Skip to main content
NCBI home page
Human Brain Mapping logoLink to Human Brain Mapping
. 2009 Jun 10;30(12):4070–4081. doi: 10.1002/hbm.20830

Functional and structural changes in the memory network associated with left temporal lobe epilepsy

Natalie L Voets 1,2,✉, Jane E Adcock 3, Richard Stacey 4, Yvonne Hart 3, Katherine Carpenter 5, Paul M Matthews 2,6, Christian F Beckmann 1,6
PMCID: PMC6870932  PMID: 19517529

Abstract

Understanding functional plasticity in memory networks associated with temporal lobe epilepsy (TLE) is central to predicting memory decline following surgery. However, the extent of functional reorganization within memory networks remains unclear. In this preliminary study, we used novel analysis methods assessing network‐level changes across the brain during memory task performance in patients with TLE to test the hypothesis that hippocampal functions may not readily shift between hemispheres, but instead may show altered intra‐hemispheric organization with unilateral damage. In addition, we wished to relate functional differences to structural changes along specific fibre pathways associated with memory function. Nine pre‐operative patients with intractable left TLE and 10 healthy controls underwent functional MRI during complex scene encoding. Diffusion tensor imaging was additionally performed in the same patients. In our study, we found no evidence of inter‐hemispheric shifts in memory‐related activity in TLE using standard general linear model analysis. However, tensor independent component analysis revealed significant reductions in functional connectivity between bilateral MTL, occipital and left orbitofrontal regions among others in left TLE. This altered orbitofrontal activity was directly related to measures of fornix tract coherence in patients (P < 0.05). Our results suggest that specific fibre pathways, potentially affected by MTL neurodegeneration, may play a central role in functional plasticity in TLE and highlight the importance of network‐based analysis approaches. Relative to standard model‐based methods, novel objective functional connectivity analyses may offer improved sensitivity to subtle changes in the distribution of memory functions relevant for surgical planning in TLE. Hum Brain Mapp, 2009. © 2009 Wiley‐Liss, Inc.

Keywords: temporal lobe epilepsy, memory, fMRI, DTI, functional connectivity, independent component analysis

INTRODUCTION

Surgical resection of epileptogenic medial temporal lobe tissue is a highly effective treatment for medically intractable unilateral temporal lobe epilepsy (TLE). However, mounting data suggest that metabolic and structural changes extend beyond the medial temporal lobe (MTL) even with localized seizures, including to temporal stem white matter [Joo et al., 2005] and ipsilateral frontal regions [Savic et al., 1997]. These extra‐temporal metabolic changes in TLE have been related to, among others, deafferentation (e.g., Cendes et al., 1994] and demyelination [Meiners et al., 2000]. Consistent with this notion, changes in fractional anisotropy (FA) and diffusivity have been noted in TLE patients along direct hippocampal efferent pathways such as the fornix [Concha et al., 2005]. However, the network‐level impact of neuronal loss in the MTL on specific fibre pathways and associated functional systems remains unclear.

Preoperatively, patients with left TLE typically demonstrate verbal memory deficits (e.g., Hermann et al., 1992]. However, their relatively limited memory impairments in the presence of potentially extensive unilateral hippocampal pathology have suggested functionally compensatory mechanisms to be associated with chronic seizures. The localization of memory functions therefore remains critical in the pre‐surgical evaluation of individual unilateral TLE patients to minimize the risk of post‐operative amnesic syndromes.

Non‐invasive imaging studies suggest the potential for functional MRI (fMRI) to predict post‐operative memory decline [Powell et al., 2008; Rabin et al., 2004]. Confident localization of memory functions using fMRI, however, remains challenging due to hardware, physiological and statistical issues. The lack of a validated, objective method for determining clinically meaningful hemispheric shifts in activation limits characterization of functional plasticity in memory networks (e.g., Chlebus et al., 2007; Jansen et al., 2006; Seghier, 2008]. For example, many studies (including ours) of language‐ and memory‐related changes in TLE have assessed inter‐hemispheric shifts using laterality indices based on a simple count of “significantly” activated voxels in a region of the left hemisphere compared with the right. Such calculations are inherently dependent on selection of arbitrary statistical thresholds at which to compare left and right hemisphere activations; a problem in itself as the “correct” way to determine significance of fMRI results remains an area of ongoing research. Unsurprisingly, evidence both supporting (e.g., Deblaere et al., 2005; Vannest et al., 2008] and contraindicating (e.g., Cheung et al., 2006; Rabin et al., 2004] absolute changes in the laterality of memory‐related activation in TLE has been reported.

In addition to methodological limitations, pathophysiological mechanisms contributing to refractory seizures such as abnormal axonal and dendritic sprouting (e.g., Mathern et al., 1996, Rao et al., 2006, Siddiqui and Joseph, 2005] could also contribute to negative findings of inter‐hemispheric shifts in memory functions. We therefore hypothesized that unilateral temporal lobe epilepsy may affect the relative communication between distinct but related anatomical regions within a wider memory network rather than inducing reorganization of memory functions to the contra‐lesional hippocampus. Independent component analysis methods measuring correlations in fMRI signals across all voxels of the brain during task performance (but independent of task models) have recently shown increased sensitivity, as compared with model‐based approaches, to disease‐related differences in functional processing [Rombouts et al., 2009]. To test our hypothesis, we studied a small group of patients with unilateral left TLE to directly compare standard model‐based methods with novel, objective functional connectivity analyses sensitive to network‐level changes across the whole brain. To our knowledge, such a comparison has not previously been conducted to assess functional reorganization with disease. In addition, we wished to relate any resulting functional differences to structural changes along specific fibre pathways associated with memory function.

A variety of fMRI memory paradigms have been described in the literature, spanning encoding and retrieval/recognition for a range of stimuli from autobiographical (hometown) and spatial navigation (virtual town) to word, face, scene, and pattern items. We chose to examine complex scene encoding, as this has been shown to engage bilateral MTL regions [Rombouts et al., 2001; Stern et al., 1996] through use of both verbal and visuo‐spatial encoding strategies [Detre et al., 1998], thus presenting an ideal task for both left and right TLE patients as part of a larger fMRI battery. As TLE is associated with hippocampal dysfunction, we set out to test whether memory‐encoding functional connectivity differences between LTLE patients and controls relate to alterations in structural connectivity along the fornix, the primary hippocampal efferent pathway. However, as the contribution of specific fibre pathways to the encoding of stimuli that may be processed through both verbal and spatial routes is unknown (see, for example, an elegant study by Gaffan et al. [ 2001] demonstrating composite effects on visual “scene” learning impairments of fornix transactions added to temporal stem resections), we also investigated uncinate structural coherence.

Using fMRI and DTI in patients with left TLE and healthy volunteers, we tested whether (i) LTLE is associated with overt inter‐hemispheric reorganization of memory functions, (ii) LTLE is associated with altered functional connectivity (for example, intra‐hemisperically) in the memory network, and (iii) changes in the relative distribution of memory activations relate to altered coherence of fibre pathways. Evidence for the latter could have potentially important implications for pre‐surgical planning.

MATERIALS AND METHODS

Subjects

Nine right‐handed pre‐operative patients with intractable LTLE (mean age 34 years, range 18–47) being evaluated by the Oxford Epilepsy Surgery Service were included (Table I). Data for an additional LTLE patient was rejected due to a technical issue with the data. A dedicated epilepsy surgery team determined that each patient had intractable left TLE on the basis of ictal video EEG, MRI, neuropsychological assessment and, where available, amytal test results. Clinical Neuropsychological examination was performed for all but one of the patients and included the Weschler Adult Intelligence Scale (WAIS‐III), verbal fluency measures and the Adult Memory and Information Processing (AMPIB) form. Ten neurologically normal healthy volunteers (mean age 29, range 24–38) also took part. All volunteers gave written informed consent to take part in the study. The research was conducted in accordance with the principles of the Declaration of Helsinki and the study was approved by the Berkshire Local Research Ethics Committee.

Table I.

LTLE patient demographics

Patient Sex Age Age at onset (years) Early complicated prolonged convulsion Duration (years) Clinical MRI diagnosis Verbal/non‐verbal IQ
1 M 35 2 Yes 33 HS 66/84
2 F 18 7 Unknown 11 HS 74/87
3 F 28 11 No 17 HS 95/101
4 F 47 43 Yes 4 HS 86/94
5 F 41 12 Yes 29 HS 84/104
6 M 32 4 No 28 HS 74/93
7 F 38 34 No 4 HS n/a
8 F 33 28 No 5 Dysplasia 105/119
9 M 38 11 No 27 HS 107/99
Open in a new tab

HS, hippocampal sclerosis; n/a, not available.

Imaging Methods

Blood‐oxygen‐level‐dependent (BOLD) fMRI data were acquired on a 3T Siemens‐Varian scanner while subjects performed a complex visual memory‐encoding task similar to that described in [Rombouts et al., 2001]. Prior to scanning, subjects were familiarized with a set of 10 color photographs depicting landscapes, animals, and unfamiliar people or buildings shown in random order a total of 10 times. Each image was presented for 4 seconds with a one second inter‐stimulus interval. To ensure each image was seen and to facilitate encoding, subjects indicated whether a building was present in the image using a button box. During the fMRI scan, subjects viewed 50 s alternating blocks of these (randomized) “familiar” images and sets of novel images. Subjects again indicated the presence of a building in each image. Single‐shot echo‐planar T2*‐weighted images were acquired continuously using the following acquisition parameters: TR = 3 s, TE = 30 ms, voxel dimensions 3 × 4 × 5 mm, 5 mm slice thickness.

Whole‐head diffusion data sets (60 slices, 2.5 mm thickness) consisted of three non‐diffusion‐weighted and 60 diffusion‐weighted images acquired with a b value of 1000 s · mm−2 (TE = 106 ms, effective TR of 20 R‐R intervals, half k‐space acquisition, matrix = 62 × 96, FOV = 240 × 240 mm2). The images were interpolated to achieve a matrix size of 128 × 128 and a final resolution of 1.875 × 1.875 × 2.5 mm3. The protocol used a doubly‐refocused spin‐echo sequence to minimize eddy currents [Reese et al., 2003] and cardiac gating to minimize pulsatile motion artefacts [Nunes et al., 2005].

fMRI Data Analysis: Model‐Based Analyses

FMRI data acquired during task performance were analyzed both using model‐based and model‐free methods. In the model‐based approach, data were analyzed using a whole‐brain General Linear Model (GLM) (FEAT, part of FSL, http://www.fmrib.ox.ac.uk/fsl/) [Smith et al., 2004]. Pre‐processing steps included: motion correction [Jenkinson et al., 2002], spatial smoothing (Gaussian kernel, full‐width half‐maximum 5 mm) and high‐pass temporal filtering. Activation maps were generated using cluster statistics performed on all voxels above a threshold of z = 2.3 with a corrected cluster extent threshold of P < 0.05.

As whole‐brain multiple comparison correction may be overly stringent when examining localized MTL activity, the GLM analysis was repeated using a voxel‐wise approach confined to a region‐of‐interest (ROI) defined on the standard MNI152 brain template incorporating hippocampal and parahippocampal labels from the Harvard‐Oxford standard brain atlas. This probabilistic atlas is based on manual segmentations of cortical and subcortical regions defined in 38 individual subject T1‐weighted structural images, carried out at the Centre for Morphometric Analysis at Harvard University, Boston, MA. Due to the proximity of MTL regions to cerebrospinal fluid surrounding the brainstem, a threshold was applied to restrict the probabilistic masks to cortex (75% for the parahippocampal gyri, 50% for hippocampi) and any gaps resulting from thresholding manually filled. The resulting bilateral ROI was applied to each subject's fMRI data and used as a mask before conducting voxel‐wise statistics within this ROI.

fMRI Data Analysis: Model‐Free Analysis

To probe group‐level differences in functional connectivity during memory‐task performance, we used novel model‐free tensor independent component analysis (TICA) [Beckmann and Smith, 2005]. TICA analyses the temporal correlation of BOLD signals across all voxels in the brain and all subjects to identify “independent components,” each consisting of a spatial map, a time course and a subject‐mode vector (encoding the relative contribution of each subject to the spatio‐temporal effect). Spatial maps were thresholded using a Gaussian/Gamma mixture model [Beckmann and Smith, 2005] fitted to the histogram of intensity values and employing an alternative hypothesis testing approach at P > 0.5. Task‐related components were identified using post‐hoc regression analysis of each component's time course against the GLM design. Similarly, significant changes between groups were assessed by testing the subject‐specific effect sizes using a GLM.

Medial Temporal Lobe Laterality Index and Regional Percent Signal Change Calculation

Voxel‐based laterality indices (LIs) were calculated based on the ROI analysis. The MTL ROI, mapped onto each subject's fMRI image, was divided into left and right hemisphere components. A LI was obtained by calculating (L − R)/(L + R), where L is the number of activated voxels at a given statistical threshold in the left hemisphere mask, and R is the number of activated voxels at the same threshold in the right hemisphere mask. Voxel‐count based LIs are inherently limited by the arbitrary selection of statistical thresholds at which to obtain voxel counts. Only three patients showed ROI activation above a z‐statistical threshold of 3.3, therefore LIs were calculated only at the z‐statistical threshold of 2.3. Additionally we quantified mean and maximum regional BOLD percent signal change within the same ROIs using Featquery.

Diffusion Tensor Image Analysis

After pre‐processing (motion and eddy current correction, averaging of same‐direction diffusion acquisitions), principal eigenvector and fractional anisotropy (FA) maps were generated using FMRIB's diffusion tool (FDT). To track the fornix in individual patients, the single voxel with the highest FA value within the fornix was identified on individual data sets as previously described by Ringman et al. [ 2007]. Pathways from this voxel were tracked in each patient using probabilistic tractography (ProbtrackX within FDT) [Behrens et al., 2007] (http://www.fmrib.ox.ac.uk/fsl/fdt/). The resulting pathways were thresholded at 90% to ensure tracts were confined to the fornix. Mean and maximum FA values were extracted from each patient's thresholded fornix tract applied to their FA map. To track the uncinate fasciculus, the location of this fibre bundle was visually identified in each patient on sagittal brain slices by overlaying the principal diffusion direction vector map onto each subject's FA map. The uncinate typically spanned 3–4 sagittal slices at the level of the temporal stem. On the central‐most slice, a 3 × 3 voxel seed mask was drawn, spanning the posterior curve of the pathway in the left hemisphere. A target mask was placed in anterior prefrontal white matter on the standard MNI 152 template, and an exclusion mask placed directly behind this to prevent paths from continuing into grey matter (introducing high variability into FA samples). For uncinate analyses, tracts via fimbria‐fornix routes, and false‐positive tracts traversing the sylvian fissure were prevented using exclusion masks placed on one coronal slice of the standard MNI template brain at the posterior level of the anterior hippocampus (y = 51, or −24 mm), and along the sylvian fissure. Tracts were thresholded at a fifth of resulting paths before mean and maximum FA values were extracted from each individual's left uncinate tract applied to their FA map.

Statistical Analyses

Statistical analyses were conducted using SPSS (v13.0). One‐tailed Pearson's correlations were used to test the relationship between signal change measures and FA within the patient group as these were predicted to decrease in parallel. Two‐tailed Pearson's correlations were used to test relationships with neuropsychological scores.

RESULTS

Standard fMRI GLM Analyses

Group FEAT analysis revealed bilateral MTL activation in both LTLE patients and healthy controls during complex scene encoding. No significant differences in activation (e.g., in the right hemisphere) were found in patients relative to controls using standard GLM analysis (see Fig. 1).

Figure 1.

Figure 1

Open in a new tab

Whole‐brain group memory task‐related activations in healthy volunteers and LTLE patients. Group average maps during memory encoding in healthy controls (left) and LTLE patients (right). No statistically significant between‐group differences were found in the lateralization or localization of memory‐task related activation (contrast maps not shown, as these were empty).

ROI analysis yielded significant voxel‐wise activation in all controls and all but one of the patients. Voxel‐count LIs did not reveal any significant difference between LTLE patients (mean: −0.1, ±0.7) and healthy controls (mean: −0.1, ±0.4) (P = 0.862) (Fig. 3A). Additionally, neither mean, nor maximum percent signal change measures differed in either the left or the right MTL ROIs in patients compared with controls (Fig. 3B).

Figure 3.

Figure 3

Open in a new tab

Memory laterality index and MTL BOLD percent signal change in LTLE patients and healthy controls. Voxel‐count laterality indices (A) did not reveal inter‐hemispheric shifts in memory task‐related activation between LTLE patients and healthy controls. Regional mean and maximum fMRI signal change measures (B) similarly failed to identify significant group differences.

TICA Results

Tensor ICA analysis performed on the same fMRI memory‐task data identified one (out of 54) component which showed (i) significant correlation with the task and (ii) significant difference in the relative activation strength between patients and controls. The subject‐mode vector revealed a significant decrease in functional connectivity between the bilateral MTLs and a region encompassing the left gyrus rectus, medial and posterior orbital gyri in LTLE patients (see Fig. 2). This network was additionally associated with decreased functional connectivity to occipital regions encompassing bilateral lateral occipital cortex, the temporo‐occipital part of inferior temporal gyrus, occipital pole and a small cluster in left lingual gyrus, as well as small regions in the paracingulate gyrus, precentral gyrus, superior parietal gyrus, and thalami.

Figure 2.

Figure 2

Open in a new tab

Functional connectivity differences between LTLE patients and healthy volunteers identified from memory‐task derived TICA. TICA group difference map (left) demonstrating significantly reduced functional connectivity between the bilateral medial temporal lobes, an ipsilesional orbitofrontal region (green arrow), occipital regions, and small clusters in the paracingulate gyrus, precentral gyrus, superior parietal gyrus, and thalami in LTLE patients relative to healthy volunteers. A boxplot illustrating the separate patient and control group effect sizes and 1 standard deviation range is presented (right). All individual patient signal amplitudes fell within a range of two standard deviations from the mean of the patient group.

Relation of Structural and Functional Memory Measures

The region showing reduced functional connectivity with the MTLs in TLE patients appeared to localize to an orbitofrontal region containing limbic afferents [Barbas and De Olmos, 1990; Porrino et al., 1981; Ongur and Price, 2000], via the fornix [Poletti and Creswell, 1977; Votaw and Lauer, 1963]. Unthresholded tractography results in a single‐subject are presented in Supporting Information Fig. 1 for illustrative purposes only. To test for a relationship between fornix white matter coherence and memory‐related differences in functional connectivity, a mask was created based on activated voxels in the TICA‐derived group difference map. After applying temporal low‐pass filtering to each subjects' data in order to remove high‐frequency noise from the BOLD recordings, mean (average 0.1 ± 0.4) and maximum (0.2 ± 0.4) percent signal change were extracted from the orbitofrontal mask. Maximum FA along the fornix (average 0.72 ± 0.17) correlated significantly with both mean percent signal change from the orbitofrontal mask (r = 0.593, P = 0.046) and maximum percent signal change in this region (r = 0.607, P = 0.041) in the LTLE patients (see Fig. 4). No relationship was found between uncinate mean (average 0.38 ± 0.06) or maximum (average 0.83 ± 0.18) FA and BOLD signal change from the orbitofrontal mask.

Figure 4.

Figure 4

Open in a new tab

Correlation between fornix coherence measures and orbitofrontal fMRI signal change during memory task performance. Measures of white matter coherence along the fornix in LTLE patients correlated significantly (P < 0.05) with task‐related mean (A) and maximum (B) BOLD percent signal change in an orbitofrontal region showing altered functional connectivity with the medial temporal lobes in LTLE patients. OFG, orbitofrontal gyrus.

Correlating our MRI measures with clinical neuropsychological measures, one patient emerged as a clear outlier (see boxplot in Fig. 5). Excluding this patient, both maximum and mean BOLD signal change correlated strongly with clinical neuropsychological measures of story immediate (r = 0.88, P < 0.009; r = 0.84, P < 0.02 for max and mean, respectively) and delayed (r = 0.91, P < 0.004 and r = 0.81, P < 0.03 for max and mean, respectively) recall (see Fig. 5). No significant correlations were found between our diffusion‐based measures and neuropsychological scores with our sample size.

Figure 5.

Figure 5

Open in a new tab

Orbitofrontal fMRI signal correlation with story recall. fMRI maximum signal change in the region of the orbitofrontal cortex showing altered functional connectivity to the medial temporal lobes in LTLE patient during memory task performance correlated significantly with clinical neuropsychological measures of story immediate (A) and delayed (B) recall assessed using the adult memory and information processing battery. Neuropsychological scores were available for 8/9 patients reported. [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

DISCUSSION

Using novel analysis methods, we aimed to clarify the extent of memory‐related functional changes associated with left temporal lobe epilepsy at the systems‐level, and compare these to standard measures of hemispheric reorganization, measures of fibre pathway coherence and behavioral measures of memory function. To our knowledge, our preliminary study offers the first direct comparison of analyses aimed at demonstrating functional reorganization in TLE. Our results suggest that (i) LTLE is associated with network‐level changes, including connectivity with the unilesional frontal lobe and bilateral occipital regions, rather than inter‐hemispheric re‐distribution, during scene memory encoding and (ii) medial‐frontal functional changes correlate with white matter coherence in the fornix.

Much interest has been focused on understanding the relationship between hippocampal pathology and memory functions to help predict patients at risk of cognitive decline following surgery. It has been advanced that in the presence of chronic seizures affecting one of the MTLs, the contralateral MTL may adopt a compensatory role. Several recent functional imaging studies employing laterality index (LI) analysis methods have lent support to this notion. However, microscopic pathophysiological changes [Mathern et al., 1996] altering local connectivity may limit the potential for complete inter‐hemispheric reorganization. Several studies using direct electrical stimulation of the hippocampus have demonstrated preserved material‐specific deficits in TLE patients as a group [Coleshill et al., 2004; Lee et al., 1990], contra‐indicating absolute hemispheric functional reorganization. Additionally, the extent of ipsilesional fMRI activation during a variety of memory tasks has been convincingly shown to predict post‐surgical memory decline [Powell et al., 2008; Rabin et al., 2004; Richardson et al., 2004, 2006], casting doubt over the clinical significance of compensatory activity.

Consistent with these observations, we found no evidence for significantly altered MTL activation relative to healthy volunteers during a complex scene encoding task either at the whole‐brain level, or within a restricted ROI. Only qualitative decreases were observed in the magnitude of activation in the LTLE patient group as compared with healthy controls. While the biological significance of differences in BOLD signal amplitude remains unknown, a recent study [Petrella et al., 2007] suggests intrinsic differences in regional BOLD signal during task performance may be of prognostic value in disease populations. Testing systems‐level changes using novel model‐free multivariate TICA, however, identified decreased functional connectivity between the bilateral MTLs and a region in the left orbitofrontal gyrus (among others) in LTLE patients. As ICA methods are more robust against disease‐related changes in the hemodynamic response function (which may result in decreased sensitivity when using GLM methods), we propose that our finding supports subtle alterations in the contribution of distinct but inter‐connected regions of the pre‐existing memory network rather than inter‐hemispheric shifts in the distribution of memory functions between the MTLs. Altered recruitment of extra‐temporal regions, particularly within frontal lobes, has been previously found using memory tasks employing verbal stimuli [Dupont et al., 2000; Maccotta et al., 2007] and suggest dynamic changes in temporo‐frontal interactions during memory performance in TLE. Using an autobiographical retrieval task, these interactions were recently directly probed using effective connectivity analysis within a priori selected hippocampal and medial prefrontal regions (among others) [Addis et al., 2007]. The authors confirmed a weakening in the strength of functional connections during autobiographical retrieval in LTLE between these regions. Our results extend these findings to the encoding network, and additionally provide a sensitive unbiased method to test for systems‐level changes in a wide range of tasks without the need to limit the search to functional‐anatomical regions defined on previous task‐derived data.

Voxel‐based morphometry (VBM) studies in unilateral TLE patients have identified grey matter volume reductions in the thalamus and prefrontal cortex [Bonilha et al., 2004; Keller et al., 2002, Riederer et al., 2008] and are supported by observations of ipsilesional frontal lobe metabolic abnormalities [Henry et al., 1993; Savic et al., 1997]. These changes have been attributed to consequences of deafferentation, particularly axonal damage to temporal lobe white matter efferent tracts [Cendes et al., 1994], possibly secondary to hippocampal cell loss [Joo et al., 2005]. Consistent with this, white matter changes affecting the uncinate [Rodrigo et al., 2007] and fornix [Baldwin et al., 1994] have been previously reported. Such structural and metabolic changes may directly explain variations in behavioral performance, for example through Wallerian degeneration (e.g., Sawlani et al., 1997]. Indeed, DTI FA measures in the left temporal lobe have been related to fluency task laterality [Briellmann et al., 2003]. These findings suggest changes in memory performance and the distribution of associated neural correlates may be explained by injury more broadly affecting the network underlying memory functions.

Our results demonstrate a direct relationship between orbitofrontal signal change during memory task performance and coherence of one of the primary hippocampal efferent pathways, the fornix, in the same population of LTLE patients. The predominance of temporo‐frontal interactions are mediated by the uncinate and fornix white matter fibre bundles. Experimental lesion data suggest the uncinate may subserve particularly verbal memory functions such as word list learning [Niogi et al., 2008] rather than visual associative learning or recognition memory [Eacott and Gaffan, 1992; Gaffan and Eacott, 1995], which instead may be supported by the fornix (e.g., Aggleton et al., 2000] through hippocampo‐frontal interactions [Gaffan et al., 2002; Ridley et al., 1996]. Consistent with these models, we found a relationship between altered functional connectivity and fornix, but not uncinate fibre coherence using a scene‐encoding task. As the stimuli in our task were selected so as not to preferentially engage verbal encoding processes, we interpret the lack of significant association between uncinate coherence and our fMRI findings as support for the notion that the uncinate may be more specifically engaged during semantic processing. Indeed, two recent studies identified a significant correlation between diffusion parameters along the uncinate [Diehl et al., 2008] or parahippocampal gyrus tract [Yogarajah et al., 2008], which extends into the inferior frontal lobe via the uncinate [Powell et al., 2004] and neuropsychological test performance in LTLE relative to controls. These findings are not in conflict with our results, as in these studies, left uncinate diffusion measures (reduced in LTLE relative to controls) related to auditory immediate and delayed verbal memory while FA along the left parahippocampal gyrus (visually decreased in LTLE relative to the contralesional side) was associated with verbal learning.

These results suggest that white matter changes, potentially secondary to neurodegeneration of MTL structures, contribute significantly to functional pathology in TLE, and highlight a central role for the fornix in memory functions. This is consistent with a recent study of memory deficits arising from colloidal cyst resections emphasizing the crucial role of hippocampal projections via the fornix in sustaining long‐term memory functions, most notably for recall [Tsivilis et al., 2008], a measure which correlated with orbitofrontal signal change in our small population (see Fig. 5). Interestingly, a recent VBM observation of reduced grey matter density in the left orbitofrontal gyrus of LTLE patients with hippocampal sclerosis [Riederer et al., 2008] appears to co‐localize with our observed orbitofrontal functional connectivity differences. These authors additionally noted grey matter changes in thalamus, superior temporal gyrus, parahippocampal gyrus, right middle cingulum, and cerebellum, lending support to the notion that chronic seizures affect widespread interconnected regions connected to the hippocampus. Better characterization of changes in functional and structural connectivity will improve our understanding of pre‐operative memory deficits in TLE, and factors impacting on cognitive surgical outcome. Further research will need to be conducted to determine the relationship of these imaging markers to neuropsychological measures of memory impairment in this patient population.

An interesting additional result to emerge from our study was a decrease in functional connectivity in LTLE patients relative to controls between occipital regions and the MT‐IFG network involved in novel scene encoding. Visual processing has not been extensively studied in TLE with regard to neural correlates; however both Focke et al. [ 2008] and Yogarajah et al. [ 2008] identified differences in diffusion‐based measures in occipital or occipito‐temporal regions. Evidence for such structural changes may lend support to the notion advanced by Vannucci et al. [2008] that despite normal visual performance, TLE patients may sub‐optimally encode visual aspects of novel stimuli. Our results are consistent with this proposal and suggest decreased temporo‐occipital interaction during processing of novel scenes may underlie this encoding difference.

A few limitations warrant description. This is a preliminary study and the sample size is small. The linear spread of signal change and FA values suggest our results were not driven by a subgroup of the patients included in this trial, but will need replication in a larger group of participants. Additionally, our (absence of) laterality results are consistent with a previous larger study in 35 TLE patients [Rabin et al., 2004]. Signal loss and distortion near sinus tissue‐air boundaries affects sensitivity to BOLD activation in the MTL. Our observations of robust activations localized to the MTLs in both populations suggest this was not the primary cause for our lack of detection of hemispheric reorganization. Thirdly, we employed a single task designed to recruit bilateral structures for enhanced sensitivity to functional shifts and suitable for both LTLE and RTLE patients. For individual LTLE patients where verbal memory loss is of particular concern, additional domain‐specific tasks may be valuable. Lastly, one patient presented with cortical dysplasia rather than hippocampal sclerosis, which could have presented a confound in our analysis. TICA, however, intrinsically manages outlier confounds affecting (a) spatial extent or (b) signal amplitudes within a given map. If this patient had a different functional connectivity pattern as compared with the rest of the TLE group, TICA would have automatically extracted this into a separate component. If the patient had been an outlier in terms of BOLD signal amplitude within a unified map, this would have been reflected in the subject‐mode boxplot (see Fig. 2). As all individual patient values fell within 2 standard deviations of the rest of the patient group, our results appeared consistent across this pathophysiological difference. Our sample size was too small to adequately investigate the potential contributions of age at onset of seizures and duration of epilepsy (years) to functional plasticity, and warrants further investigation.

In conclusion, our results demonstrate a loss of functional connectivity between distinct anatomical regions involved in memory processing in patients with left TLE. These changes were not observed using commonly employed general linear model analysis and emphasize the value of a network‐based analysis to understanding memory (and possibly other) deficits in TLE. Signal change during task performance in the orbitofrontal region of LTLE patients was significantly related to coherence along the primary hippocampal efferent pathway, the fornix. These findings may have important implications for understanding disease‐related adaptive functional plasticity in TLE and has potentially important implications for predicting cognitive outcome following surgical intervention in TLE patients.

Supporting information

Additional Supporting Information may be found in the online version of this article.

Supplementary Figure 1: Illustrative single‐subject fornix and uncinate tractography results. Legend: Single subject unthresholded fornix (light blue) and uncinate tracts (dark blue), rendered onto the standard 1mm MNI template depicting the group memory‐task TICA result (red‐yellow). These single‐subject tracts are presented unthresholded for illustrative purposes only. All manuscript results are derived from thresholded pathways, which were restricted to white matter and thus did not extend into the cortical regions as presented here.

Click here for additional data file. (452.2KB, tif)

Acknowledgements

We acknowledge the Medical Research Council (MRC) and a private donation to the Cairns Fund for personal support to N.L.V. and P.M.M. J.E.A. received support from the Epilepsy Research Foundation (UK). We thank Laura Stamper and colleagues at FMRIB, particularly Heidi Johansen‐Berg, Tim Behrens, and Steve Smith; the participants; and Peter Hobden, the senior radiographer. P.M.M. is a full‐time employee of GlaxoSmithKline.

REFERENCES

  1. Addis DR, Moscovitch M, McAndrews MP ( 2007): Consequences of hippocampal damage across the autobiographical memory network in left temporal lobe epilepsy. Brain 130: 2327–2342. [DOI] [PubMed] [Google Scholar]
  2. Aggleton JP, McMackin D, Carpenter K, Hornak J, Kapur N, Halpin S, Wiles CM, Kamel H, Brennan P, Carton S, Gaffan D ( 2000): Differential cognitive effects of colloidal cysts in the third ventricle that spare or compromise the fornix. Brain 234 ( Part 4): 800–815. [DOI] [PubMed] [Google Scholar]
  3. Baldwin GN, Tsuruda JS, Maravilla KR, Hamill GS, Hayes CE ( 1994): The fornix in patients with seizures caused by unilateral hippocampal sclerosis: Detection of unilateral volume loss on MR images. Am J Roentgenol 162: 1185–1189. [DOI] [PubMed] [Google Scholar]
  4. Barbas H, De Olmos J ( 1990): Projections from the amygdala to basoventral and mediodorsal prefrontal regions in the rhesus monkey. J Comp Neurol 300: 549–571. [DOI] [PubMed] [Google Scholar]
  5. Beckmann CF, Smith SM ( 2005): Tensorial extension of independent component analysis for multisubject fMRI analysis. NeuroImage 25: 294–311. [DOI] [PubMed] [Google Scholar]
  6. Behrens TE, Johansen‐Berg H, Jbabdi S, Rushworth MF, Woolrich MW ( 2007): Probabilistic diffusion tractography with multiple fibre orientations: What can we gain? NeuroImage 34: 144–155. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Bonilha L, Rorden C, Castellano G, Pereira F, Rio PA, Cendes F, Li LM ( 2004): Voxel‐based morphometry reveals gray matter network atrophy in refractory medial temporal lobe epilepsy. Arch Neurol 61: 1379–1384. [DOI] [PubMed] [Google Scholar]
  8. Briellmann RS, Mitchell LA, Waites AB, Abbott DF, Pell GS, Saling MM, Jackson GD ( 2003): Correlation between language organization and diffusion tensor abnormalities in refractory partial epilepsy. Epilepsia 44: 1541–1545. [DOI] [PubMed] [Google Scholar]
  9. Cendes F, Andermann F, Preul MC, Arnold DL ( 1994): Lateralization of temporal lobe epilepsy based on regional metabolic abnormalities in proton magnetic resonance spectroscopic images. Ann Neurol 35: 211–216. [DOI] [PubMed] [Google Scholar]
  10. Cheung MC, Chan AS, Chan YL, Lam JM, Lam W ( 2006): Effects of illness duration on memory processing of patients with temporal lobe epilepsy. Epilepsia 47: 1320–1328. [DOI] [PubMed] [Google Scholar]
  11. Chlebus P, Mikl M, Brazdil M, Pazourkove M, Krupa P, Rektor I ( 2007): fMRI evaluation of hemispheric language dominance using various methods of laterality index calculation. Exp Brain Res 179: 365–374. [DOI] [PubMed] [Google Scholar]
  12. Coleshill SG, Binnie CD, Morris RG, Alarcon G, van Emde Boas W, Velis DN, Simmons A, Polkey CE, van Veelen CW, van Rijen PC ( 2004): Material‐specific recognition memory deficits elicited by unilateral hippocampal electrical stimulation. J Neurosci 24: 1612–1616. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Concha L, Beaulieu C, Gross DW ( 2005): Bilateral limbic diffusion abnormalities in unilateral temporal lobe epilepsy. Ann Neurol 57: 188–196. [DOI] [PubMed] [Google Scholar]
  14. Deblaere K, Backes WH, Tieleman A, Vandemaele P, Defreyne L, Vonck K, Hofman P, Boon P, Vermeulen J, Wilmink J, Aldenkamp A, Boon PA, Vingerhoets G, Achten E ( 2005): Lateralized anterior mesiotemporal lobe activation: Semirandom functional MR imaging encoding paradigm in patients with temporal lobe epilepsy—Initial experience. Radiology 236: 996–1003. [DOI] [PubMed] [Google Scholar]
  15. Detre JA, Maccotta L, King D, Alsop DC, Glosser G, D'Esposito M, Zarahn E, Aguirre GK, French JA ( 1998): Functional MRI lateralization of memory in temporal lobe epilepsy. Neurology 50: 926–932. [DOI] [PubMed] [Google Scholar]
  16. Diehl B, Busch RM, Duncan JS, Piao Z, Tkach J, Luders HA ( 2008): Abnormalities in diffusion tensor imaging of the uncinate fasiculus relate to reduced memory in temporal lobe epilepsy. Epilepsia 49: 1409–1418. [DOI] [PubMed] [Google Scholar]
  17. Dudek FE, Sutula TP ( 2007): Epileptogenesis in the dentate gyrus: A critical perspective. Prog Brain Res 163: 755–773. [DOI] [PubMed] [Google Scholar]
  18. Dupont S, Van de Moortele PF, Samson S, Hasboun D, Poline JB, Adam C, Lehericy S, Le Bihan D, Samson Y, Baulac M ( 2000): Episodic memory in left temporal lobe epilepsy: A functional MRI study. Brain 123: 1722–1732. [DOI] [PubMed] [Google Scholar]
  19. Eacott MJ, Gaffan D ( 1992): Inferotemporal‐frontal disconnection: The uncinate fascicle and visual associative learning in monkeys. Eur J Neurosci 4: 1320–1332. [DOI] [PubMed] [Google Scholar]
  20. Focke NK, Yogarajah M, Bonelli SB, Bartlett PA, Symms MR, Duncan JS ( 2008): Voxel‐based diffusion tensor imaging in patients with mesial temporal lobe epilepsy and hippocampal sclerosis. NeuroImage 40: 728–737. [DOI] [PubMed] [Google Scholar]
  21. Gaffan D, Eacott MJ ( 1995): Uncinate fascicle section leaves delayed matching‐to‐sample intact, with both large and small stimulus sets. Exp Brain Res 105: 175–180. [DOI] [PubMed] [Google Scholar]
  22. Gaffan D, Parker A, Easton A ( 2001): Dense amnesia in the monkey after transection of fornix, amygdala and anterior temporal stem. Neuropsychologia 39: 51–70. [DOI] [PubMed] [Google Scholar]
  23. Henry TR, Engel J Jr, Mazziotta JC ( 1993): Clinical evaluation of interictal fluorine‐18‐fluorodeoxyglucose PET in partial epilepsy. J Nucl Med 34: 1892–1898. [PubMed] [Google Scholar]
  24. Hermann BP, Seidenberg M, Haltiner A, Wyler AR ( 1992): Adequacy of language function and verbal memory performance in unilateral temporal lobe epilepsy. Cortex 28: 423–433. [DOI] [PubMed] [Google Scholar]
  25. Jansen A, Menke R, Sommer J, Forster AF, Bruchmann S, Hempleman J, Weber B, Knecht S ( 2006): The assessment of hemispheric lateralization in functional MRI‐robustness and reproducibility. Neuroimage 33: 204–217. [DOI] [PubMed] [Google Scholar]
  26. Jenkinson M, Bannister P, Brady M, Smith S ( 2002): Improved optimization for the robust and accurate linear registration and motion correction of brain images. Neuroimage 17: 825–841. [DOI] [PubMed] [Google Scholar]
  27. Joo EY, Hong SB, Han HJ, Tae WS, Kim JH, Han SJ, Seo DW, Lee KH, Hong SC, Lee M, Kim S, Kim BT ( 2005): Postoperative alteration of cerebral glucose metabolism in mesial temporal lobe epilepsy. Brain 128: 1802–1810. [DOI] [PubMed] [Google Scholar]
  28. Keller SS, Mackay CE, Barrick TR, Wieshmann EC, Howard MA, Roberts N ( 2002): Voxel‐based morphometric comparison of hippocampal and extrahippocampal abnormalities in patients with left and right hippocampal atrophy. Neuroimage 16: 23–31. [DOI] [PubMed] [Google Scholar]
  29. Lee GP, Loring DW, Smith JR, Flanigin HF ( 1990): Material specific learning during electrical stimulation of the human hippocampus. Cortex 26: 433–442. [DOI] [PubMed] [Google Scholar]
  30. Maccotta L, Buckner RL, Gilliam FG, Ojemann JG ( 2007): Changing frontal contributions to memory before and after medial temporal lobectomy. Cereb Cortex 17: 443–456. [DOI] [PubMed] [Google Scholar]
  31. Mathern GW, Babb TL, Leite JP, Pretorius K, Yeoman KM, Kuhlman PA ( 1996): The pathogenic and progressive features of chronic human hippocampal epilepsy. Epilepsy Res 26: 151–161. [DOI] [PubMed] [Google Scholar]
  32. Meiners LC, van der Grond J, van Rijen PC, Springorum R, de Kort GA, Jansen GH ( 2000): Proton magnetic resonance spectroscopy of temporal lobe white matter in patients with histologically proven hippocampal sclerosis. J Magn Reson Imaging 11: 25–31. [DOI] [PubMed] [Google Scholar]
  33. Niogi SN, Mukherjee P, Ghajar J, Johnson CE, Kolster R, Lee H, Suh M, Zimmerman RD, Manley GT, McCandliss BD ( 2008): Structural dissociation of attentional control and memory in adults with and without mild traumatic brain injury. Brain 131: 3209–3221. [DOI] [PubMed] [Google Scholar]
  34. Nunes RG, Jezzard P, Clare S ( 2005): Investigations on the efficiency of cardiac‐gated methods for the acquisition of diffusion‐weighted images. J Magn Reson 177: 102–110. [DOI] [PubMed] [Google Scholar]
  35. Ongur D, Price JL ( 2000): The organization of networks within the orbital and medial prefrontal cortex of rats, monkeys and humans. Cereb Cortex 10: 206–219. [DOI] [PubMed] [Google Scholar]
  36. Petrella JR, Prince SE, Wang L, Hellegers C, Doraiswamy PM ( 2007): Prognostic value of posteromedial cortex deactivation in mild cognitive impairment. PloS One 2: e1104. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Poletti CE, Creswell G ( 1977): Fornix system efferent projections in the squirrel monkey: An experimental degeneration study. J Comp Neurol 175: 101–128. [DOI] [PubMed] [Google Scholar]
  38. Porrino LJ, Crane AM, Goldman‐Rakic PS ( 1981): Direct and indirect pathways from the amygdala to the frontal lobe in rhesus monkeys. J Comp Neurol 198: 121–136. [DOI] [PubMed] [Google Scholar]
  39. Powell HWR, Guye M, Parker GJM, Symms MR, Boulby P, Koepp MJ, Barker GJ, Duncan JS ( 2004): Noninvasive in vivo demonstration of the connections of the human parahippocampal gyrus. NeuroImage 22: 740–747. [DOI] [PubMed] [Google Scholar]
  40. Powell HW, Richardson MP, Symms MR, Boulby PA, Thompson PJ, Duncan JS, Koepp MJ ( 2008): Preoperative fMRI predicts memory decline following anterior temporal lobe resection. J Neurol Neurosurg Psychiatry 79: 686–693. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Rabin ML, Narayan VM, Kimberg DY, Glosser G, Tracy JI, French JA, Sperling MR, Detre JA ( 2004): Functional MRI predicts post‐surgical memory following temporal lobectomy. Brain 127: 2286–2298. [DOI] [PubMed] [Google Scholar]
  42. Rao MS, Hattiangady B, Reddy DS, Shetty AK ( 2006): Hippocampal neurodegeneration, spontaneous seizures, and mossy fiber sprouting in the F344 rat model of temporal lobe epilepsy. J Neurosci Res 83: 1088–1105. [DOI] [PubMed] [Google Scholar]
  43. Reese TG, Heid O, Weisskoff RM, Wedeen VJ ( 2003): Reduction of eddy‐current‐induced distortion in diffusion MRI using a twice‐refocused spin echo. Magn Reson Med 49: 177–182. [DOI] [PubMed] [Google Scholar]
  44. Richardson MP, Strange BA, Thompson PJ, Baxendale SA, Duncan JS, Dolan RJ ( 2004): Pre‐operative verbal memory fMRI predicts post‐operative memory decline after left temporal lobe resection. Brain 127: 2419–2426. [DOI] [PubMed] [Google Scholar]
  45. Richardson MP, Strange BA, Duncan JS, Dolan RJ ( 2006): Memory fMRI in left hippocampal sclerosis: Optimizing the approach to predicting postsurgical memory. Neurology 66: 699–705. [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Ridley RM, Baker HF, Harder JA, Pearson C ( 1996): Effects of lesions of different parts of the septo‐hippocampal system in primates on learning and retention of information acquired before or after surgery. Brain Res Bull 40: 21–32. [DOI] [PubMed] [Google Scholar]
  47. Riederer F, Lanzenberger R, Kaya M, Prayer D, Serles W, Baumgartner C ( 2008): Network atrophy in temporal lobe epilepsy: A voxel‐based morphometry study. Neurology 71: 419–425. [DOI] [PubMed] [Google Scholar]
  48. Ringman JM, O'Neill J, Geschwind D, Apostolova LG, Rodriguez Y, Schaffer B, Varpetian A, Tseng B, Ortiz F, Fitten J, Cummings JL, Bartzokis G ( 2007): Diffusion tensor imaging in preclinical and presymptomatic carriers of familial Alzheimer's disease mutations. Brain 130: 1767–1776. [DOI] [PubMed] [Google Scholar]
  49. Rodrigo S, Oppenheim C, Chasoux F, Golestani N, Cointepas Y, Poupon C, Semah F, Mangin J‐F, Le Bihan D, Meder J‐F ( 2007): Uncinate fasiculus fibre tracking in mesial temporal lobe epilepsy. Initial findings. Eur Radiol 17: 1663–1668. [DOI] [PubMed] [Google Scholar]
  50. Rombouts SA, Barkhof F, Witter MP, Machielsen WC, Scheltens P ( 2001): Anterior medial temporal lobe activation during attempted retrieval of encoded visuospatial scenes: An event‐related fMRI study. Neuroimage 14: 67–76. [DOI] [PubMed] [Google Scholar]
  51. Rombouts SA, Damoiseaux JS, Goekoop R, Barkhof F, Scheltens P, Smith SM, Beckmann CF ( 2009): Model‐free group analysis shows altered BOLD fMRI networks in dementia. Hum Brain Mapp 30: 256–266. [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Savic I, Altshuler L, Baxter L, Engel J Jr ( 1997): Pattern of interictal hypometabolism in PET scans with fludeoxyglucose F 18 reflects prior seizure types in patients with mesial temporal lobe seizures. Arch Neurol 54: 129–136. [DOI] [PubMed] [Google Scholar]
  53. Sawlani V, Gupta RK, Singh MK, Kohli A ( 1997): MRI demonstration of Wallerian degeneration in various intracranial lesions and its clinical implications. J Neurol Sci 146: 103–108. [DOI] [PubMed] [Google Scholar]
  54. Seghier ML ( 2008): Laterality index in functional MRI: Methodological issues. Magn Reson Imaging 26: 594–601. [DOI] [PMC free article] [PubMed] [Google Scholar]
  55. Siddiqui AH, Joseph SA ( 2005): CA3 axonal sprouting in kainate‐induced chronic epilepsy. Brain Res 1066: 129–146. [DOI] [PubMed] [Google Scholar]
  56. Smith SM, Jenkinson M, Woolrich MW, Beckmann CF, Behrens TEJ, Johansen‐Berg H, Bannister PR, De Luca M, Drobnjak I, Flitney DE, Niazy R, Saunders J, Vickers J, Zhang Y, De Stefano N, Brady JM, and Matthews PM ( 2004): Advances in functional and structural MR image analysis and implementation as FSL. NeuroImage 23( S1): 208–219. [DOI] [PubMed] [Google Scholar]
  57. Stern CE, Corkin S, Gonzalez RG, Guimaraes AR, Baker JR, Jennings PJ, Carr CA, Suqiura RM, Vedantham V, Rosen BR ( 1996): The hippocampal formation participates in novel picture encoding: Evidence from functional magnetic resonance imaging. Proc Natl Acad Sci USA 93: 8660–8665. [DOI] [PMC free article] [PubMed] [Google Scholar]
  58. Vannest J, Szaflarski JP, Privitera MD, Schefft BK, Holland SK ( 2008): Medial temporal fMRI activation reflects memory lateralization and memory performance in patients with epilepsy. Epilepsy Behav 12: 410–418. [DOI] [PubMed] [Google Scholar]
  59. Vannucci M, Pezer N, Helmstaedter C, Schaller K, Viggiano MP, Elger CE, Grunwald T ( 2008): Hippocampal response to visual objects is related to visual memory functioning. Neuroreport 19: 965–968. [DOI] [PubMed] [Google Scholar]
  60. Votaw CL, Lauer EW ( 1963): An afferent hippocampal fiber system in the fornix of the monkey. J Comp Neurol 121: 195–206. [DOI] [PubMed] [Google Scholar]
  61. Yogarajah M, Powell HWR, Parker GJM, Alexander DC, Thompson PJ, Symms MR, Boulby P, Wheeler‐Kingshott CA, Barker GJ, Koepp MJ, Duncan JS ( 2008): Tractography of the parahippocampal gyrus and material specific memory impairment in unilateral temporal lobe epilepsy. NeuroImage 40: 1755–1764. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Additional Supporting Information may be found in the online version of this article.

Supplementary Figure 1: Illustrative single‐subject fornix and uncinate tractography results. Legend: Single subject unthresholded fornix (light blue) and uncinate tracts (dark blue), rendered onto the standard 1mm MNI template depicting the group memory‐task TICA result (red‐yellow). These single‐subject tracts are presented unthresholded for illustrative purposes only. All manuscript results are derived from thresholded pathways, which were restricted to white matter and thus did not extend into the cortical regions as presented here.

Click here for additional data file. (452.2KB, tif)

Articles from Human Brain Mapping are provided here courtesy of Wiley

RESOURCES