a property called anisotropy. To fully determine the direction of diffusion, a tensor is used, which describes the mobility of water molecules along each direction. Diffusion-weighted images have to be acquired along at least 6 directions along with an image acquired without diffusion weighting. Diffusion anisotropy is subsequently characterized using several indices made of combinations of the terms of the diffusion tensor, that is, the eigenvalues λ ₁, λ ₂, and λ ₃, which characterize the main diffusion directions and associated diffusivities. The most popular measure is fractional anisotropy (FA), which ranges from 0 (isotropy) to 1 (maximum anisotropy). DTI enables to extract the tridimensional orientation of the underlying fibers within each voxel and to quantify the motional anisotropy.

Diffusion tensor color-coded maps and tractography

Diffusion anisotropy can be displayed in several ways. In FA maps, signal intensity codes for the degree of anisotropy (ranging from 0–1). Isotropic (CSF) or weakly anisotropic voxels (gray matter) usually present an FA less than 0.2, whereas anisotropic white matter presents an FA greater than 0.2. However, FA maps do not provide information on diffusion direction. The white matter tract organization is therefore better represented using ellipsoids or directionally color-coded schematic maps of major diffusion orientation. Ellipsoids are 3D representations of the diffusion distance covered in space by molecules in a given time. In color-coded maps, different colors (red, green, or blue) are attributed to different fiber orientation along the 3 orthogonal spatial axes ( Fig. 1 ). Also, the brightness of each color is modulated by the degree of anisotropy.

Fiber tractography enables 3D visualization of fiber bundles. Because water diffusion is preferentially oriented along the direction of fiber tracts, tracts are reconstructed by using this directional property of water diffusion in each voxel. A large number of DTI tractography algorithms have been proposed to reconstruct fiber tracts. The most commonly used algorithm in clinical practice is a deterministic algorithm based on the FACT (fiber assignment by contiguous tracking) algorithm. With this method, tracking is performed on a voxel-by-voxel basis. The entire track is determined from a seed point after the successive orientations associated with adjacent voxels. Tractography necessitates the definition of a seed region of interest (ROI) located on the path of the investigated fiber tract to initiate the fiber tracking procedure. The termination of line propagation is determined by 2 parameters corresponding to stopping thresholds: (1) when a voxel is less than a predetermined minimum FA value, typically 0.2, or (2) when the maximum angle of the main diffusion orientation between 2 contiguous voxels is reached, typically 30° ( Fig. 2 ).

DTI in brain tumors

Reconstructing DTI fiber tracts for tumoral presurgical planning is challenging because in the vicinity of the tumor, fibers can be displaced, infiltrated (by the tumor and/or edema), or destroyed ( Fig. 3 ). The tumoral cells and the peritumoral edema cause changes in the brain structure. Typically, measurement of diffusion anisotropy in the normal brain parenchyma up to near the tumor demonstrates a decrease in FA values ( Fig. 4 ). The reduction in anisotropy can result from fiber depletion (tumor destroys fibers, reducing their absolute numbers), fiber dilution (tumor or vasogenic edema spreads intact fibers apart, reducing their density), or fiber degradation (fibers themselves become intrinsically less anisotropic, retaining normal numbers and density). Reduced anisotropy explains why DTI tractography can be limited in a tumoral environment ( Fig. 5 ).

DTI Artifacts and Limitations

It is important to keep in mind that DTI reconstructed fibers and tracts are not the visualization of an anatomic structure but an estimation of these structures via the diffusion anisotropy. Therefore, there are several limitations inherent to DTI and tractography techniques, which must be known before interpreting the information given by orientation color maps and reconstructed fiber tracts. These limitations have already been described extensively.

Crossing fibers

In DTI acquisitions, each voxel can contain a large number of fibers. The estimate of orientation of the tensor is therefore an average of the orientations of all the fibers contained within the voxel. When the fibers are coherently organized in a parallel fashion, the orientation of the tensor truly reflects the orientation of the underlying fibers. However, when fibers cross within a voxel, the voxel-averaged estimate of orientation cannot accurately summarize the orientation of the underlying fibers.

This explains that DTI tractography fails to reconstruct some existing fibers. For example, the corticospinal tract has a fan-shaped morphology in the centrum semiovale. However, DTI fiber tracking most often reconstructs only fibers originating from the vertex. This reconstruction is attributable to the existence of multiple crossing fibers in the centrum semiovale, which leads to inaccuracy in the estimation of the direction anisotropy in these voxels.

This problem can be overcome with other diffusion techniques, such as the high angular resolution diffusion imaging (HARDI), which has the ability to extract different orientations of fibers in voxels containing multioriented fiber populations. Several algorithms can be used to reconstruct diffusion orientations from HARDI images, the most popular being Q-ball reconstruction. Q-ball can distinguish distinct fiber populations in each voxel and reconstruct portions of tracts that could not be evidenced with DTI.

The main limitation of HARDI in clinical practice is the length of the scan acquisition because it requires a large number of gradient directions and higher b values.

Spurious fibers

Acquisition noise and crossing fibers can induce a corrupted estimate of orientation in a voxel from which a spurious aberrant fiber can be generated. These fibers have the same appearance as the fibers of the expected tract, except that they are anatomically incorrect ( Fig. 6 ). Determining whether some fibers are spurious or anatomically coherent can be difficult. Strong a priori knowledge of fiber tract anatomy provides valuable information in this context.