Abnormalities in diffusion tensor imaging of the uncinate fasciculus relate to reduced memory in temporal lobe epilepsy


Address correspondence to Beate Diehl, M.D., Department of Clinical Neurophysiology, The National Hospital for Neurology and Neurosurgery, Queen Square, London WC1N 3BG, United Kingdom. E-mail: diehlbeate@yahoo.com


Aims: To use Diffusion Tensor Imaging (DTI) to explore structural integrity and connectivity of the uncinate fasciculus (UF) in patients with temporal lobe epilepsy (TLE) and its relationship to memory performance.

Methods: DTI and UF reconstruction were performed in 28 patients with TLE (18 left, 10 right) and 10 normal controls. Differences between left and right UF fractional anisotropy (FA) and apparent diffusion coefficient (ADC) values and correlations between DTI measures and memory scores in the TLE groups were computed.

Results: In controls, FA was higher in the left than right UF (p < 0.01). In left TLE, FA values were lower and ADC values higher than controls in the left UF and ADC values were higher in the right UF (all p < 0.05). In right TLE, ADCs were higher in the left and right UF compared to controls, and FA was reduced in the left UF (all p < 0.05). In left TLE, ADCs in the left UF were negatively correlated with Auditory Immediate (p < 0.05) and Delayed Memory (p < 0.01). Visual Delayed Memory was positively correlated with reduced FA in the ROI of the right UF and increased radial diffusivities (p < 0.05). No significant correlations were found in right TLE. Thus, DTI values correlated with memory scores in the expected direction in patients with left TLE.

Conclusions: Abnormal diffusion measures in the UF ipsilateral to the epileptogenic zone suggest that integrity of the UF is related to memory performance in patients with left TLE. Larger sample sizes are needed to evaluate structure-function correlations further.

Temporal lobe epilepsy (TLE) is the most common intractable focal epilepsy and is often associated with significant memory deficits. The latter often correlate with the type and extent of structural abnormality observed on MRI, such as hippocampal atrophy (Baxendale et al., 1998b; Baxendale et al., 1998a; Sawrie et al., 2001). In vivo imaging studies have shown structural abnormalities in, adjacent to, and remote from areas of ictal onset (Moran et al., 2001).

Diffusion Tensor Imaging (DTI) is an MRI technique that measures the molecular motion of water in tissue in six or more directions, allowing characterization of the magnitude of diffusion as well as its direction in every single voxel, thus providing insight into the microstructure of white matter (Basser et al., 2000; Basser & Jones, 2002; Beaulieu, 2002). Widespread DTI abnormalities have been reported in patients with focal epilepsy (Arfanakis et al., 2002; Diehl et al., 2005; Gross et al., 2006). The integrity of white matter tracts, as measured by DTI, is related to individual differences in performance across a wide range of cognitive skills. For example, the regional brain connectivity in left temporoparietal white matter correlates with a wide range of reading abilities in children (Beaulieu et al., 2005).

The uncinate fasciulus (UF) is a major white matter tract connecting the anterior temporal and frontal lobes (Schmahmann et al., 2007). It has the form of a curved dumbbell and links the anterior three temporal convolutions and the amygdala with the gyrus rectus, medial retro orbital cortex, and subcallosal area (Ebeling & von Cramon, 1992). The UF has an important role in the formation and retrieval of episodic memories (Nestor et al., 2004; Squire & Zola-Morgan, 1991) and is a pathway of seizure spread to the frontal lobe in TLE (Mayanagi et al., 1996).

The purpose of this study was to test the hypothesis that DTI would reveal structural abnormalities of the UF ipsilateral to the seizure focus in TLE and that the degree of abnormality would correlate with functional abnormality, as shown by reduced memory scores.

Patients and Methods

This study included 28 patients with medically intractable TLE and 10 age- and sex-matched controls (please see Table 1 for demographic and seizure information). Eighteen patients had left TLE (13 mesial temporal, 5 lateral temporal) and 10 patients had right TLE (five mesial temporal, five lateral temporal). All patients underwent temporal lobe resection for treatment of epilepsy and all but one patient had good seizure outcome [ILAE class 1, n = 25; ILAE class 2, n = 2; ILAE class 3, n = 1 (Wieser et al., 2001)] at 6 months to 2 years (median 1 year) follow-up. Nine patients had pathologically proven hippocampal sclerosis, 14 nonspecific gliosis, and two Type 1A cortical dysplasia, characterized by architectural disorganization (Palmini et al., 2004). The study was approved by the Institutional Review Board of the Cleveland Clinic Foundation, and all patients gave informed consent prior to enrollment in the study.

Table 1.  Demographic and seizure data for study patients
VariableLeft temporalRight temporal
 Median (range)Median (range)
Age (years)37.00 (24–47) 42.50 (29--55)
Education (years)14.00 (12–19)12.00 (8–17)
Age of seizure onset (years) 23.00 (0.5–41)16.50 (5–42)
Duration of epilepsy (years)14.50 (2–28) 22.50 (1–41)
Sex  Male = 5 (28%)  Male = 6 (60%)
  Female = 13 (72%)Female = 4 (40%)
RaceCaucasian = 18 (100%)Caucasian = 10 (100%)

MRI protocol

MR-images were acquired on a 1.5T whole body MR scanner (Vision Siemens, Medizintechnik, Erlangen, Germany). For the TLE patients, the protocol included a volumetric T1-weighted gradient echo, coronal FLAIR and axial T2-weighted fast spin echo scans and DTI sequences. The control subjects had volumetric T1 and DTI acquisitions. The DTI acquisition comprised axial 2D echo planar imaging (2D EPI) diffusion-weighted sequence with TR/TE = 6000/112 ms, FOV = 24 cm, matrix = 128 × 128, 3 mm contiguous slices without gap, 6 averages, two b values = 0 and 1000 s/mm2; 12 directions. In order to ascertain consistent quality, routine preventative maintenance is performed on the MRI scanner. The measures always remained well within specifications for main field stability, gradient stability, rf stability, and eddy current compensation performance.

DTI quantitation

Data were transferred to a “Leonardo” workstation (Siemens Medizintechnik AG, Erlangen, Germany) from each set of diffusion/orientation-weighted images and processed using DTI task card software (Massachusetts General Hospital, https://www.nmr.mgh.harvard.edu). Specifically, multiple linear regression was used to generate the diffusion tensor D, [Sb= S0× exp(−bD); ln(Sb/S0) =−b(D)] from each set of diffusion/orientation-weighted images. Sb, is the MR signal measured for a given b value, S0 is the MR signal for b = 0, b is the b matrix characterizing the diffusion gradient pluses (timing, amplitude; shape) along each direction(s/mm2); D is the diffusion tensor which describes the molecular mobility along each direction and correlation between these directions. The diffusion tensor D is then diagonalized to obtain the eigenvectors and eigenvalues (λi, i = 1, 2, 3). The eigenvectors represent the major diffusion directions and the eigenvalues are the associated diffusivities. The apparent diffusion coefficient (ADC; units = mm2/s) is calculated from the trace of the diagonalized diffusion tensor [(λ123)/3]. For isotopic diffusion, λ123; for anisotropic diffusion, λ1 > λ2≥λ3. Parametric maps of the ADC and fractional anisotropy (FA) were generated. FA, a scalar (unit-less) quantity, indicates the degree of directionality of the diffusion within a given voxel; it ranges from 0 to 1, with an FA of 0 indicating full isotropy and FA of 1 indicating complete anisotropic diffusion. Similarly, parametric maps for the axial or parallel (main direction of diffusion, E1 =λ1) and radial [perpendicular to the main axis, λT= (λ23)/2] diffusivities were also created. Together, these quantitative measures help to characterize the integrity of the underlying white matter.

Region of interest analysis and tractography

The UF was reconstructed using a two ROI approach to restrict fiber assignment to the UF (Figs. 1 and 2). On the axial colorized FA map, a ROI was placed encompassing the perpendicular fibers passing through the temporal stem in the anterior temporal lobe toward the orbitofrontal cortex (Mori et al., 2005). A second ROI was placed in each patient on an inferior axial slice closer to the inferior and anterior portion of the temporal lobe, encompassing the fibers of the UF, in order to restrict fiber assignment to the UF. Fibers were reconstructed that passed through both ROIs. Fiber tracking was performed using the FACT algorithm (Mori et al., 1999; Stieltjes et al., 2001) implemented within the DTI task card software. The algorithm developed fiber tracts by following the direction of the principle eigenvector at each step starting from a ROI. Tracking propagates on the basis of the orientation of the eigenvector that is associated with the largest eigenvalue. In this study, tracking is terminated when it reaches a voxel with FA lower than a threshold of 0.2 and when the angle between the two principal eigenvectors is greater than 50°. Both of these thresholds are user defined. Measures of FA and ADC were obtained for the entire reconstructed tract.

Figure 1.

(A) The figure illustrates placement of the two ROIs to reconstruct the UF. The white arrow shows the location of the fibers of the right UF in the axial slice of the colorized fiber orientation map. The ROI in orange on the left shows the superior of the two ROIs used to reconstruct the UF. (B) The reconstructed left UF is displayed from a left lateral angle. The two ROIs used for reconstruction are visible in orange and pink color.

Figure 2.

(A) Sagittal and axial cuts of colorized fiber orientation map of a 34-year-old woman with intractable left temporal neocortical epilepsy. The UF is displayed in yellow (left UF) and red (right UF). (B) The UFs were coregistered and overlaid onto the patient's T1 volumetric study.

In order to gain insight into the underlying microstructural sources of the observed differences in the FA and ADC values measured for the tracts, the diffusion along each of the main three directions, i.e., eigenvalues (λ1, λ2, λ3) (mean ± SD), was examined along with the FA and ADC, for a ROI contained within the rostrocaudal course of the UF within the temporal stem. This ROI was selected after reconstruction of the UF to only include fibers that were felt to follow the course of the UF.

The axial and radial diffusivities were computed for each individual ROI within the UF in order to independently evaluate the degree of diffusion parallel and perpendicular to the UF tract at that location. Such measurements in the axial and perpendicular direction allow to elucidate the mechanism producing the observed changes in anisotropy, thus providing insights into the underlying pathology. Due to the nonlinear trajectory of the white matter tracts, such estimates of axial and perpendicular diffusivities are only meaningful in one single plane. Consequently, due to the complex geometric shape of the entire fasciculus, we chose to report on these changes in one ROI, contained within the tract. This region was located within the temporal stem, where the UF has a rostrocaudal orientation and can be easily identified.

Neuropsychological protocol

All TLE patients underwent a comprehensive neuropsychological evaluation as part of their presurgical investigations. The Wechsler Memory Scale-Third Edition (WMS-III) was administered as part of the neuropsychological battery. Four memory indices from the WMS-III were used in the current study to evaluate memory performance. The Auditory Immediate Memory Index and the Auditory Delayed Memory Index were used to assess verbal memory. The Visual Immediate Memory Index and the Visual Delayed Memory Index were used to assess visual memory.


In order to compare age at seizure onset and duration of epilepsy in the TLE groups (left, right), U tests were computed.

To evaluate DTI measures, two-tailed t-tests were conducted to examine differences in FA and ADC between left and right UF among the study groups and differences in FA and ADC values between the groups. Then, Spearman correlations between DTI and memory measures in the TLE groups were examined.

Given the exploratory nature of this study, no correction for Type I error was made.

To obtain measures of reliability, the UF was reconstructed in 10 controls on both sides on two separate occasions, 4 months apart, by the same rater (BD) and reliability was assessed using Cronbach's alpha values.

In all tests, statistical significance was set to p < 0.05. All analyses were performed using the SPSS software package (SPSS, Chicago, IL, U.S.A.).


Demographics were comparable between the study groups (Table 1). Specifically, no difference was found in age at onset of epilepsy in the left versus right TLE group.

All controls and the majority of epilepsy patients were right-handed. A total of 10 patients were left-handed or ambidextrous. These patients were confirmed to be left hemisphere dominant for speech on Wada testing or functional magnetic resonance imaging (fMRI); therefore all subjects are likely left hemisphere dominant for language.

DTI values in controls and patients with left and right TLE


Mean FA in the left UF tract was higher than in the right UF (left UF FA 0.3654 ± 0.033; right UF FA 0.33 ± 0.02; p < 0.01). No other differences in DTI values were identified. Tract volume was symmetric bilaterally. The reliability measure (Cronbach's alpha) for reconstruction of the bilateral UF in the 10 controls on two occasions 4 months apart was excellent (ADC left UF, 0.9920; FA left UF, 0.9950; ADC right UF, 0.9983; FA right UF, 0.9950).

Comparison between TLE patients and controls

In the left TLE group (n = 18), FA was reduced in the left UF, but not in the right as compared to controls. ADCs and radial diffusivities were increased bilaterally (Table 2). In the right TLE group (n = 10), the FA was lower in the left UF than the controls. FA in the right UF, although lower than in left TLE and in controls, was not statistically significant. ADCs were increased in both the left and right UF. Subanalysis of the eigenvalues of the diffusivities within the left ROI revealed significant increases in the radial diffusivities. In the right ROI, although nominally higher, this difference failed to reach statistical significance. Volume of the left and right UF was symmetric in both left and right TLE.

Table 2.  DTI values in controls and patients with left and right TLE
 Controls (n = 10) Mean (SD)Left TLE (n = 18) Mean (SD)Right TLE (n = 10) Mean (SD)
  1. SD, Standard deviation; Lt, left; Rt, right; ROI, Region of Interest; E1, eigenvalue 1. ADC, E1 and radial diffusivities all in 10−4 mm2/s.

  2. *p < 0.05; **p < 0.01 (unpaired t-test between controls and patients with left and right TLE).

lt UF tract, FA0.36540  0.33572*   0.32630**
lt UF tract, ADC8.2047  8.7142**  8.8154**
(0.29958)(0.4658) (0.5972)
ROI within lt UF, FA0.506500    0.416278**    0.387900**
ROI within lt UF, ADC7.5952    8.060556**  7.899300
ROI within lt UF, E112.401012.0833311.60300
ROI within lt UF, radial diffusivities5.19205    6.051750**    6.047250**
(0.6334) (0.61239)(0.6258)
rt UF tract, FA0.330100  0.324778  0.317400
(0.0194)(0.0294) (0.03319)
rt UF tract, ADC8.33870   8.664056*   8.883900*
ROI within rt UF, FA0.469700  0.423167 0.40060
ROI within rt UF, ADC7.51780   7.878583*  8.201700
ROI within rt UF, E111.783011.8544411.99900
(0.630)(0.798) (0.735) 
ROI within rt UF, radial diffusivities5.38495   5.895944*  6.302350

Comparison between TLE patients

There were no differences in DTI measurements between patients with mesial versus lateral TLE. Patients with hippocampal sclerosis and patients without any specific pathology within the resected tissue had comparable DTI measurements to those without either pathology.

Correlations between duration of epilepsy and DTI measures

We examined correlations between age at seizure onset, duration of epilepsy, and DTI measures. These correlations were calculated on all TLE patients as a group as well as separately for those with left TLE and those with right TLE. There were no significant correlations between age at seizure onset or disease duration and DTI measures.

Correlations between DTI measures and memory scores

Left TLE patients

The following correlations with DTI measures in the left UF were found: The Auditory Immediate Memory Index score was negatively correlated with ADC (Table 3). Performance on the Auditory Delayed Memory Index score was negatively correlated with ADC and radial diffusivities, and positively correlated with FA in the ROI in the left UF. In summary, evidence of damage to the left UF based on DTI measurements is associated with reduced performance on measures of Auditory Immediate and Delayed Memory in left TLE patients.

Table 3.  Correlations between DTI measurements and auditory and visual memory scores in left TLE and right TLE
 Auditory immediate index scoreAuditory delayed index scoreVisual immediate index scoreVisual delayed index score
  1. aSpearman's correlation is significant at the 0.01 level (2-tailed).

  2. bSpearman's correlation is significant at the 0.05 level (2-tailed).

Left TLE
lt UF tract, FA0.2720.3820.1090.210
lt UF tract, ADC−0.512b−0.535a−0.0830.126
ROI within lt UF, ADC−0.304−0.435−0.1120.032
ROI within lt UF, FA0.4120.534b0.2580.314
ROI within lt UF, E10.1910.1970.1890.377
ROI within lt UF, radial diffusivity−0.437−0.571b−0.279−0.262
rt UF tract, FA0.0570.1450.0320.090
rt UF tract, ADC−0.107−0.286−0.0850.139
ROI within rt UF, FA0.0060.0790.3970.547b
ROI within rt UF, ADC−0.318−0.369−0.366−0.322
ROI within rt UF, E1−0.160−0.1980.0550.108
ROI within rt UF, radial diffusivity−0.287−0.333−0.376−0.539b
Right TLE
lt UF tract, FA0.4830.0910.4310.254
ADC in lt UF tract−0.250−0.0060.0000.100
ROI within lt UF, ADC−0.268−0.310−0.311−0.044
ROI within lt UF, FA0.1280.1580.116−0.069
ROI within lt UF, E10.189−0.097−0.140−0.263
ROI within lt UF, Perpendicular−0.183−0.219−0.2870.088
FA in rt UF tract0.3670.3900.5500.386
ADC in rt UF−0.1400.359−0.1950.382
ROI within rt UF, FA0.2380.152−0.024−0.138
ROI within rt UF, ADC0.1400.0790.1460.200
ROI within rt UF, E10.5750.3350.3150.320
ROI within rt UF, radial diffusivity−0.152−0.1520.0180.156

The following correlations with DTI measures in the right UF (ROI only) were found: The Visual Delayed Memory Index score was positively correlated with the FA in the ROI and negatively correlated with radial diffusivities. In summary, evidence of damage to the right UF based on DTI measurements is associated with reduced performance on measures of Visual Delayed Memory. Please see also Supplementary material.

Right TLE

No significant correlations between DTI values and memory scores were found in the right TLE group.


DTI of the UF in controls

In this study we found that FA in controls was greater in the left than the right UF with symmetric tract volume. The literature on UF asymmetries in controls remains controversial. Some groups have demonstrated a left greater than right asymmetry in UF FA (Kubicki et al., 2002); others, however, found that the right UF had higher FA (Rodrigo et al., 2007). These differences may in part be due to methodological differences both in image acquisition and analysis. It is likely that there is a variability of diffusion values that can be measured at different locations within the UF; some authors describe a right greater than left asymmetry in the stem and the inferior (temporal) aspect of the UF (Park et al., 2004; Rodrigo et al., 2007), and a left greater than right asymmetry in the frontal aspect of the UF. Detailed analysis in one study showed that right greater than left asymmetry was present in the middle and inferior portion, and left-greater-than-right in the superior portion of the UF (Park et al., 2004). The methodology used in the current study did not allow for separation of those two parts of the UF.

DTI of the UF in patients with Epilepsy

This study showed that patients with TLE have abnormal measures of diffusivity and anisotropy in the UF bilaterally. Reduced FA was noted in the left UF as well as the ROI in the left UF in patients with left and right TLE.

The only study reporting on DTI of the UF in patients with epilepsy reports on 10 patients with right TLE due to right hippocampal sclerosis compared to 10 controls (Rodrigo et al., 2007). It showed that the right, but not left, FA was lower in the epilepsy patients as compared to the controls. There is no report on diffusivity measures.

Our study shows more bilateral involvement in the UF with significantly increased ADCs in the right UF and decreased FA and increased ADCs in the left UF in both right and left TLE patients. This is in concordance with reports of bilateral diffusion abnormalities in limbic structures in patients with TLE. Also, several groups have demonstrated that diffusion abnormalities in TLE exist in areas remote and even contralateral to the presumed seizure focus (Arfanakis et al., 2002; Concha et al., 2005; Gross et al., 2006).

Preferential pathways for seizure spread in TLE may be the fornix and stria terminalis, amygdalofugal fibers, and UF (Mayanagi et al., 1996). Therefore, it is conceivable that the abnormal DTI values may be related to damage of the axonal pathways that are involved in ictal spread. Alternatively, neuronal damage from seizures may lead to secondary white matter loss in connected areas. Interestingly, the current study failed to demonstrate any difference between mesial versus lateral TLE or between the patients with and without hippocampal sclerosis. It is possible, however, that changes may not be apparent because of the small sample size. Studies with larger sample sizes are required to definitively answer this question.

Both epilepsy groups were comparable in age at seizure onset and duration of the epilepsy; hence, assuming similar seizure burden, we would expect comparable damage in both left and right TLE. Indeed, both TLE groups show evidence of comparable pattern of DTI abnormalities in both UFs. However, as we did not prospectively investigate the number of seizures or seizure types, we cannot determine whether there is a correlation between degree of DTI abnormalities and severity of epilepsy. Other variables such as history of status epilepticus and febrile seizures should also be evaluated in a larger group of patients with epilepsy to understand their impact on DTI abnormalities in patients with epilepsy.

To date, the exact mechanism of such seizure-induced damage is unknown. In this study, the characteristics of the diffusion changes in a ROI within the UF were examined to gain further insight into the type of changes. Analyzing the pattern of diffusion changes with respect to diffusivities parallel and radial to the main axonal direction provides in vivo insights into the underlying cause of decreased FA. This study found unchanged parallel diffusivity and increased perpendicular diffusivity. In order to understand the contributions of axonal versus myelin damage, serial diffusion measurements have been performed on the optic nerve in a mouse model of retinal ischemia (Song et al., 2003). According to this model, parallel diffusivity shows a significant decrease in the first days of degeneration, which corresponds to the disintegration of the axonal microstructure, whereas myelin remains intact. Five days after the initial injury perpendicular diffusion increases, which corresponds to the degradation of myelin sheaths. As demonstrated using an in vitro model of Wallerian degeneration in frog sciatic nerve, axonal and myelin degeneration causes a decrease in diffusion anisotropy due to reduced parallel and increased radial diffusion (Beaulieu et al., 1996). In humans, reductions in the principal direction and increases in radial diffusivities have been shown in chronically degenerated white matter tracts (Pierpaoli et al., 2001). Serial DTI measurements in three patients who underwent corpus callosotomy to treat medically refractory seizures and drop attacks revealed interesting insights into the diffusion changes in the corpus callosum after the surgery (Concha et al., 2006). After 1 week, a decrease in parallel diffusivities was seen, evidencing the breakdown of the axons (Kerschensteiner et al., 2005; Concha et al., 2006), creating barriers in the longitudinal displacement of the water molecules. In the chronic stage, 2–4 months after corpus callosotomy, an increase of the radial diffusivities was observed. Most likely at that stage, axonal membranes became more degraded and myelin sheaths showed degeneration, leading to preferential increase in radial diffusivities. It would appear that the overall pattern of FA changes seen in this study is most consistent with chronic Wallerian degeneration, possibly due to cell loss in the temporal lobe secondary to seizure-induced cell death.

Correlations with neuropsychological dysfunction

The role of the UF in memory

The UF is the major fiber tract connecting the inferior frontal, anterior and mesial temporal lobes (Ebeling & von Cramon, 1992). A multitude of functional neuroimaging data has implicated the temporal lobes, particularly mesial temporal and frontal structures in encoding and retrieval of memories. The anterior temporal area receives information from sensory association areas as well as the limbic nuclei and integrates sensory input (Damasio et al., 1985; Markowitsch et al., 1985). In healthy subjects, fMRI has confirmed that episodic memory is associated with both mesial temporal and frontal lobe activation (Markowitsch et al., 1985; Brewer et al., 1998; Wagner et al., 1998; Kirchhoff et al., 2000). There is material-specific lateralization of memory in both healthy volunteers and patients with unilateral mesial temporal lobe lesions. Encoding of verbal information activates the left medial temporal structures, encoding of less verbalizable stimuli, such as patterns, activates the right mesial temporal structures, with encoding of intermediate verbalizable stimuli, such as faces and scenes, result in approximately symmetric activation (Brewer et al., 1998; Golby et al., 2001; Golby et al., 2002; Hwang & Golby, 2006). In general, the lateralization of memory performance regarding verbal material appears stronger; conversely, there is less a firm association of right TLE with disturbed figural learning (Helmstaedter et al., 1995; Powell et al., 2007b).

The medial temporal lobes have been consistently implicated not only in encoding, but also in retrieval (Wagner et al., 1998; Schacter & Wagner, 1999). Prefrontal regions in the left hemisphere are differentially activated during episodic encoding and semantic retrieval, whereas right prefrontal areas are differentially involved during episodic memory retrieval. It therefore seems reasonable to assume that the integrity of the UF linking the frontal, anterior and mesial temporal lobes is important for optimal performance on memory tasks.

Correlations of DTI abnormalities in the UF in disease

In line with the above hypothesis, we found correlations between DTI measures suggesting damage to the UF and dysfunction in lateralized memory tasks. Specifically, the current study suggests that in patients with left TLE, left UF diffusivity is related to reduced verbal memory performance, whereas right UF DTI measures are related to reduced visual memory performance. In patients with right TLE, such correlations could not be demonstrated. This may in large part be due to the small sample size in the right TLE group. Although none of the correlations in the right TLE group reached statistical significance, it should be noted that the variance of correlation coefficients observed in this group was large and some of the correlations were in the medium to large range (see Table 3). This suggests that if the sample sizes had been bigger in the right TLE group, these correlations would likely have reached statistical significance.

In the dominant hemisphere, strong structure and function relationships have been found in the left hemisphere for both language and memory in a variety of DTI and fMRI studies in TLE patients (Powell et al., 2007a, 2007b, 2008; Focke et al., 2008). Subjects with more lateralized functional activation had also more highly lateralized DTI values. In left TLE, more symmetrical language activations were seen on fMRI, along with reduced left hemisphere and increased right hemisphere structural connections. fMRI in the patients undergoing nondominant anterior temporal lobe resection showed no significant correlation between right hippocampal encoding activation for faces or pictures and postoperative change in design learning, suggesting a less strong structure-function relationship in nondominant TLE. Therefore, lack of correlation between visual memory performance and DTI values in the UF in the smaller right TLE group may not be surprising.

Correlations between lateralized memory performance evaluating both verbal and visual memory paradigms and DTI abnormalities have been shown in other diseases: Patients with schizophrenia have reduced levels of functioning across all neuropsychological measures and selective relationships between memory performance and DTI measures have been demonstrated. Reduced left UF FA correlated with reduced scores in measures of declarative-episodic memory and reduced right UF FA correlated with lower scores on measures of working memory, general intelligence, verbal intelligence, and verbal comprehension. The authors felt that the latter finding underscored the widely distributed nature of higher cognition in the brain, thus cautioning against simple isomorphic relationships between function and anatomy (Nestor et al., 2004). Another study reported that lower FA in the right UF correlated with reduced performance on measures of visual attention (Kubicki et al., 2002). In five subjects with schizotypal personality disorder, bilateral reductions of FA in the UF were reported. Correlations were found between right UF abnormalities and clinical symptoms such as restricted affect and social anxiety. Left UF measurements indicative of microstructural damage were correlated with lower performance on measures of verbal and visual memory (Nakamura et al., 2005).

In a group of TLE patients suffering from psychosis (Flugel et al., 2006), a positive correlation was found between verbal fluency and DTI measurements in the left frontal, right frontal, and left temporal regions. Prediction of poor fluency could be made using FA of left frontal and bilateral temporal regions. It was felt that the significant association between impairment on particular executive tests and reductions of frontotemporal FA may reflect the contribution of frontotemporal white-matter abnormalities to the cognitive deficits in these patients. This argument is further strengthened by data from diseases mostly affecting white matter, such as multiple sclerosis, where lesion burden and abnormal diffusivity measures correlate with cognitive performance (Rovaris et al., 2002).

Microstructural abnormalities within the UF therefore could contribute to memory dysfunction in patients with TLE. Furthermore, the UF carries cholinergic fibers from the basal nucleus of Meynert, as part of a cholinergic pathway that supplies frontal, parietal, and temporal neocortices and the perisylvian division of the frontotemporal operculum, insula, and superior temporal gyrus. Altered cholinergic innervation through the UF may contribute to disturbed memory functions (Selden et al., 1998).

One of the shortcomings of the current study is that neuropsychological measures were available only for the patient groups and not for the controls. Therefore, it cannot be determined whether similar correlations exist between memory performance and UF diffusion measures in a healthy control population. The current study is also limited by the rather small sample size, and larger prospective studies will need to be undertaken to confirm the results. Furthermore, due to the exploratory nature of the study, no correction for Type I error was made. However, the strong correlations in the expected direction despite small sample size are a good indicator that correlations between memory performance and integrity of the UF is a robust finding, particularly in the dominant hemisphere.


BD was supported by the Early Career Clinician Scientist Award from the Milken Family Foundation and the Cleveland Clinic Investigator Development Award.

Conflict of interest: We confirm that we have read the Journal's position on issues involved in ethical publication and affirm that this report is consistent with those guidelines. None of the authors has any conflicts of interest to disclose.