Diffusion tensor MRI in temporal lobe epilepsy

Magnetic Resonance Imaging 20 (2002) 511–519

Diffusion tensor MRI in temporal lobe epilepsy Konstantinos Arfanakisa,*, Bruce P. Hermannb, Baxter P. Rogersa, John D. Carewc, Michael Seidenbergd, Mary E. Meyeranda a

Department of Medical Physics, University of Wisconsin, USA b Department of Neurology, University of Wisconsin, USA c Department of Statistics, University of Wisconsin, USA d Department of Psychology, Chicago Medical School, North Chicago, IL, USA Received 2 April 2002; accepted 21 May 2002

Abstract The purpose of this study was to investigate the diffusion characteristics of white matter in patients with focal temporal lobe epilepsy (TLE). Diffusion tensor imaging (DTI) was applied to patients and normal controls. Rotationally invariant mean diffusivity and diffusion anisotropy maps were calculated for all subjects. Comparisons between the two groups were performed for several white matter structures. Mean diffusivity and diffusion anisotropy of each selected structure were tested for correlations with age at onset and duration of epilepsy. Significantly lower diffusion anisotropy, and higher diffusivity in directions perpendicular to the axons, was detected in several white matter structures of the patients when compared to the controls. These structures were not located in the temporal lobes. No significant difference in mean diffusivity was detected between the selected structures from the two groups. Diffusion anisotropy was significantly correlated with age at onset of epilepsy in the posterior corpus callosum. Duration of epilepsy was not significantly correlated with the diffusion indices from any of the selected structures. The results of this study suggest that diffusion anisotropy may reveal abnormalities in patients with focal TLE. In addition, these abnormal changes are not necessarily restricted to the temporal lobes but might extend in other brain regions as well. Furthermore, the age at onset of epilepsy may be an important factor in determining the extent of the effect of epilepsy on white matter. © 2002 Elsevier Science Inc. All rights reserved. Keywords: DTI; TLE; Age at onset; Duration; Diffusion anisotropy

1. Introduction Diffusion tensor imaging (DTI), as implemented in MRI [1,2], is a noninvasive imaging technique that can be used to probe, in vivo, the intrinsic diffusion properties of deep tissues. Unlike conventional diffusion imaging (DI) [3], where diffusion-weighted (DW) images are used to calculate an apparent scalar diffusion constant (ADC), DTI characterizes diffusive transport of water by an effective diffusion tensor D. This symmetric 3 ⫻ 3 tensor is of great importance since it contains useful structural information about the tissue. The eigenvalues of D are the three principal diffusivities and the eigenvectors define the local fiber tract direction field [1]. Moreover, one can derive from D rotationally invariant scalar quantities that describe the intrinsic * Corresponding author. Tel.: ⫹1-608-265-5742; fax: ⫹1-608-2659840. E-mail address: [email protected] (K. Arfanakis).

diffusion properties of the tissue. The most commonly used are the trace of the tensor [1,4,5], which measures mean diffusivity, and Fractional Anisotropy (FA) [4 – 6], which characterizes the anisotropy of the fiber structure. Diffusion properties of brain tissue have demonstrated high sensitivity to pathological changes. DTI has been applied in the diagnosis of disease conditions such as cerebral ischemia [7], acute stroke [8], multiple sclerosis [9], schizophrenia [10] and traumatic brain injury [11]. In epilepsy, reduced diffusivity has been shown with DI, in areas that electrocorticography indicated as related to the seizures [12]. Increased diffusivity and reduced diffusion anisotropy was demonstrated with DI in hippocampi of patients with hippocampal sclerosis [13]. Increased diffusivity and reduced anisotropy was also shown with DTI in some patients with malformations of cortical development, which is commonly associated with epilepsy [14]. The purpose of our study was to investigate the diffusion characteristics of white matter in patients with focal tempo-

0730-725X/02/$ – see front matter © 2002 Elsevier Science Inc. All rights reserved. PII: S 0 7 3 0 - 7 2 5 X ( 0 2 ) 0 0 5 1 2 - X

512

K. Arfanakis et al. / Magnetic Resonance Imaging 20 (2002) 511–519

ral lobe epilepsy (TLE). DTI was applied to a group of patients with focal TLE, and a group of normal controls. Comparisons between the two groups were performed in several white matter structures by means of the diffusivity of protons in directions parallel and perpendicular to the axons, the mean diffusivity, and the diffusion anisotropy of each structure. Additionally, diffusivity in directions parallel and perpendicular to the axons, mean diffusivity, and diffusion anisotropy of each selected white matter structure, were tested for correlations with age at onset and duration of epilepsy.

2. Materials and methods

Table 1 The x, y, z components of unit vectors with origin (0, 0, 0) that define the 23 DW gradient orientations used in the DTI acquisitions {(0, ⫺0.607, 0.795), (0.577, ⫺0.188, 0.795), (0.357, 0.491, 0.795), (⫺0.357, 0.491, 0.795), (⫺0.577, ⫺0.188, 0.795), (0, ⫺0.982, 0.188), (0.742, ⫺0.653, 0.149), (0.350, 0.780, 0.518), (⫺0.934, ⫺0.303, 0.188), (0.985, 0.092, 0.149), (0, 0, 1), (0.392, ⫺0.539, 0.745), (0.634, 0.206, 0.745), (0, 0.667, 0.745), (⫺0.634, 0.206, 0.745), (⫺0.392, ⫺0.539, 0.745), (0.175, ⫺0.837, 0.518), (0.577, 0.795, 0.188), (⫺0.217, 0.965, 0.149), (⫺0.742, ⫺0.425, 0.518), (⫺0.851, 0.504, 0.149), (⫺0.851, 0.276, 0.447), (0.526, ⫺0.724, 0.447)}

of developmental learning disorder. All subjects, or their legal guardian, signed an informed consent form in accordance with institutional policy.

2.1. Subjects 2.2. DTI scans Fifteen patients (mean age 31.6 ⫾ 8.9 years) with temporal lobe epilepsy and fifteen age matched healthy controls (mean age 29.7 ⫾ 11.2 years) participated in this study. Initial selection criteria for epilepsy patients included the following: a) chronological age from 14 to 60 years, b) complex partial seizures of definite or probable temporal lobe origin, c) absence of MRI abnormalities other than atrophy on clinical reading, and d) no other neurological disorder. A board certified neurologist with special expertise in epileptology reviewed patients’ medical records. This review, blinded to all quantitative imaging and cognitive data, included seizure semiology, previous EEGs, clinical neuroimaging reports, and all available medical records. Based on this review, each patient was classified as having complex partial seizures (CPS) of definite, probable, or possible temporal lobe (TL) origin. Definite CPS/TL was defined by continuous video/EEG monitoring of spontaneous seizures demonstrating temporal lobe seizure onset; probable CPS/TL was determined by review of clinical semiology with features reported to reliably identify complex partial seizures of temporal lobe origin versus onset in other regions (e.g., frontal) in conjunction with interictal EEGs, neuroimaging findings, and developmental and clinical history. Only those meeting criteria for definite and probable CPS/TL proceeded to recruitment for study participation. Patients with possible CPS/TL were excluded. Whenever possible, patients were also interviewed in the presence of family regarding details of their epilepsy history and clinical course. Medical records were requested concerning all previous epilepsy-related hospitalizations and records were requested from physicians who had treated the patients’ epilepsy. These records were reviewed and abstracted by an individual blinded to the MRI findings. Selection criteria for healthy controls included the following: a) chronological age from 14 to 60, b) either a friend or family member of the patient, c) no current substance abuse, medical or acute psychiatric condition that could affect cognitive functioning, and d) no psychotropic medications, loss of consciousness ⬎5 minutes, or history

DTI scans were performed on a clinical 1.5T GE Signa LX MRI scanner (General Electric, Waukesha, WI) with Echospeed imaging gradients (22 mT/m maximum amplitude, 120 mT/m/msec slew rate) and the standard GE head coil. The product spin-echo EPI pulse sequence was modified to obtain DTI image datasets with 23 different DW gradient orientations (Table 1). The imaging parameters were TR ⫽ 4500 msec, TE ⫽ 114 msec, field of view ⫽ 240 ⫻ 240 mm2, 20 contiguous slices, 3 mm slice thickness. The amplitude of the total DW gradient was 22 mT/m. The duration of each DW gradient lobe was 32.2 msecs and their temporal spacing was 38.8 msecs. The effective diffusion weighting was b ⫽ 1000 sec/mm2. Echo planar readout was performed with fractional ky encoding (80/128) and later all images were reconstructed to a 128 ⫻ 128 matrix using homodyne reconstruction. Reference images of b ⫽ 0 sec/ mm2 (T2 images) were also acquired. All images (T2 and DW) were repeated 5 times. DTI images were acquired only in the axial plane. The duration of the DTI scan was 9 minutes. All DTI images were transferred to an offline workstation for post processing. The 5 repetitions of the images were first averaged to increase the signal to noise ration (SNR). Then, the T2 images with no diffusion weighting were thresholded to remove CSF, and the averaged DW images were registered to them, using a two-dimensional perspective 8-parameter registration algorithm [16], to correct for distortions induced by eddy currents [17]. For the DTI experiments the following equation applied for each voxel: ln

冉 冊 冘冘

3 3 S bijDij, ⫽⫺ S0 i⫽1 j⫽1

(1)

where S was the DW signal intensity, S0 was the spin-echo signal when no diffusion gradient was applied, bij were elements of the b matrix [2] and Dij were elements of the symmetric 3 ⫻ 3 diffusion tensor D. The only unknowns in

K. Arfanakis et al. / Magnetic Resonance Imaging 20 (2002) 511–519

513

Fig. 1. (a) FA map from a normal volunteer. (EC ⫽ external capsule, PCC ⫽ posterior corpus callosum, ACC ⫽ anterior corpus callosum, PIC ⫽ posterior limb of the internal capsule, AIC ⫽ anterior limb of the internal capsule). The border between AIC and PIC is not obvious from this image. (b) Absolute value color map of the same slice. The arrows show the exact level of the border between AIC and PIC. Color representation of directions in three-dimensional space helps to clearly separate the two structures. (c) Color circle that demonstrates the correspondence of colors and directions in three-dimensional space. This circle should be thought of as a three-dimensional dome that is viewed from below.

Eq. [1] were 6 elements of D. Since DW images were acquired in 23 DW gradient orientations the system was overdetermined and was solved using a non-linear approach (Levenberg-Marquardt) [15]. The eigenvalues ␭1, ␭2, ␭3, and eigenvectors ␧1, ␧2, ␧3, were derived for the diffusion tensor D of each voxel: D ␧ i ⫽ ␭i␧ i

,

i ⫽ 1, 2, 3.

(2)

The eigenvectors of D define the local fiber tract direction field and the corresponding eigenvalues are the effective diffusivities in these directions. Eigenvalue ␭1 represents the diffusivity along the primary diffusion direction ␧1. In theory, the primary diffusion direction in a voxel coincides with the mean of the unit vectors that are parallel to the physical axes of the neurons included in that voxel. The second and third eigenvalues (␭2, ␭3) represent the diffusivities along directions that are perpendicular to the primary diffusion direction, and to each other. The trace [1,4,5] of the diffusion tensor D is equal to: trace ⫽ ␭1 ⫹ ␭2 ⫹ ␭3 ⫽ 3具D典

internal capsule and anterior limb of the internal capsule, in all subjects. The selection was performed based on FA maps where the structures were easily identified (Fig. 1a). Since it was not trivial to detect the border between the posterior and the anterior limb of the internal capsule based only on FA maps, where the two structures appear to be continuous, absolute value color maps [18] were also produced for all subjects (Fig. 1b, 1c). In these maps different colors represent specific directions in 3-D space, and the intensity at each voxel is weighted by its FA value. The difference between the orientation of neurons in the posterior and the anterior limb of the internal capsule is such that the corresponding colors are also significantly different, therefore simplifying the selection of the voxels (Fig. 1b, 1c). The same rules of voxel selection were applied to all patients and normal volunteers. Each side (left, right) of the 5 afore-

(3)

where 具D典 is the mean diffusivity. FA [4 – 6] measures diffusion anisotropy of fiber structure and is equal to:

FA ⫽

冑冑 3 2

1 ((␭ ⫺ ␭2)2 ⫹ (␭2 ⫺ ␭3)2 ⫹ (␭3 ⫺ ␭1)2) 3 1

冑␭21 ⫹ ␭22 ⫹ ␭23

(4)

For any material, 0 ⱕ FA ⱕ 1. The minimum value of FA can occur only in a perfectly isotropic medium. The maximum value can occur only when ␭1 ⬎⬎ ␭2 ⫽ ␭3 [6]. Maps of ␭1, ␭2, ␭3, trace and FA were constructed for all subjects. 2.3. Data analysis A series of voxels was selected systematically in the left and right sides of the external capsule, posterior corpus callosum, anterior corpus callosum, posterior limb of the

Fig. 2. This image shows an example of the selection of voxels from the 5 structures under study, in one slice of a normal volunteer. The background image is an FA map. The colored dots correspond to the selected voxels. The dots are made larger than the in-plane dimensions of a single voxel for visualization purposes. Blue dots are used for the left and right posterior corpus callosum, red dots for the left and right external capsule, purple dots for the left and right posterior limb of the internal capsule, orange dots for the left and right anterior limb of the internal capsule, green dots for the left and right anterior corpus callosum.

514

K. Arfanakis et al. / Magnetic Resonance Imaging 20 (2002) 511–519

the parameter value is zero, were considered significant. For all quantities for which Pearson’s correlation coefficient was computed, graphs of these quantities were also produced.

mentioned structures was considered as a separate region of interest (ROI). Fifteen non-contiguous voxels were selected from each ROI. A total of 150 voxels (15 voxels ⫻ 5 structures ⫻ 2 hemispheres) were chosen in each subject. The selection was performed according to the following protocol. Since thin slices (3 mm) were used in the DTI scans, portions of each ROI appeared in more than 3 slices. After identifying all the slices that contained each ROI, voxels from the 3 central ones were selected. In each one of these 3 slices, 5 non-contiguous voxels were selected per ROI. The voxels were evenly spaced from each other within the same slice and ROI (Fig. 2). All voxels were not close to the edges of the ROI. Measurements from homologous ROIs in the two hemispheres were combined, and mean ␭1, ␭2, ␭3, trace, FA and their standard deviations were calculated for each structure in all subjects. The mean ␭1, ␭2, ␭3, trace, FA and their standard deviations were also calculated for each structure in the patient group and the control group. Comparisons were performed in each structure between the two groups. The significance of the differences was assessed using a Student’s t-test. Only differences with p ⬍ 0.05 were considered significant. Bar graphs were constructed, which compared DTI results from both groups. It has been previously reported that, there exists a significant negative correlation between diffusion anisotropy in several brain regions and the age of the subject [19]. Therefore, Pearson’s correlation coefficient (R) was first computed between the DTI indices (␭1, ␭2, ␭3, FA, trace) from all structures under investigation and the age of the control subjects. Then, Pearson’s correlation coefficient (R) was estimated between DTI indices from white matter structures of the patients and the age at onset of recurrent seizures, as well as the duration of epilepsy. To investigate the significance of a correlation between two quantities, a linear regression model was fitted to the data. To control for the effect of all other parameters an additional linear regression model was applied, which contained a grouping variable for the two populations, and the parameters of age, age at onset and duration. The significance of the regression model parameters was assessed with a t-test. Parameters with a test statistic at a p ⬍ 0.05 level, under the null hypothesis that

3. Results In the patients, significantly increased mean ␭2 was measured in the posterior ( p ⫽ 0.028) and anterior ( p ⫽ 0.033) corpus callosum, and significantly increased ␭2 ( p ⫽ 0.013) and ␭3 ( p ⫽ 0.009) were measured in the external capsule compared to the controls. Significantly reduced mean FA values were detected in the external capsule and the posterior corpus callosum of the patients compared to the controls ( p ⬍ 0.05) (Table 2). In all other white matter structures under study, mean FA was also reduced in the patient group compared to the control group, however the differences were not significant ( p ⬎ 0.05) (Fig. 3) (Table 2). No significant differences were detected in the mean trace values from any of the selected white matter structures of the two groups ( p ⬎ 0.05) (Table 2). For the control group, significant negative correlation was detected between FA and the age of the subjects in the posterior corpus callosum, anterior corpus callosum and posterior limb of the internal capsule ( p ⬍ 0.05) (Table 3) (Fig. 4). There was no significant correlation between FA and the age of the subjects in the external capsule or the anterior limb of the internal capsule of the controls (Table 3) (Fig. 4). Significant positive correlation was detected between the trace, ␭2, ␭3 values and the age of the subjects only in the anterior corpus callosum of the controls (Table 3) (Fig. 4). In the other selected white matter structures there was no significant correlation between the trace values, or the eigenvalues, and the age of the control subjects. In the patient group, significant positive correlation was demonstrated between FA and the age at onset of recurrent seizures only in the posterior corpus callosum and anterior limb of the internal capsule (Table 4) (Fig. 5). No significant correlation was detected between FA and the age at onset in the rest of the selected structures (Table 4) (Fig. 5). In the

Table 2 This table presents the mean FA and trace values, and their standard deviations, for several white matter structures of the epilepsy patients and the healthy controls Trace ⫻ 103 (mm2/sec)

FA Structure

Epilepsy patients

Controls

p

Epilepsy patients

Controls

p

EC PCC ACC PIC AIC

0.325 ⫾ 0.025 0.66 ⫾ 0.07 0.60 ⫾ 0.07 0.64 ⫾ 0.05 0.51 ⫾ 0.04

0.38 ⫾ 0.05 0.71 ⫾ 0.05 0.65 ⫾ 0.06 0.64 ⫾ 0.04 0.54 ⫾ 0.06

0.0011 0.042 0.076 0.84 0.08

2.06 ⫾ 0.09 2.26 ⫾ 0.15 2.55 ⫾ 0.23 2.01 ⫾ 0.19 1.95 ⫾ 0.09

1.98 ⫾ 0.16 2.22 ⫾ 0.12 2.40 ⫾ 0.18 2.01 ⫾ 0.19 1.96 ⫾ 0.11

0.094 0.45 0.066 0.93 0.88

(EC ⫽ external capsule, PCC ⫽ posterior corpus callosum, ACC ⫽ anterior corpus callosum, PIC ⫽ posterior limb of the internal capsule, AIC ⫽ anterior limb of the internal capsule). There is significant reduction ( p ⬍ 0.05) of FA in EC and PCC of the epilepsy patients compared to the controls. No other significant difference was detected in the FA values of the two groups. However, FA values in the epilepsy patients are lower than those of the controls for ACC and for AIC also, and the corresponding p-values are low. No significant differences were detected in the trace values of the two groups.

K. Arfanakis et al. / Magnetic Resonance Imaging 20 (2002) 511–519

515

epilepsy, all the non-significant correlations between DTI indices and the age of the subjects remained non-significant. Significant negative correlation was detected between FA and the age of the subjects in the anterior corpus callosum ( p ⫽ 0.020) and the posterior limb of the internal capsule ( p ⫽ 0.039). Significant positive correlation was detected between the trace ( p ⫽ 0.032) and the age of the subjects only in the anterior corpus callosum. Also in the anterior corpus callosum, there was significant positive correlation between ␭2 ( p ⫽ 0.013), ␭3 ( p ⫽ 0.014) and the age of the subjects. In the posterior limb of the internal capsule, ␭2 was positively correlated with age ( p ⫽ 0.028). When the correlations of DTI indices with age at onset were repeated controlling for the effects of duration and age, all the non-significant correlations remained non-significant. For the posterior corpus callosum the positive correlation of FA with age at onset remained significant ( p ⫽ 0.022). Also in the posterior corpus callosum, significant negative correlation was detected between ␭2 and the age at onset ( p ⫽ 0.021). The trace, ␭1, ␭3 values were not significantly correlated with the age at onset of epilepsy, controlling for the effects of duration and age, for any of the selected white matter structures. When the correlations of DTI indices with duration were repeated controlling for the effects of age and age at onset, all the non-significant correlations remained non-significant. Almost significant negative correlation was detected only between FA and the duration of epilepsy in the anterior limb of the internal capsule ( p ⫽ 0.058). The trace, ␭1, ␭2, ␭3 values and the duration of epilepsy were not significantly correlated in any of the selected structures.

Fig. 3. This graph shows the mean FA values and their standard deviations from the selected white matter structures in patients and controls. (EC ⫽ external capsule, PCC ⫽ posterior corpus callosum, ACC ⫽ anterior corpus callosum, PIC ⫽ posterior limb of the internal capsule, AIC ⫽ anterior limb of the internal capsule). In all structures the mean FA is lower in the patients compared to the controls. However, only in EC and PCC the difference is significant.

posterior corpus callosum, significant negative correlation was measured between ␭2 and the age at onset (Table 4). The trace, ␭1, ␭3 values and the age at onset were not significantly correlated for any of the selected white matter structures in the patients (Table 4) (Fig. 5). Significant negative correlation was detected between FA and duration of epilepsy in the posterior corpus callosum and anterior limb of the internal capsule (Table 5) (Fig. 6). In no other selected structure FA was significantly correlated with duration (Table 5) (Fig. 6). In the posterior corpus callosum there was significant positive correlation between ␭2 and the duration of epilepsy. In the anterior limb of the internal capsule there was significant negative correlation between ␭1 and duration, and positive correlation between ␭3 and duration (Table 5). The trace values and the duration of epilepsy were not significantly correlated in any of the selected white matter structures (Table 5) (Fig. 6). Controlling for the effects of age at onset and duration of

4. Discussion In this study, mean DTI characteristics were first compared between the group of TLE patients and the group of normal controls. Although these quantities vary with the age of the subjects [19], comparisons between the two groups were possible since the mean age and the age distributions were similar for the patients and the controls. In contrast, comparisons of the DTI results between single patients and

Table 3 Pearson’s correlation coefficient (R) between the DTI indices and the age of the subjects, for several white matter structures of the controls FA vs. Age

␭1 vs. Age

Trace vs. Age

␭2 vs. Age

␭3 vs. Age

Structure

R

p

R

p

R

p

R

p

R

p

EC PCC ACC PIC AIC

0.157 ⫺0.529 ⫺0.643 ⫺0.595 0.131

0.577 0.0425 0.0096 0.0194 0.643

⫺0.38 0.337 0.685 0.199 ⫺0.229

0.162 0.219 0.0048 0.477 0.411

⫺0.054 ⫺0.184 ⫺0.047 ⫺0.191 ⫺0.269

0.793 0.368 0.819 0.350 0.183

⫺0.100 0.376 0.540 0.254 0.249

0.627 0.058 0.004 0.210 0.219

0.158 0.336 0.555 0.016 0.351

0.442 0.093 0.003 0.538 0.079

(EC ⫽ external capsule, PCC ⫽ posterior corpus callosum, ACC ⫽ anterior corpus callosum, PIC ⫽ posterior limb of the internal capsule, AIC ⫽ anterior limb of the internal capsule). There is significant negative correlation ( p ⬍ 0.05) between FA and the age of the subjects in PCC, ACC and PIC. There is also significant positive correlation ( p ⬍ 0.05) between the trace, ␭2, ␭3 values and the age of the control subjects in ACC.

516

K. Arfanakis et al. / Magnetic Resonance Imaging 20 (2002) 511–519

Fig. 4. Graphs of FA and trace values from 5 white matter structures of 15 controls versus the age of the subjects. (EC ⫽ external capsule, PCC ⫽ posterior corpus callosum, ACC ⫽ anterior corpus callosum, PIC ⫽ posterior limb of the internal capsule, AIC ⫽ anterior limb of the internal capsule). There is significant negative correlation between FA and the age of the subjects in PCC, ACC and PIC. There is also significant positive correlation between the trace values and the age of the control subjects in ACC.

0.05) only for the external capsule and the posterior corpus callosum, Fig. 3 and Table 2 show a reduction of FA in the anterior corpus callosum and the anterior limb of the internal capsule with very low p-values ( p ⫽ 0.076, p ⫽ 0.08). In order to make a definitive statement about the significance of these changes, additional measurements are necessary. It is possible that if more data is acquired, the changes of FA in the anterior corpus callosum and the anterior limb of the internal capsule may also become significant ( p ⬍ 0.05). The patients that participated in this study were diagnosed with TLE. However, changes of FA were detected in regions that are not part of the temporal lobes. Therefore, although the origin of the seizures was the temporal lobes, other brain regions might also be related to TLE. This finding may be important for cases where surgical treatment is considered. No significant differences were detected between the mean trace values from any of the selected structures of the

the mean values from the control group [14] were not performed. These were avoided due to the fact that it was possible to produce significant differences just due to averaging of results from controls with ages very different than the age of the single patient. A study on single patients would be appropriate only if for each patient there existed a group of controls with similar age. Our results showed that in patients with focal TLE there was significant reduction of diffusion anisotropy in several white matter regions compared to normal controls. In addition, it was evident from our data that this reduction of diffusion anisotropy was due to a significant increase of diffusivity in directions perpendicular to the axons (␭2, ␭3). Thus, in these particular white matter regions of the patients, restriction of diffusion in directions perpendicular to the axons was reduced. This might be due to a deficit of myelin, or an increased permeability in the membranes of the axons, or a less tightly packed neuronal network. Although the reduction of FA was significant ( p ⬍

Table 4 Pearson’s correlation coefficient (R) between the DTI indices and the age at onset of recurrent seizures, for several white matter structures of the patients FA vs. Age at onset

Trace vs. Age at onset

␭1 vs. Age at onset

␭2 vs. Age at onset

␭3 vs. Age at onset

Structure

R

p

R

p

R

p

R

p

R

p

EC PCC ACC PIC AIC

0.351 0.595 0.221 ⫺0.078 0.554

0.200 0.0192 0.429 0.784 0.0320

⫺0.507 ⫺0.265 0.198 0.245 0.184

0.0536 0.340 0.480 0.380 0.512

⫺0.371 0.439 0.481 0.186 0.450

0.173 0.102 0.069 0.506 0.093

⫺0.505 ⫺0.627 ⫺0.080 0.412 ⫺0.016

0.055 0.012 0.776 0.127 0.955

⫺0.471 ⫺0.405 0.036 0.099 ⫺0.413

0.076 0.134 0.899 0.725 0.126

(EC ⫽ external capsule, PCC ⫽ posterior corpus callosum, ACC ⫽ anterior corpus callosum, PIC ⫽ posterior limb of the internal capsule, AIC ⫽ anterior limb of the internal capsule). Significant positive correlation between FA and the age at onset is detected only in the PCC and the AIC. Also in the PCC, significant negative correlation exists between ␭2 and the age at onset. There is no significant correlation between the trace, ␭1, ␭3 values and the age at onset in any of the selected structures.

K. Arfanakis et al. / Magnetic Resonance Imaging 20 (2002) 511–519

517

Fig. 5. Graphs of FA and trace values from 5 white matter structures of 15 epilepsy patients versus the age at onset of recurrent seizures. (EC ⫽ external capsule, PCC ⫽ posterior corpus callosum, ACC ⫽ anterior corpus callosum, PIC ⫽ posterior limb of the internal capsule, AIC ⫽ anterior limb of the internal capsule). Significant correlation exists only between FA and the age at onset in the PCC and AIC.

in the human brain to complete myelination [20,21], at an age of approximately 20 years, then the observed negative correlation of diffusion in directions perpendicular to the axons with the age at onset of epilepsy might be a result of incomplete myelination. Future studies with magnetizationtransfer imaging [22] may provide more myelin specific information. In contrast to the age at onset of epilepsy, the duration of epilepsy did not appear to have an effect on the diffusion properties of the white matter structures that were studied. In this study, all analysis was based on measurements in systematically selected voxels. The selection was done manually and followed specific rules that were described in the Methods section. Selection of all the voxels that belonged to each structure was avoided, so that these voxels that lie next to the edges of the structures, and which are affected by partial volume effects, would not be included. For different brain volumes the percentage of such voxels would be different, thus affecting the final mean results.

patients and controls. However, this does not mean that there are no changes of trace values anywhere in white matter, in association with TLE. Our measurements were restricted to a few easily identifiable white matter structures and they did not extend over all white matter. Correlation of DTI indices from specific white matter structures with the age of the controls resulted in findings similar with previously published research [19]. In several structures FA reduced significantly with age. The new information that arose from our study was that trace values in the anterior corpus callosum increase significantly with age. In the posterior corpus callosum, the diffusivity in directions perpendicular to the axons was higher, and the diffusion anisotropy was lower with an earlier age at onset of recurrent seizures. Therefore, in patients with early age at onset of epilepsy, there was less restriction of diffusion in directions perpendicular to the axons of the posterior corpus callosum, compared to patients with late onset. If we consider that the posterior corpus callosum is the last structure

Table 5 Pearson’s correlation coefficient (R) between the DTI indices and the duration of epilepsy, for several white matter structures of the patients FA vs. Duration

Trace vs. Duration

␭1 vs. Duration

␭2 vs. Duration

␭3 vs. Duration

Structure

R

p

R

p

R

p

R

p

R

p

EC PCC ACC PIC AIC

⫺0.163 ⫺0.514 ⫺0.439 0.137 ⫺0.828

0.562 0.0498 0.101 0.625 0.0001

0.365 0.347 ⫺0.0129 ⫺0.479 ⫺0.187

0.181 0.205 0.964 0.071 0.504

0.316 ⫺0.261 ⫺0.460 ⫺0.447 ⫺0.638

0.251 0.348 0.085 0.095 0.010

0.260 0.572 0.291 ⫺0.457 0.298

0.350 0.026 0.292 0.086 0.280

0.369 0.405 0.215 ⫺0.343 0.539

0.176 0.134 0.441 0.211 0.038

(EC ⫽ external capsule, PCC ⫽ posterior corpus callosum, ACC ⫽ anterior corpus callosum, PIC ⫽ posterior limb of the internal capsule, AIC ⫽ anterior limb of the internal capsule). Significant negative correlation between FA and the duration is detected only in the PCC and the AIC. In the PCC there is also significant positive correlation between ␭2 and the duration of epilepsy. In the AIC there is significant negative correlation between ␭1 and duration, and positive correlation between ␭3 and duration. There is no significant correlation between the trace values and the duration in any of the selected structures.

518

K. Arfanakis et al. / Magnetic Resonance Imaging 20 (2002) 511–519

Fig. 6. Graphs of FA and trace values from 5 white matter structures of 15 epilepsy patients versus the duration of epilepsy. (EC ⫽ external capsule, PCC ⫽ posterior corpus callosum, ACC ⫽ anterior corpus callosum, PIC ⫽ posterior limb of the internal capsule, AIC ⫽ anterior limb of the internal capsule). Significant correlation exists only between FA and the duration in the PCC and AIC.

Partial volume effects were avoided with our method, since all selected voxels were away from the edges of the structures. At the same time enough voxels were selected from each structure to reduce the variance. Although our method was manual it was quite systematic, thus producing reproducible results. When the same method was applied multiple times on the same subject the results were very similar. Any variations were minimal and would not alter the final results of the study. Voxel selection was performed following the same rules for all subjects, instead of selecting pixels from one subject and then applying a voxel by voxel analysis using such methods as statistical parametric mapping (SPM, Wellcome Department of Cognitive Neurology, UK). In order to use a voxel by voxel analysis all FA maps would have to be registered into a standard space. Such a template does not exist for FA maps. Therefore, one solution would be to register first a T2 weighted dataset to the Talairach coordinates and then use the same transformation for the FA dataset. When this was attempted, registration of the FA datasets was relatively crude. This was somewhat expected since the criterion in that process was not matching the characteristic patterns of FA datasets, but the patterns of CSF, which dominate T2 weighted images. The other solution would be to select one of the FA datasets as the standard space and register the datasets from all other subjects to that. However, spatial smoothing would have to be performed prior to registration, which would affect the final results. Therefore, voxel by voxel analysis was considered inappropriate for this study. Due to the intrinsic relationship between age, age at onset and duration of epilepsy, when testing for correlations between DTI indices and the age at onset or duration, corrections had to be performed to control for the effects of

age, and duration or age at onset respectively. This could be avoided to some extent if the subjects were selected as follows. In order to study the effect of the age at onset on DTI indices, patients should have different ages at onset and the same duration of epilepsy. In order to study the effect of duration on DTI indices, patients should have the same age at onset and different duration of epilepsy. If patients would be selected in this manner then in each case we would have to control only for the effect of age.

5. Conclusion This study suggests that diffusion anisotropy may reveal abnormalities in patients with TLE. These abnormalities are due to decreased restriction on diffusion in directions perpendicular to the axons. In addition, these changes are not necessarily restricted to the temporal lobes but might extend to other brain regions as well. Finally, the age at onset of recurrent seizures may be an important factor in determining the extent of the effect of epilepsy on white matter.

Acknowledgments The authors gratefully acknowledge Dr. Jason Fine for his help with the statistical analysis. This work was supported in part by NS37738 and M01 RR03186.

References [1] Basser PJ, Mattiello J, Le Bihan D. MR diffusion tensor spectroscopy and imaging. Biophys J 1994;66:259 – 67.

K. Arfanakis et al. / Magnetic Resonance Imaging 20 (2002) 511–519 [2] Basser PJ, Mattiello J, Le Bihan D. Estimation of the effective self-diffusion tensor from the NMR spin echo. J Magn Reson B 1994;103:247–54. [3] Le Bihan D. Diffusion NMR imaging. Magn Reson Q 1991;7:1–30. [4] Pierpaoli C, Jezzard P, Basser PJ, Barnett A, Di Chiro G. Diffusion tensor MR imaging of the human brain. Radiology 1996;201:637– 48. [5] Basser PJ. Inferring microstructural features and the physiological state of tissues from diffusion-weighted images. NMR in Biomedicine 1995;8:333– 44. [6] Basser PJ, Pierpaoli C. Microstructural and physiological features of tissues elucidated by quantitative-diffusion-tensor MRI. J Magn Reson B 1996;111:209 –19. [7] Lythgoe MF, Busza AL, Calamante F, Sotak CH, King MD, Bingham AC, Williams SR, Gadian DG. Effects of diffusion anisotropy on lesion delineation in a rat model of cerebral ischemia. Magn Reson Med 1997;38:662– 8. [8] van Gelderen P, de Vleeschouwer MHM, DesPres D, Pekar J, van Zijl PCM, Moonen CTW. Water diffusion and acute stroke. Magn Reson Med 1994;31:154 – 63. [9] Werring DJ, Clark CA, Barker GJ, Thompson AJ, Miller DH. Diffusion tensor imaging of lesions and normal-appearing white matter in multiple sclerosis. Neurology 1999;52:1626 –32. [10] Lim KO, Hedehus M, Moseley M, de Crespigny A, Sullivan EV, Pfefferbaum A. Compromised white matter tract integrity in schizophrenia inferred from diffusion tensor imaging. Arch Gen Psychiatry 1999;56:367–74. [11] Werring DJ, Clark CA, Barker GJ, Miller DH, Parker GJM, Brammer MJ, Bullmore ET, Giampietro VP, Thompson AJ. The structural and functional mechanisms of motor recovery: complementary use of diffusion tensor and functional magnetic resonance imaging in a traumatic injury of the internal capsule. J Neurol Neurosurg Psychiatry 1998;65:863–9. [12] Diehl B, Najm I, Ruggieri P, Foldvary N, Mohamed A, Tkach J, Morris H, Barnett G, Fisher E, Duda J, Luders HO. Periictal diffu-

[13]

[14]

[15]

[16]

[17]

[18]

[19]

[20]

[21]

[22]

519

sion-weighted imaging in a case of lesional epilepsy. Epilepsia 1999; 40(11):1667–71. Wieshmann UC, Clark CA, Symms MR, Barker GJ, Birnie KD, Shorvon SD. Water diffusion in the human hippocampus in epilepsy. Magn Reson Imaging 1999;17:29 –36. Eriksson SH, Rugg-Gunn FJ, Symms MR, Barker GJ, Duncan JS. Diffusion tensor imaging in patients with epilepsy and malformations of cortical development. Brain 2001;124:617–26. Press WH, Flannery BP, Teukolsky SA, Vetterling WT. Numerical recipes in C: The art of scientific computing. 2nd ed. Cambridge: Cambridge University Press, 1992. Woods RP, Grafton ST, Holmes CJ, Cherry SR, Mazziotta JC. Automated image registration: I. General methods and intrasubject, intramodality validation. J Comp Assist Tomogr 1998;22:141–54. Haselgrove JC, Moore JR. Correction for distortion of echo-planar images used to calculate the apparent diffusion coefficient. Magn Reson Med 1996;36:960 – 4. Pajevic S, Pierpaoli C. Color schemes to represent the orientation of anisotropic tissues from diffusion tensor data: application to white matter fiber tract mapping in the human brain. Magn Reson Med 1999;42:526 – 40. Pfefferbaum A, Sullivan EV, Hedehus M, Lim KO, Adalsteinsson E, Moseley M. Age-related decline in brain white matter anisotropy measured with spatially corrected echo-planar diffusion tensor imaging. Magn Reson Med 2000;44:259 – 68. Giedd JN, Rumsey JM, Castellanos FX, Rajapakse JC, Kaysen D, Vaituzis AC, Vauss YC, Hamburger SD, Rapoport JL. A quantitative MRI study of the corpus callosum in children and adolescents. Dev Brain Res 1996;91:274 – 80. Paus T, Collins DL, Evans AC, Leonard G, Pike B, Zijdenbos A. Maturation of white matter in the human brain: A review of magnetic resonance studies. Brain Res Bul 2001;54(3):255– 66. Balaban RS, Ceckler TL. Magnetization transfer contrast in magnetic resonance imaging. Magn Reson Quart 1992;8(2):116 –37.