The role of diffusion tensor imaging in the evaluation of ischemic brain injury - a review.

Water diffusion in brain tissue is affected by the presence of barriers to translational motion such as cell membranes and myelin fibers. The measured water apparent diffusion coefficient (ADC) value is therefore frequently anisotropic and varies depending upon the orientation of restricting barriers (such as white matter tracts) relative to the diffusion-sensitive-gradient direction. Anisotropic water diffusion can be specified using indices of diffusion anisotropy [e.g. standard deviation of the individual ADC values, fractional anisotropy (FA), lattice index (LI)], which are derived from measurements of the full diffusion tensor. The rotationally invariant nature of particular diffusion anisotropy indices (e.g. FA, LI) allows orientation-independent comparisons of these parameters between different subjects. Pathophysiological processes (such as cerebral ischemia) that modify the integrity of the tissue microstructure result in significant alterations in tissue anisotropy and make this metric a useful endpoint for characterizing the temporal evolution of the disease. Diffusion-tensor imaging (DTI) studies of both experimental and human stroke suggest that DTI may provide additional information about the evolution of the disease that is not available from diffusion-weighted MRI (DWI) alone. Acute reductions in the average diffusivity [<D> = (lambda(1) + lambda(2) + lambda(3))/3 where lambda(1), lambda(2), and lambda(3) are the eigenvalues of the diffusion tensor] following the onset of cerebral ischemia are often accompanied by increases in diffusion anisotropy. In the transition from acute to sub-acute and chronic stroke, <D> renormalizes and subsequently increases whereas diffusion anisotropy measures (e.g. FA) decline and remained reduced in chronic infarcts. Overall isotropic ADC changes during infarct evolution have been observed to be greater in white matter (WM) than in gray matter (GM) lesions (although there have been conflicting reports on this issue) and GM lesions tend to renormalize prior to WM lesions as the infarct evolves. Ischemic WM exhibits a significant decrease in diffusion anisotropy (relative to normal WM) during ischemic evolution whereas that of ischemic GM remains statistically unchanged. Furthermore, the percentage decrease in ischemic WM <D> is largely determined by reductions in lambda(1), the eigenvalue that coincides with the long axis of the WM fiber tract. Variations in unidirectional ADC or <D> over the ischemic time course limit the usefulness of this parameter alone as a predictor of ischemic injury. Consequently, ADC information has been combined with that of other MR parameters (including DTI) to unambiguously stage and predict ischemic brain injury over its entire temporal evolution. Combined <D> and diffusion anisotropy measurements have identified three phases of diffusion abnormality: (1) reduced <D> and elevated anisotropy; (2) reduced <D> and reduced anisotropy; and (3) elevated <D> and reduced anisotropy. However, variations in the differential patterns of <D> and diffusion anisotropy evolution have been observed by a number of investigators and more work is needed to clarify the role of these measurements in characterizing the severity of the ischemic insult as well as the potential outcome in response to the initial ischemic injury. The use of DTI, in combination with more sophisticated analysis methods for performing multiparametric segmentation, such as multispectral analysis, may enhance the use of MRI for accurate diagnosis and prognosis of stroke. Furthermore, these techniques may also play an important role in the clinical evaluation of new stroke treatments. Copyright 2002 John Wiley & Sons, Ltd.