The phenomenon of nuclear magnetic resonance was discovered in the 1940s, leading to the development of magnetic resonance (MR) imaging of human patients in the 1970s. The 1980s and 1990s were years of learning to adapt this tool to clinical use. Advances made its use easier and broadened the range of clinically useful information that could be obtained. The most commonly encountered types of MR imaging in clinical practice are the T₂-weighted images that provide visualization of many pathologies and the T₁-weighted images that show anatomy best. As researchers discover ways to visualize different aspects of tissue structure, new methods of MR imaging are developed. One such advance was the development of diffusion tensor imaging (DTI), a more sophisticated version of diffusion-weighted (DW) MR imaging.^+3—+5 Although this technique has been under study since the mid-1980s, it is now just beginning to find clinical usefulness.

DTI provides a way to examine the microstructure of the brain, particularly of the white matter.^+6,+7 It is based on the finding that water diffuses differently along (parallel to) than across (perpendicular to) axons (+Figure 1). Thus the direction and integrity of fibers can be assessed. A basic understanding of the principles of DTI will help the clinician follow the development of this technique as a useful clinical tool.

In clinical DW imaging, two scans are collected for each brain section. One scan is simply a standard T₂-weighted image. The second image is modified during collection to make it sensitive to water movement (diffusion) in the chosen direction. An analysis is then done of the signal intensity change from the first to the second image for every location (volume element; voxel) within the brain section. The differences between the two images are used to calculate an index of the average speed of water diffusion (apparent diffusion constant or coefficient; ADC) for each voxel in the image. DW MR imaging is quite sensitive to processes that alter the size of the extracellular space, such as the development of cytotoxic edema in an area of ischemia.^+7,+8 It has proven very valuable, particularly in the imaging of acute stroke.⁺⁹ However, DW MR images provide limited information about the direction of water diffusion.

In DTI a minimum of seven images is acquired for each brain section. As in DW MR imaging, one image is simply a standard T₂-weighted image as is used in clinical imaging. The rest of the images are modified during collection to make them sensitive to water movement in different directions. From the complete set of seven images, a matrix that describes diffusional speed in each direction is calculated for every voxel in the image. This matrix is the diffusion tensor and gives the technique its name. It can take 20 to 30 minutes to collect all the images, depending on the number of diffusion-sensitive images acquired. In gray matter, the speed of diffusion is usually similar in all directions. This is termed isotropic diffusion (+Figure 1, left). In white matter, the diffusion of water is significantly faster parallel to axons than across axons; it is directional. This is termed anisotropic diffusion (+Figure 1, right). It is not yet clear what factors about white matter contribute to the constrained (and therefore directional) diffusion of water, but it is present prior to myelination, although it does increase as myelin is formed.^+10—+12

Many quantities related to diffusion can be calculated from the tensor (matrix). The trace of the tensor provides a measure of average diffusion that is better than what can be obtained from simple DW imaging because it is not influenced by patient positioning or fiber orientation. Average diffusion is altered in areas of ischemia and may also be sensitive to gliosis. The degree of anisotropy within each voxel provides information on structural integrity of white matter. It can be used to identify areas of pathology or damage, as occur in multiple sclerosis or following traumatic brain injury. The principal direction of diffusion (eigenvector of the tensor associated with the largest eigenvalue) provides information about fiber direction. It can be used for mapping fiber tracts, which may be altered in areas of pathology or damage and by developmental changes. This technique also allows visualization of fiber tracts too small to be seen on conventional MR images.^+1,+13,+14 One way of displaying this information is by using directionally encoded color (see +Figure 3 and +Cover). The principal direction of diffusion in each voxel is represented by a color scheme in which a set color is assigned to each major direction (anterior—posterior, left—right, superior—inferior).⁺¹

While promising, DTI is relatively early in its development and still suffers from some significant weaknesses. Most commonly the very fast echo-planar MR method is used to collect the images. This method results in artifacts in areas of magnetic field inhomogeneity, such as interfaces between brain, bone, and air. DTI requires combining information from many images and therefore is quite sensitive to patient movement. Most commonly the imaging matrix acquired is rather coarse (96×96, 128×128). As a result, voxels are large relative to some of the white matter structures that are being examined, and so images suffer from partial volume artifacts (inclusion of a mixture of tissues or fiber tracts within a voxel). Refinements are under development that will allow higher resolution while maintaining adequate signal in each voxel.⁺¹⁵ In addition, the measures average what is occurring within each voxel. In areas in which fibers are crossing or diverging, the resultant average can be quite misleading.^{+7,+14,+16,+17} Techniques are being developed that extract measures of dispersion from each voxel, which should help with this problem.⁺¹⁴

Although still in its infancy, DTI has the potential to expand our knowledge of the brain many-fold. Currently, reports exist in the literature that demonstrate its potential use in both the acute and the chronic patient. Below is a short summary of selected disease processes and cases in which DTI proved helpful clinically in either diagnosis, prognosis, or treatment planning.

Jones et al. found changes in average diffusion similar to those found in acute stroke in the area surrounding sites of acute brain injury in all 4 cases they examined.⁺¹⁸ These areas appeared normal on standard T₂-weighted MR images, just as acute stroke does. The results suggest that these areas are ischemic and therefore might be responsive to treatment. The authors proposed that identification of such areas may lead to changes in acute treatment that renders those brain areas "potentially salvagable." Rugg-Gunn et al. presented 2 cases of chronic traumatic brain injury in which diffuse axonal injuries distant from the major injury site were present on DTI that were not visible on conventional MRI.⁺² Of great importance, the sites of diffuse axonal injury were consistent with the motor and neuropsychiatric deficits evident on examination. In another case report, the patient's excellent motor recovery by 18 months after traumatic brain injury correlated well with the preservation of normal anisotropy in a portion of the posterior limb of the internal capsule, an indication that the pathways were intact.⁺¹⁹ The authors suggested that this imaging method may well be able to differentiate between injuries that cause a transitory loss of function (as a result of temporary inflammation, edema, and/or shock) and permanent damage. If they are correct, this would allow patients to be separated into those who will recover function relatively quickly and those who will require extensive rehabilitation. Similarly, in another case, decreased anisotropy and increased diffusion were measured within the corticospinal tract for its entire length 18 months after cortical stroke.⁺²⁰ In contrast, 3 weeks after hemorrhage into the putamen, anisotropy was decreased in the internal capsule but average diffusion was within normal limits.⁺²⁰ These authors suggested that decreased anisotropy indicates axonal disruption, whereas increased diffusion indicates gliosis. Toxins can also alter the microstructure of the brain. A preliminary report of 15 patients with alcohol dependence indicates that DTI may provide insight into the anatomic basis of cognitive changes in alcoholism.⁺²¹

Mapping of alterations in fiber tracts that are related to particular functions is of primary interest in the application of DTI to the study of developmental disorders. Several groups have used DTI to test the theory that schizophrenia occurs as a result of frontal disconnection.^+22—+24 All three studies found decreases in the normal anisotropy of white matter, an indication of axonal disruption or disorganization. One found lower anisotropy in the prefrontal white matter of 5 patients with schizophrenia compared with normal subjects, as well as lower metabolic rates (as measured by positron emission tomography) in both frontal cortex and striatum.⁺²² In another study of 10 patients with schizophrenia, an overall decrease in anisotropy of white matter was found that was similar across all regions (prefrontal, temporal—parietal, parietal—occipital).⁺²³ A third group looked only at the corpus callosum in a group of 20 patients with schizophrenia.⁺²⁴ They found a reduction in anisotropy and an increased mean diffusivity in the splenium but not the genu of the corpus callosum in the patient group. DTI has also been used to show widespread abnormalities in tissue organization as a result of cortical maldevelopment with accompanying seizures.^+25,+26 Some of these areas of abnormal organization appeared normal on traditional imaging. Both authors note the significance of this in planning surgical correction of the accompanying epilepsy. DTI may also be sensitive to much more subtle white matter abnormalities. Adults with poor reading ability (previous diagnosis of developmental dyslexia) demonstrated decreased anisotropy in the left temporoparietal region that correlated well with reading skills.⁺²⁷ No abnormalities were visible on high-resolution T₁-weighted MR images.

Rose et al. proposed DTI as a means to identify Alzheimer's disease (AD).⁺²⁸ They imaged 11 patients who needed investigation for "dementia" and had been given a diagnosis of probable AD after meeting appropriate diagnostic criteria. When compared with 9 age-matched control subjects, the patients with probable AD demonstrated reduced anisotropy in the splenium of the corpus callosum, superior longitudinal fasciculus, and left cingulum (+Figure 3). Anisotropy of the splenium correlated well with the MMSE scores. The authors note that this region contains fibers that originated from the temporoparietal region—an area known to be affected in AD.

Ulug et al. included 2 patients with degenerative disease in their case series.⁺²⁰ In the patient with amyotrophic lateral sclerosis, there was decreased anisotropy with no change in diffusion in the posterior limb of the internal capsule, although the area appeared normal on clinical imaging. In the patient with progressive bulbar paralysis, there was little change in anisotropy but an increase in diffusion. The authors suggest that this latter pattern may be associated with gliosis. Two studies suggest that DTI may help with the staging of lesions in multiple sclerosis and may provide insight into underlying disease mechanisms.^+29,+30 Wieshmann et al. report the examination of a patient with seizures and a tumor of the right frontal lobe.⁺³¹ DTI imaging brought to light distant mass effect and displacement of white matter fibers adjacent to the tumor rather than destruction, a finding consistent with the patient's mild motor impairment. Again, the authors note the importance of this information in planning surgical interventions.

DTI is a powerful new imaging technique that provides a means to assess the integrity of white matter at the microstructural level. As the scattered case reports mentioned above indicate, it has broad applications in the study of both normal and abnormal brain development as well as acquired pathology. Some of these studies suggest an important role for DTI in guiding the planning of neurosurgical interventions based on the displacement or reorganization of fiber tracts around areas of pathology. If DTI can identify the extent of brain injuries and/or predict recovery potential, then TBI treatment might be improved during both the acute and chronic stages. If lesions can be documented in patients with disease not generally respected by third-party payers or disability examiners, then patients may be able to receive benefits once denied them. And last but not least, patients and families may be comforted by seeing a lesion or abnormality on a clinical film that explains their symptoms.