Scientific Principles of DTI

Diffusion MRI scans image the molecular diffusion of water at the same scale as cellular dimensions and therefore allow the microarchitecture of the brain to be investigated. The constant random motion of molecules is described by Brownian motion and is exploited by diffusion imaging to specifically detect the displacement of water molecules through the brain tissue medium. Diffusion-weighted scanning consists of a T2-weighted spin-echo sequence with the addition of two diffusion-sensitizing gradients applied before and after the 180^o refocusing pulse, through an identical axis. Therefore, there is a loss of signal intensity as a result of incomplete rephasing of water proton spins after they have moved during the time elapsed between the two diffusion-sensitizing gradients.²⁰ Diffusion times in the region of 10 to 50 ms are used which provides microscopic detail, capturing average molecular displacements of 10 μm.²⁰ Scan acquisition using standard MRI systems takes 3 to 10 min,²¹ and therefore is minimally burdensome on patient, radiographer, or scanner time.

The direction of the passage of water is different depending on the nature of tissue in which it is found. Where no structural boundaries exist nearby, the molecular motion of water is unimpeded and equal in all directions. This is known as isotropic diffusion. This is exhibited within the cerebrospinal fluid spaces of the brain, with the exception of sites of bulk flow such as the aqueduct of Sylvius or foramen of Munro.²⁰ Isotropic diffusion is also believed to occur in grey matter.^22,23 In contrast, myelinated white matter fiber tracts are arranged into parallel, densely packed bundles that impede the diffusion of water molecules perpendicular to the fibers’ direction. Therefore, diffusion of water molecules in this situation is not equal in all directions and is defined as anisotropic diffusion. Detection of water molecule anisotropy is the basis of diffusion tensor imaging and tractography.

The diffusion tensor is a 3 × 3 matrix of vectors which mathematically describes the three-dimensional (3D) directionality and magnitude, or diffusion anisotropy, of water molecules.^20,21,24 The three principal axes of the diffusion tensor are termed eigenvectors. When plotted as an ellipsoid, isotropic diffusion is a sphere whereas anisotropic diffusion forms an elongated ellipsoid, becoming a prolate (cigar) shape when the eigenvector of greatest magnitude is much larger than the other two. Prolate diffusion within a brain voxel is assumed to represent a white matter fiber bundle where the primary eigenvector is aligned with the axonal axis. Tracing of white matter tracts to produce functional tractograms uses each voxel’s diffusion tensor to link it to adjacent voxels and in this way trace out the likely path of a fiber bundle in 3D space (Fig. 3-1).

FIGURE 3-1 Main eigenvectors in adjacent voxels providing the basis for tractogram construction (A). Fractional anisotropy map derived from diffusion tensor image (B). Examples of seed points are seen overlying both thalami.

DTI fiber tract data can be presented in two forms. Functional anisotropy maps provide information on fiber anatomy in cross-sectional two-dimensional (2D) images with color-coded axes where the brightness is proportional to the degree of anisotropy (see Fig. 3- 1). By convention, the anteroposterior axis is represented by green, left-right by red, and up-down by blue. Therefore, the corpus callosum will appear red, for example. Alternatively, deterministic or probabilistic²⁵ functional tractography performs a 3Ddimensional reconstruction and portrayal of the fiber pathways based on following a white matter tract from voxel to voxel as described above (Fig. 3-2). Specified anatomic points, known as “seeds” (see Fig. 3-1), can be selected by the user from where the tractogram can be plotted by the processing software to delineate proposed neural connectivity with the selected site. Alternatively, larger volumes of brain can be selected as regions of interest or “masks.” To reduce dependence on the user and therefore the inherent subjectivity of seed selection while also increasing the likelihood of depicting functionally relevant tracts, Schonberg et al. used functional MRI to define where seed points should be sited.²⁶ Although this represents an extra stage of patient assessment, they found that it enabled a more comprehensive mapping of fiber systems such as the pyramidal tract and the superior longitudinal fasciculus. See glossary of terms in Table 3-1.

FIGURE 3-2 Tractogram representing the ascending and descending pathways among cortex, brainstem, and cerebellum after seed selected in pons.

(Courtesy of Prof. Peter Silburn and Dr. Terry Coyne, University of Queensland, Brisbane, Australia.)

Table 3-1 Glossary of Diffusion Tensor Imaging Terms

Preoperative Planning Applications

DTI shows the surgeon the relationship of the intra-axial tumor to local white matter tracts in multiple planes. A variety of aspects of the tumor–tract relationship can therefore be demonstrated. The identity of the tract can be surmised from its position and course, such as the corticospinal tract and optic radiations. The proximity of the tumor to the tract can be appreciated. Also the position of the tumor can be seen in relation to the tract, for example superior, lateral, medial etc., allowing optimal approach to be determined to highly eloquent and complex areas such as the pons.²⁷ Displacement of the tract by the tumor or edema can also be demonstrated.^28,29 This is crucial information when planning a surgical trajectory in order to avoid eloquent tissue. DTI has been found to provide important preoperative warning of this surgical hazard in situations where a precisely planned trajectory is imperative such as during resection of thalamic juvenile pilocytic astrocytoma with displacement of the posterior limb of the internal capsule.³⁰ Incorporation of white matter fibers within the tumor mass, seen especially in low-grade tumors,³¹ and destruction of white matter fibers by the tumor can also be depicted. These features will have profound implications for the extent of resection amenable for the individual tumor.

DTI can also help elucidate the anatomy of poorly described pathways in vivo in the human to inform and advance established surgical strategies. Resection strategies that aim to excise normal as well as neoplastic tissue with a view to minimizing the likelihood of recurrence such as frontal and temporal lobectomy can be enhanced by DTI to maintain safety. The anatomico-functional connectivity of the dominant temporal lobe, for example, was reviewed by Duffau et al. using a combination of DTI and subcortical intraoperative stimulation studies to elucidate the white matter pathways, which should represent the resection boundaries of temporal lobectomy such as the pyramidal tract and the anterior wall of the temporal part of the superior longitudinal fasciculus.³² Indeed, DTI has been proposed as the preoperative investigation to assess individual patients’ risk of visual field defect prior to anterior temporal lobe resection as it images the Meyer loop of the optic radiation as it courses anteriorly from the lateral geniculate nucleus and around the tip of the temporal horn before projecting to the visual cortex. The individual variation of this white matter pathway³³ increases the risk of a deficit which can permanently disqualify the patient from holding a driving license. Therefore, preoperative warning of a more ventral position of the Meyer loop along its course anterior to the temporal horn should identify those with a higher likelihood of postoperative deficit.^34,35

Although DTI visualizes white matter tracts, it is possible to extrapolate these projections and therefore visualize their grey matter cortical projections/origins. Kamada et al. applied this technique to map the primary motor area (PMA) preoperatively in thirty patients with supratentorial lesion affecting the motor system.³⁶ By selecting seed points within the corticospinal tract at the cerebral peduncle, plus the medial lemniscus to differentiate from somatosensory projections, a PMA map was produced that was successfully validated against subsequent intraoperative cortical somatosensory evoked potentials. Indeed, fMRI failed to identify the PMA in eight patients. The reasons for this were inherent in the patients’ pathology through its effect on the motor system in that they were incapable of successfully completing the self-paced finger tapping task required to elicit the blood oxygenation level dependent signal that fMRI detects. In contrast, DTI requires no patient tasks to acquire its data and therefore offers an important alternative for preoperative non-invasive cortical mapping in patients who, for whatever reason, cannot complete them.

DTI has been applied in neuro-oncology beyond not only functional mapping but for noninvasive assessment of tumor architecture in terms of cell density, white matter invasion and even histologic discrimination such as the distinction between primary and secondary intra-axial tumors.^37–41 It has been proposed that DTI can distinguish between vasogenic edema and tumor-infiltrated edema. Edematous tissue surrounding glioma is generally accepted to be infiltrated by tumor cells. In contrast, edema surrounding cerebral metastases or meningioma is considered to be vasogenic.^40,42 Therefore, hyperintensity surrounding tumor on T2-weighted MRI may reflect any of glioma, metastasis or meningioma. However, as the FA at the voxels corresponding to the site of edema has been shown in some studies to be of a lower value in infiltrative pathologies such as glioma, a tumor infiltration index was derived by Lu et al. to help distinguish against pathologies producing only vasogenic edema.³⁹ There have been contradictory reports including a PET-labeling study questioning whether this DTI analysis is specific enough to differentiate between tumor-infiltrated edema and vasogenic edema.^41,43 Further investigation will determine whether DTI can fulfill this potential and provide reliable presurgical histologic tumor characterization.

Therefore, DTI provides advanced warning of potential intraoperative misadventures to help surgeons adapt their approach subsequently in theater to minimize these. Even prior to this stage, DTI can inform the surgeon of how amenable the tumor is to surgery by virtue of its relationship to eloquent brain and even potentially its histologic nature, and therefore what surgery can offer in terms of likely benefits and associated risk of adverse effects.

Intraoperative Neuronavigation

The prospect of intraoperative tract navigation is possibly the most exciting application of DTI for the surgeon. It is logical to expect that an intraoperative map of functionally important subcortical tracts should reduce the likelihood of inadvertent straying into white matter pathways, reassure the surgeon to be more aggressive allowing optimization of resection limits or, conversely, restraint when critical tracts are close leading to an associated reduction in the incidence of postoperative neurologic deficits and increase in tumor volume reduction. A number of studies have sought to establish whether current DTI techniques fulfill this promise.

Various neuronavigation systems capable of integrating DTI with frameless stereotaxy exist. The DTI FA sequence can be subjected to predefined thresholds to delineate in three dimensions the white matter tracts of interest and the surgeon can then also manipulate the final renditions using the drawing tools available in the software. A standard 3D stereotactic neuronavigational MRI sequence is selected to provide the reference images for navigating and the DTI sequence is selected to provide the working images to be merged with it. The intensity of the tractograms can be altered to optimize the prominence of the tracts with respect to the structures of interest, usually the tumor, and the surgeon’s preference. Standard patient registration and navigating strategies are then employed as in conventional navigation. Depending on the surgeon’s preferences and the particular software and hardware facilities available, the navigational display can be presented on the workstation beside the patient or projected through the microscope’s heads-up display.

The feasibility of intraoperative guidance by incorporation of DTI fiber tracking into neuronavigation systems has been demonstrated by a number of investigators within the last decade.^44–49 White matter pathways such as the pyramidal tract and the optic radiation were successfully portrayed in relation to intra-axial tumors such as cavernoma and glioma. Coenen et al. were the first to report the use of intraoperative neuronavigation with 3D tract reconstruction to assist the resection of glioblastoma associated with the pyramidal tract. They found fiber tract navigation to be a helpful adjunct to resection in the four patients in whom it was applied. Subsequent studies have also underlined its potential for efficiency and patient safety.^47,48 The ability of DTI to reliably predict the true location of critical white matter pathways intraoperatively is crucial for the technique to be applied with confidence during surgery. Investigators have evaluated intraoperative DTI’s accuracy in depicting motor pathways by comparing it to intraoperative electrophysiologic methods, in particular cortical stimulation.^50,51 One particular study of 13 patients employed electrical motor cortex stimulation to verify the location of the precentral gyrus and indirectly the pyramidal tract, DTI neuronavigation correctly predicted the principal motor pathways’ position in 92%.⁵¹ DTI has not only been applied to surgery for supratentorial tumors but also to the resection of brainstem lesions such as cavernoma with promise of improving operative safety.⁵²

The perceived benefits of fiber tract neuronavigation need to be translated into objective improvements in aspects of patient care; however, few studies have addressed this rigorously with objective endpoints. Notably, Wu et al. performed a prospective, randomized controlled trial to attend to this deficiency in the literature.⁵³