Portal de Periódicos da CAPES

Diffusion and Perfusion MRI in Epilepsy

2002; Wiley; Volume: 43; Issue: s1 Linguagem: Inglês

10.1046/j.1528-1157.2002.043s1069.x

1528-1167

David C. Alsop, Alan Connelly, John S. Duncan, A. Hufnagel, Carlo Pierpaoli, Fergus Rugg‐Gunn,

MRI in cancer diagnosis

EpilepsiaVolume 43, Issue s1 p. 69-77 Free Access Diffusion and Perfusion MRI in Epilepsy David C. Alsop, David C. Alsop Harvard Medical School, Boston, Massachusetts, U.S.A.;Search for more papers by this authorAlan Connelly, Alan Connelly Radiology and Physics Unit, Institute of Child Health, UCL, andSearch for more papers by this authorJohn S. Duncan, John S. Duncan National Society for Epilepsy MR Unit and UCL, London, England;Search for more papers by this authorAndreas Hufnagel, Andreas Hufnagel Department of Neurology, University Hospital, Essen, Germany; andSearch for more papers by this authorCarlo Pierpaoli, Carlo Pierpaoli NICHD, National Institutes of Health, Bethesda, Maryland, U.S.A.Search for more papers by this authorFergus J. Rugg-Gunn, Fergus J. Rugg-Gunn National Society for Epilepsy MR Unit and UCL, London, England;Search for more papers by this author David C. Alsop, David C. Alsop Harvard Medical School, Boston, Massachusetts, U.S.A.;Search for more papers by this authorAlan Connelly, Alan Connelly Radiology and Physics Unit, Institute of Child Health, UCL, andSearch for more papers by this authorJohn S. Duncan, John S. Duncan National Society for Epilepsy MR Unit and UCL, London, England;Search for more papers by this authorAndreas Hufnagel, Andreas Hufnagel Department of Neurology, University Hospital, Essen, Germany; andSearch for more papers by this authorCarlo Pierpaoli, Carlo Pierpaoli NICHD, National Institutes of Health, Bethesda, Maryland, U.S.A.Search for more papers by this authorFergus J. Rugg-Gunn, Fergus J. Rugg-Gunn National Society for Epilepsy MR Unit and UCL, London, England;Search for more papers by this author First published: 06 February 2002 https://doi.org/10.1046/j.1528-1157.2002.043s1069.xCitations: 20 Address correspondence and reprint requests to Dr. A. Connelly, Radiology and Physics Unit, Institute of Child Health, University College London, 30 Guilford Street, London WC1N 1EH, U.K. E-mail: [email protected] AboutSectionsPDF ToolsRequest permissionExport citationAdd to favoritesTrack citation ShareShare Give accessShare full text accessShare full-text accessPlease review our Terms and Conditions of Use and check box below to share full-text version of article.I have read and accept the Wiley Online Library Terms and Conditions of UseShareable LinkUse the link below to share a full-text version of this article with your friends and colleagues. Learn more.Copy URL Diffusion and perfusion imaging offer new insights into the microstructural organization and the delivery of blood to the brain. In this article we review the recent technical developments of diffusion and perfusion imaging of the brain using magnetic resonance (MR) methods and the application and limitations of these techniques to the study of those with epilepsy. THE PRINCIPLES OF MR DIFFUSION IMAGING The study of the diffusion properties of brain tissue water using MR has provided important information regarding tissue status in the context of cerebral ischemia. After earlier diffusion studies of induced seizures in animal models (1–4), diffusion imaging is now being applied to the investigation of patients with epilepsy. Diffusion describes the random microscopic translational motion of molecules (brownian motion), in which the net mean displacement of a population of molecules remains zero. In a large population of freely diffusing molecules, such as in a mobile liquid, the "diffusion volume" would be expected to be spherical, with the radius of the sphere increasing as the square root of the diffusion time. Diffusion behavior in an in vivo environment, however, is more complex, and is dependent on the nature of any barriers to diffusion. The signal intensity on an MR image can be made dependent on the rate of water diffusion in the brain. In its simplest form, this can be achieved by placing a pair of equivalent magnetic field gradient pulses on either side of the refocusing radiofrequency pulse in a spin-echo (SE) image. The resulting MR image intensity would be weighted according to the rate of water diffusion, with high-diffusivity regions giving relatively low signal intensity, whereas regions of restricted diffusion would give a relatively high signal intensity. However, diffusion-weighted (DW) images are highly sensitive to the effects of brain motion, such that conventional multishot images have severe degradation when DW gradients are added. There are several potential solutions to this problem, the most common of which are (a) very rapid single-shot imaging [e.g., DW echo-planar imaging (EPI)](5), and (b) the use of navigator echo correction of SE or other multishot images (6–8). Although the latter can potentially produce higher-quality images, this is still an area of development; therefore, single-shot DW-EPI is by far the most common method in use at present, with the concomitant limitation of image distortions in regions of high magnetic susceptibility gradients such as the temporal lobes. The degree of DW in an MR image is dependent (9) on a combination of the strength of the diffusion gradient (G), the length of time the diffusion gradient is on (δ), and the time between the front edges of the gradients (Δ). These contributions to DW are collectively described by a "b-value," where b = γ2G2 δ2(Δ − δ/3) The higher the value of b, the greater the degree of DW. In DW imaging (DWI), the image intensity is weighted not only by water diffusivity, but also by proton density and the relaxation times, T1 and T2. The contribution of diffusion can be isolated by means of quantitation [i.e., by calculation of the apparent diffusion coefficient (ADC)]. This term is used because the measured diffusion is dependent not only on water diffusivity, but also on factors such as barrier permeability and diffusion time. MR DIFFUSION TENSOR IMAGING AND TRACTOGRAPHY Physical basis of DTI Isotropic diffusion can be described by a scalar quantity D, such that the DW signal intensity is given by S DW = S0 exp(−bD) where S0 is the non-DW signal intensity. D can be determined by making two measurements at different b values, with the diffusion gradient in any arbitrary direction (because diffusion is isotropic). This is not the case for anisotropic diffusion, in which the water diffusivity is different in different directions; this can confound data reproducibility but also may provide further information. Diffusion in an anisotropic system cannot be described by a simple scalar (as earlier), but is characterized by a 3 × 3 diffusion tensor matrix, D in which the diagonal elements indicate the molecular mobility in orthogonal directions, and the off-diagonal elements express how diffusion in one direction is correlated with displacement in a perpendicular direction. In effect, rather than diffusivity being characterized by a sphere (as in the isotropic case), anisotropic diffusion is characterized by the size, shape, and orientation of a diffusion ellipsoid, with three potentially unequal axes representing different diffusivities in orthogonal directions. In tissues with randomly composed microstructure, water diffusivity appears to be the same in all directions (isotropic diffusion). However, in tissues with highly organized microstructure, such as brain white matter, the measured diffusivity of tissue water varies with the tissue's orientation (anisotropic diffusion) (10,11). Therefore, to obtain a reproducible measure of diffusion in the brain, it is not meaningful to calculate the ADC in a single direction. The axes of the cells (or their diffusion ellipsoids) will be arbitrarily oriented with respect to the fixed x, y, and z axes of the magnet, resulting in a misleadingly inhomogeneous (and orientation-dependent) DW signal intensity. Diffusion tensor MRI (DTI) provides all of the information required to construct a diffusion ellipsoid in each voxel of an imaging volume (12). Specifically, it provides a measurement of the effective diffusion tensor, D, in each voxel of an imaging volume. From D it is possible to calculate new quantitative scalar and vector-valued parameters that characterize specific features of the diffusion process. These scalar quantities are designed to be rotationally invariant (i.e., independent of the coordinate system in which the MR measurement is made, the orientation of subjects in the magnet, and many of the specifics of the experimental design). The most basic rotationally invariant quantities are the three principal diffusivities (or eigenvalues) of D, which are the principal diffusion coefficients measured along the three (intrinsic) coordinate directions that constitute the local "fiber" frame of reference in each voxel. Each eigenvalue is associated with a principal direction (eigenvector) that also is intrinsic to the tissue. The three eigenvectors of D are mutually perpendicular and define a local fiber frame of reference in which the description of diffusion in that voxel is the simplest and most natural. In each voxel, these eigenvalues can be sorted in order of decreasing magnitude (λ1, highest diffusivity; λ2, intermediate diffusivity, and 13, lowest diffusivity). In anisotropic tissues consisting of ordered parallel bundles, the largest eigenvalue, λ1, represents the diffusion coefficient along the direction parallel to the fibers (D‖), whereas λ2 and λ3 represent the diffusion coefficients (D⊥ and D⊥') in the transverse directions (12). The eigenvectors of the diffusion tensor provide unique directional information that can be used to infer interesting features of living tissue. In particular, the eigenvector associated with the largest eigenvalue, λ1, represents the direction parallel to a bundle of fibers within a voxel. We can use measurements of this eigenvector in each voxel to construct vector maps of white-matter fiber direction within an imaging volume (13,14). Moreover, from this discrete 3-dimensional (3D) vector field, one can calculate trajectories of different fiber tracts in a manner similar to that which is used in fluid mechanics to obtain fluid streamlines from a discrete velocity vector field (15–18). A rotationally invariant scalar measure that characterizes the bulk diffusion properties of the tissue, independent of the orientation of the fibers, is Trace(D). (The trace of a matrix is the sum of the diagonal elements [i.e. Trace(D) = Dxx+ Dyy+ Dzz]. This quantity is proportional to the orientationally averaged diffusion coefficient (an average of diffusion coefficients measured in all spatial directions). It can also be shown to be the arithmetic average of the three eigenvalues of D (λ1, λ2, and λ3). Other widely used rotationally invariant scalars measure different features of anisotropic diffusion. These include the Relative- (19), Fractional- (19), and Lattice Anisotropy indices and the Skewness of the eigenvalues (20). The first three measure the degree to which the diffusion ellipsoid's shape deviates from being spherical, whereas the fourth measures whether the ellipsoid is prolate (cigar-shaped) or oblate (pancake-like). Prolate water-displacements profiles are typically found in white-matter regions with parallel arrangement of fibers, such as the corpus callosum and the pyramidal tract. Oblate ellipsoids correspond to white-matter regions having a particular architectural arrangement of fibers, such as sheets of parallel fibers with different orientations, or bundles of fibers that are randomly oriented in a plane. Oblate water-displacement profiles are found in the centrum semiovale and subcortical white matter regions (13). Biologic and radiologic findings using diffusion tensor imaging DTI is an imaging modality that combines features of in vivo anatomic MRI and histopathology. This is possible because water-diffusion properties in tissues, as measured by DTI, are affected by tissue constituents, such as macromolecules, membranes, organelles, as well as by tissue microstructure, architecture, and organization (1). A DTI study furnishes microstructural and architectural information that can not be obtained with conventional, composition-based (e.g., proton density and chemical spectroscopy) or relaxometry-based (e.g., magnetization transfer, T1, T2, and T2*) MRI methods. Important results have been obtained in the assessment of cerebrovascular diseases, in particular in acute ischemia, in which tissue water diffusivity decreases shortly after the onset of the ischemic event (21). This finding has allowed the efficacy of stroke-protective agents to be evaluated in vivo in clinical trials (22). Trace(D), calculated from DTI methods, has become the most commonly used diffusion parameter to measure water-diffusion changes in brain parenchyma, and Trace(D) has been found to be quite uniform in normal brain parenchyma across diverse brain regions (13). This finding makes Trace(D) a particularly good parameter to use to detect structural abnormalities in vivo because its value appears to be so tightly controlled under normal conditions. Measurements of diffusion anisotropy have been critical for the advancement of a number of clinical applications, particularly in the diagnosis and evaluation of neurodegenerative disorders. For example, brain regions having a homogeneous signal intensity in T1- or T2-weighted images may have heterogeneous water-diffusion displacement profiles depending on how the architectural arrangement of the fibers changes within the region (Fig. 1)(13). DTI anisotropy measurements also have been particularly powerful in evaluating the loss of white matter in wallerian degeneration after a focal lesion (23–25). Figure 1Open in figure viewerPowerPoint Conventional T2-weighted image (a) and maps of quantities computed from a DT-MRI study (b, c, and d) in an axial section of the human brain. Trace (D) is homogeneous in normal brain parenchyma. In the anisotropy map bright voxels correspond to regions where water diffusion depends upon direction (anisotropy), whereas dark voxels indicate regions where water diffusion is the same in all directions (isotropy). In the fiber orientation map, different colors are associated with different fiber tract orientations as indicated in the color circle shown in the bottom right of the figure. The left-right, anterior-posterior, and superior-inferior directions are associated with pure red, green, and blue, respectively. Obliquely oriented fibers are represented by colors resulting from a combination of red, green, and blue. While identifying the location and orientation of white matter pathways is virtually impossible in the T2-weighted images, DT-MRI fiber orientation mapping provides a clear depiction of different pathways. DTI is the only available method that allows mapping of white-matter fiber orientation in vivo and noninvasively. In the brain, fiber-direction mapping is useful to identify and differentiate anatomic white-matter pathways that have similar structure and composition but different spatial orientation (14,26–28). Historically, such studies of the brains structural anatomy could be performed only by using histologic methods (29). Several groups have presented white-matter orientation maps of the human brain that clearly depict the main association, projection, and commissural white-matter pathways. Even in the brainstem, which is a relatively small anatomic structure, motor and sensory pathways can be easily identified and can be differentiated from the transverse pontine fibers and the cerebellar peduncles (28). Recently DTI has been used to study connectivity noninvasively in the living human brain. There are still many open questions about whether this goal can be fully achieved by currently available methods, mainly because inferring continuity of particular pathways from diffusion measurements in regions where different fiber bundles cross is problematic. Nevertheless, robust algorithms for fiber tracking with DTI data have been proposed recently and can be used to explore the potential usefulness of DTI in assessing brain connectivity and degenerative changes of neural tissue (15–18). Significant diffusion changes occur in brain development, and several groups have demonstrated that changes in the cyto- and myeloarchitecture of the developing brain are accompanied by changes in the diffusion properties of tissue water (30,31). For instance, DT-MRI has been used to follow the development of white-matter tracts in neonates, and to visualize the nonmyelinated fibers of the corpus callosum as early as 28 weeks in vivo (30). In conclusion, there is evidence that DTI provides unique information about brain anatomy, microstructure, organization, and architecture. APPLICATION OF INTERICTAL DIFFUSION IMAGING The mechanism for signal change in DWI is not fully understood. DWI studies in ischemia have shown an early decrease in ADC, progressing at a later stage to a high ADC if infarction occurs (32). Animal models suggest that the decrease in ADC in ischemia is associated with cell swelling in conjunction with energy failure below a critical perfusion threshold (11). DWI investigations of animal models of epilepsy have reported restricted diffusion, associated with seizure activity. Status epilepticus induced by the administration of either bicuculline (2) or flurothyl (3) has been shown to result in reduced ADC, that was partially reversible in the latter by barbiturate administration. Cortical electroshocks in rats also have been shown to lead to reduced ADC, even in the absence of epileptic afterdischarge (4). The epilepsy models have important differences from ischemia models, in that the former showed no reduction in ATP, and no hypoperfusion, whereas in the latter, these appeared to be contributory factors. However, a common feature to both models is the presence of cell swelling, suggesting that the reduction in diffusion may result from an increase in intracellular water, or a reduction in the extracellular space. Two interictal DWI studies have shown increased ADCAV and decreased AI in the epileptic, atrophic hippocampus in patients with mesial TLE (32,33). In contrast, ADCAV was reported to be decreased interictally in a group of 21 patients with nocturnal frontal lobe epilepsy and normal conventional MRI, compared with healthy controls and with patients with generalized idiopathic epilepsy. This study, however, measured whole brain mean diffusivity and diffusion measures that were not orientation independent (34). Relatively few studies in patients with epilepsy have used DTI. In a study of three patients after temporal lobe resection, one of whom developed a homonymous hemianopia postoperatively, DTI demonstrated significantly reduced anisotropy in the relevant optic radiation of the affected patient compared with the other two patients and a control group of 22 normal subjects. The hypothesis was that postoperative wallerian degeneration resulted in myelin loss with less restricted diffusion transverse to the fiber bundles and hence reduced anisotropy (35). A single patient with subcortical nodular heterotopia was studied with a number of imaging sequences. DTI revealed a heterogeneous structure of the malformation and demonstrated displacement of a major white-matter tract by the malformation. Neither feature was evident on conventional imaging. Interestingly the part of the malformation with the most abnormal diffusion was associated with the lowest NAA (N-acetyl aspartate) concentration, but was distinct from the site of fMRI activation (36). Two articles examined DTI and epileptogenic structural abnormalities. The first studied 10 patients, each with a hemiparesis as a result of one of a number of different supratentorial lesions, for example, infarcts, cerebral injury, and malformations of cortical development (MCDs). In all, reduced anisotropy was noted in the relevant corona radiata and in patients with severe impairment also in the cerebral peduncle. Structural disorganization as a result of the abnormality and wallerian degeneration in the cerebral peduncle were the suggested causes for the reduced anisotropy (37). The second study examined 18 patients with structural lesions, including infarcts, injuries, MCDs, and tumors. Anisotropy was reduced in all lesions, and diffusivity was increased in ∼70%. In particular, the majority of MCDs had normal mean diffusivity. This suggests that despite the loss of structural/directional organization in MCDs, cell density is relatively preserved (38). Most studies have investigated only structural abnormalities in single patients or small series using a region-of-interest approach. Voxel-by-voxel analysis is more objective, and in addition, it is possible to evaluate the whole brain rather than just specific regions. Investigating patients with localization-related epilepsy and normal conventional MRI (MRI-negative patients) is important because surgical treatment without an abnormality on preoperative imaging is often associated with a poor outcome. The hypothesis that DTI analyzed on a voxel-by-voxel basis would identify areas of abnormal diffusion [fractional anisotropy (FA) and mean diffusivity (MD)] has been addressed in a study of 22 patients with MCDs, 10 patients with epileptogenic acquired lesions, and in 30 MRI-negative patients. A standard DTI scanning protocol was used that gave a resolution of 2.5 × 2.5-mm in-plane and 5-mm slice thickness. The diffusion maps were calculated for each subject according to the method proposed by Pierpaoli et al. (39). Using SPM96 (Wellcome Dept. of Cognitive Neurology, Institute of Neurology, London, U.K.), each subject's map was then normalized to Talairach space, which corrected for differences in rotation, shear, scale, and position between individuals, thereby allowing meaningful analyses. Subsequently the maps were smoothed, and each patient was statistically compared on a voxel-by-voxel basis with 30 control subjects, and significant increases or decreases were detected at an individual voxel threshold of 0.001, which was corrected for multiple comparisons to 0.05 (40). DTI analyzed with SPM was sensitive in patients with acquired cerebral damage, identifying significant increases in diffusivity and significant reductions of anisotropy in all patients (41). Areas of reduced anisotropy were found in 17 patients, with MCDs and areas of increased diffusivity in 10 (42). Increased or abnormally located grey matter, pathologic white matter with abnormal myelination, or defective neurogenesis and cell loss may have resulted in abnormal anisotropy or diffusivity in these patients. In addition, these new methods conferred additional sensitivity by detecting abnormal diffusion in cerebral tissue that appeared normal on conventional imaging in patients with both acquired lesions and MCDs. Individual analyses of the 30 patients with partial seizures and normal conventional MRI identified areas of significantly abnormal diffusion in nine patients. Seven of these concurred with the localization of epileptiform EEG abnormality (Fig. 2). Group analysis of nine MRI-negative patients with electroclinical seizure onset localizing to the left temporal region revealed a significant increase of diffusivity and a significant reduction in anisotropy within the white matter of the left temporal lobe. Significant differences in the diffusion indices in individual MRI-negative patients and the group effect in patients with left TLE suggests that minor structural disorganization may exist in occult epileptogenic cerebral lesions. This could be due to either etiologic factors, for example, occult dysgenesis, or chronic seizures, for example, atrophy, gliosis, and expansion of the extracellular space. These techniques are promising, noninvasive imaging methods for identifying cerebral areas associated with partial seizures, and with improvements in DTI and analysis methods, further occult epileptogenic regions may be identified in individual patients before more invasive diagnostic procedures and possible epilepsy surgery. Figure 2Open in figure viewerPowerPoint Diffusion tensor imaging of a patient with right frontal lobe epilepsy and normal conventional MRI. Normalized axial diffusivity maps at the same slice position for the averaged 30 control subjects (a) and the patient (b and c). Note that the difference in signal to noise between the maps is due to averaging of the 30 control subjects. The region of significantly increased diffusivity is superimposed on map (c). The region of increased diffusivity is localized to the normal-appearing cerebral tissue of the right frontal lobe. The equivalent slice of the patient's T1-weighted image (d). Right on the images is patient's right. PERFUSION MRI IN EPILEPSY Perfusion MRI (43) offers advantages in spatial and temporal resolution over the widely used nuclear medicine perfusion technique of single-photon emission computed tomography (SPECT) using Tc-HMPAO. To the extent that perfusion and metabolism are well coupled, perfusion MRI also provides an alternative to 18FDG positron emission tomography (PET). Because MRI can also provide additional structural and functional information, a complete structural/functional MRI (fMRI) examination is attractive in the evaluation of patients with epilepsy. However, because all currently available MRI perfusion methods use short-lived tracers, imaging of ictal events requires the patient to already be in the MRI scanner. This makes ictal imaging more challenging than that with SPECT or even PET, unless seizures are frequent. Perfusion MRI can be performed either with the bolus injection of a contrast agent or without any injection. The contrast-agent bolus approach measures the time varying attenuation of T2 or T2* images caused by the passage of an agent with high magnetic susceptibility through the vasculature (44). This method is often referred to as dynamic susceptibility contrast (DSC) MRI. The contrast-free technique uses the spatial selectivity of radiofrequency and gradient pulses to perturb the magnetic properties of water nuclei selectively within the inflowing arterial blood (45–47). Because this electromagnetic perturbation effectively labels the endogenous arterial water spins, it is known as arterial spin labeling (ASL). Dynamic susceptibility contrast MRI All clinically approved MR contrast agents do not cross the normal blood–brain barrier and can be considered intravascular contrast agents in most studies of epilepsy. When a high-susceptibility agent is present in the vasculature, it causes very nonuniform magnetic fields, which lead to attenuation on T2*- and T2-sensitive MR images. If, as is commonly done, the change in 1/T2 or 1/T2* is assumed to be linear in tissue concentration of the agent, then a series of MR images acquired during bolus passage can be converted into concentration-versus-time images (Fig. 3). Although many clinical uses of DSC MRI use descriptive measures of the bolus passage, such as the time to peak concentration, these dynamic contrast concentration images can be converted into quantitative measures of physiologic variables (48–49). Figure 3Open in figure viewerPowerPoint Images of signal attenuation during the passage of a bolus obtained from a single axial slice every 2 s. These images are the difference between an image acquired prebolus and the image acquired during the bolus passage. Such further quantification of the bolus passage requires removal of the spread of the bolus that results from the intravenous injection and flow through the lungs. Typically, the arterial contrast concentration is inferred from the same images used for perfusion measurement by measuring the signal change in an arterial region of interest. This region can be defined by anatomic criteria, by the early arrival of contrast, by the high blood volume in voxels containing a major vessel, or more typically by a combination of all three (48). Because the volume of artery in the region is uncertain and the proportionality between concentration and relaxation is likely different in the arterial region than in other tissue, the absolute concentration cannot be inferred from the signal; only the relative concentration as a function of time is determined. The effect of the temporal spread of the arterial input can then be removed by a mathematical process know as deconvolution. The deconvolved tissue curves can then be used to determine physiologic quantities (49). The peak height of the curve is a measure of perfusion (cerebral blood flow; CBF) relative to other regions in the brain; the area under the curve is a measure of relative blood volume; and the ratio of the two is the absolute mean transit time (MTT). It should be noted, however, that delay and dispersion of the bolus between the site of measurement of the arterial input function and the region of interest can introduce significant errors in the CBF and MTT measures [e.g., CBF is underestimated (50) by 45% for a delay of 1.5 s]; concentration–time curves must therefore be examined for such characteristics before calculated maps can be regarded as reliable. Arterial spin labeling ASL uses spatially selective radiofrequency and gradient pulses to perturb the nuclear spins of hydrogen in the water molecules of inflowing arterial blood. If time is allowed for the perturbed spins to flow into the tissue being imaged, the perturbation in the inflowing spins will lead to a perfusion-dependent perturbation in the image signal intensity. Because the perturbation decays with T1, an MR relaxation time on the order of 1 s, the experiment can be repeated rapidly to obtain high temporal resolution. The greatest disadvantage of ASL is the small