[Correction added after online publication 5-Jul-2012: Dr. O'Muircheartaigh's name has been updated.]
Address correspondence to Matthias Koepp, Department of Clinical & Experimental Epilepsy, UCL Institute of Neurology, Queen Square, London WC1N 3BG, U.K. E-mail: firstname.lastname@example.org
Purpose: Focal epilepsies are often associated with structural and functional changes that may extend beyond the area of seizure onset. In this study we investigated the functional anatomy of memory in patients with frontal lobe epilepsy (FLE), focusing on the local and remote effects of FLE on the networks supporting memory encoding.
Methods: We studied 32 patients with drug-resistant FLE and 18 controls using a functional magnetic resonance imaging (fMRI) memory encoding paradigm.
Key Findings: During encoding of stimuli, patients with FLE recruited more widely distributed areas than healthy controls, in particular within the frontal lobe contralateral to the seizure onset. Normal memory performance was associated with increased recruitment of frontal areas, and conversely a poor performance was associated with an absence of this increased recruitment and decreased activation in mesial temporal lobe areas.
Significance: In patients with FLE, recruitment of wider areas, particularly in the contralateral frontal lobe, appears to be an effective compensatory mechanism to maintain memory function. Impaired hippocampal activation is relatively rare and, in turn, associated with poor recognition memory.
Focal epilepsies are frequently accompanied by cognitive dysfunction, the causes of which are a matter of ongoing debate. Neuropsychological studies have shown that focal epilepsy may have both a local effect in the area of the seizure focus, and also remote effects on networks beyond the lobe containing the focus. A sizable proportion of temporal lobe epilepsy patients with temporal lobe epilepsy (TLE) have “frontal lobe dysfunction” on neuropsychological testing (Hermann & Seidenberg, 1995; Helmstaedter et al., 1996; Upton & Thompson, 1996; Martin et al., 2000; Bernhardt et al., 2008). Fluorodeoxyglucose–positron emission tomography studies have demonstrated reduced glucose metabolism in the frontal lobes of patients with drug-resistant TLE that correlates with cognitive dysfunction (Takaya et al., 2006), and this was reversible after successful epilepsy surgery (Spanaki et al., 2000). The rich interconnectivity between the temporal and frontal lobes may facilitate epileptic activity propagation and subsequent dysfunction in distant structures.
In contrast to the well-described cognitive profile of TLE, the cognitive profile in frontal lobe epilepsy (FLE) is less well characterized. Neuropsychological studies in FLE have focused on frontal lobe dysfunction (Helmstaedter et al., 1996; Upton & Thompson, 1996), but impairment of functions that are not typically “frontal,” such as memory encoding, have not been assessed systematically. Some studies have reported long-term memory impairment in patients with FLE showing dysfunction during encoding, free recall, and retrieval (Exner et al., 2002; Nolan et al., 2004), and there is also evidence of memory dysfunction following frontal lobe resection for epilepsy (McDonald et al., 2001).
The prevalence of memory impairment in FLE and whether the underlying mechanism is a frontal lobe dysfunction “per se” or a remote effect of FLE on temporal lobe regions remain unclear.
Functional magnetic resonance imaging (fMRI) studies using memory encoding paradigms are useful tools to characterize the networks involved in the encoding process in healthy population (Cabeza & Nyberg, 2000; Kim et al., 2010). These paradigms have been used to study the integrity of these networks in patients with TLE (Richardson et al., 2003, 2004; Wagner et al., 2008; Bonelli et al., 2010). In this study we employed an fMRI memory encoding paradigm together with out-of-scanner recognition testing to investigate the following:
1 Memory encoding and recognition performance in patients with FLE.
2 Characterization of the brain regions involved in memory encoding in patients with FLE.
3 The effect of FLE on the frontotemporal brain networks involved in memory encoding, assessing the local effect of the epileptic focus in the frontal lobe and the remote effect on the medial temporal lobe.
4 The effect of seizure focus lateralization on changes observed due to FLE.
5 The functional brain correlates of memory impairment in FLE.
We studied 32 patients (16 female) with a diagnosis of drug-refractory FLE recruited from the epilepsy clinics at the National Hospital for Neurology and Neurosurgery and King’s College Hospital (London, United Kingdom). Diagnosis of the patients was based on prolonged video-electroencephalography (EEG) monitoring, seizure semiology, and MRI. In addition, some patients had FDG-PET and ictal/interictal single-photon emission computed tomography (SPECT). Epileptic focus was located on the left frontal lobe in 19 patients and on the right frontal lobe in the remaining 13. Clinical data regarding the age at seizure onset, duration of epilepsy, antiepileptic medication and number and type of seizures as well as etiology of epilepsy were collected for each patient. Etiology was cryptogenic in 75% of patients. Small areas of focal cortical dysplasia were found in six patients and a single area of MRI signal abnormality of unknown nature was found in two patients. Location of lesions was concordant with the presumed epileptic focus. See Table 1 for population details.
Table 1. Sample characteristics
Seizures per month
Type of seizures
Median and mean values are shown for each variable. Full scale IQ was estimated with the National Adult Reading Test scale. SPS, simple partial seizures; SPS > SGTC, simple partial seizures progressing to secondary generalization; FCD, focal cortical dysplasia.
SPS SPS > SGTCS
16 2 1
Cryptogenic FCD Unspecific lesion
SPS SPS > SGTCS
8 4 1
Cryptogenic FCD Unspecific lesion
We recruited 18 healthy controls (12 female) with no history of neurologic disease, no family history of epilepsy, and normal structural MRI.
The study was approved by the Research Ethics Committee of the UCL Institute of Neurology and UCL Hospitals. Written informed consent was obtained from each subject.
MRI was acquired on a 3T General Electric Excite HD scanner (General Electric, Milwaukee, WI, U.S.A.).
For the fMRI paradigm gradient-echo echo-planar T2*-weighted images were acquired using the following parameters: Echo time (TE) of 25 msec and repetition time of 2.5 s. A total of 294 volumes were acquired. Each volume comprised 50 interleaved 2.4-mm slices with a 0.1-mm interslice gap, with an orientation parallel to the anterior to posterior cingulate (AC-PC) line. Images had a 64 × 64 matrix with a 24-cm field of view giving an in-plane pixel size of 3.75 mm. The scanner’s body coil was used for radiofrequency transmission, and the manufacturer’s standard eight-channel head coil was used for signal reception. An array spatial sensitivity encoding technique (ASSET) (parallel imaging) speed up factor of 2 was employed.
We used a memory encoding paradigm containing visual stimuli of different types. A total of 210 items were presented inside the scanner grouped in 30 s blocks of 10 pictures (black and white nameable line drawn objects), 10 words (single concrete nouns), or 10 faces (photographs unfamiliar to the subjects). Items were presented every 3 s and encoding blocks were separated by 15 s of cross-hair fixation. Subjects were instructed to actively memorize the items and indicate whether each item was pleasant or unpleasant by a right hand joystick response.
Subjects underwent a recognition test 60 min after the scanning in which the 210 presented items were randomly mixed with an additional 50% novel items.
Images were analyzed using statistic parametric mapping (SPM5) (http://www.fil.ion.ucl.ac.uk/spm/). Each subject’s images were realigned using the mean image as a reference, spatially normalized into Montreal Neurological Institute (MNI) space (using a scanner-specific template created from patient and control data) and smoothed with a Gaussian kernel of 8 mm full-width at half maximum.
Statistical fMRI analyses were performed first at the single subject level and then at the group level. In the single subject level analyses, trial-related activity was modeled by convolving a vector of block onsets with a canonical hemodynamic response function (HRF) to create regressors of interest. One regressor was modeled for each type of material (pictures, words and faces). Each subject’s movement parameters were included as confounds, and parameter estimates pertaining to the height of the HRF for each regressor of interest were calculated for each voxel. Contrasts for the effect of encoding pictures, words, and faces were built for each subject.
Second level group analyses were carried out as random-effects analysis. Individual contrasts were entered into three different factorial designs models to test the following questions:
Effect of frontal lobe epilepsy
Individual contrasts were submitted to a 2 × 3 analysis of variance (ANOVA) within SPM5. Cells were specified as 2 (group: controls vs. all 32 FLE patients) ×3 (material type: pictures, words and faces). The effect of FLE was investigated collapsed across the different material types.
Effect of the laterality of the focus
Patients with frontal lobe epilepsy were subdivided according to the laterality of the focus: 19 left, 13 right). Individual contrasts were submitted to a 3 × 3 ANOVA. Cells were specified as 3 (group-laterality: controls vs. left FLE vs. right FLE) ×3 (material type: pictures, words, and faces).The effect of epileptic focus lateralization was assessed relative to controls.
Lateralization indices of the maps left FLE versus controls and right FLE versus controls were calculated using the Lateralization Index (LI) toolbox for SPM5 (Wilke & Lidzba, 2007), that follows the equation:
Functional correlates of performance
Patients were subdivided into those with and without memory impairment according to their performance on the recognition test after the scan. To determine the impairment in a patient we compared the recognition accuracy of each patient with FLE to the control group mean by calculating the modified t-test developed for single case studies (Crawford & Garthwaite, 2002; Crawford et al., 2003). A patient was classified as having recognition memory impairment if the performance was significantly poorer than that of the controls when the alpha level was set at 0.05 (two-tailed). The modified t-test has been proposed as an appropriate tool to establish the abnormality of a score when the control sample against which the patient is compared is modest in size.
Individual contrasts were submitted to a 3 × 3 ANOVA, with cells specified as 3 (performance-group: controls vs. FLE good-performers vs. FLE poor performers) × 3 (material type: pictures, words, and faces).
Activations at the group level were reported as significant at a threshold of p < 0.05 corrected for multiple comparisons (family-wise error correction) in a whole brain analysis.
Because of the low signal-to-noise ratio in the anterior temporal lobe, activations in medial temporal lobe structures were reported as significant at a threshold of p < 0.001 uncorrected for multiple comparisons.
Hippocampal volumes were measured manually for each subject on T1-weighted scans. Volumes were estimated by measuring the area of the hippocampus on contiguous 1.5-mm–thick coronal slices throughout the whole anterior-posterior extent by using manually drawn boundaries (Woermann et al., 1998). Normal range of hippocampal volume measures was defined as control mean ± 2 standard deviations (SDs).
Responses of the recognition test were classified as remembered (hits), forgotten, and false alarms for the falsely recognized items. Recognition accuracy (RA) rates were calculated for each subject as: hit rate minus false alarm rate.
To investigate whether there was evidence of material specificity on memory dysfunction related to the side of epileptic focus in patients with FLE we performed a mixed-design ANOVA analysis. For this purpose we used the RA scores for words and faces as measures of verbal and nonverbal recognition as the within-subject factor and laterality of epileptic focus (left and right) as the between-subject factor.
We investigated the distribution of recognition memory performance in patients with FLE by comparing each patient’s mean RA across the three type of stimuli with the controls mean RA calculating z-scores. Patients were classified as significantly impaired when the alpha level was set at 0.05 (two-tailed).
Correlations of RA with the different clinical variables and IQ were investigated using Pearson’s correlation test.
Recognition accuracy (RA) was significantly lower in patients with FLE than in controls for all the items: Pictures (F1,46 = 4.5, p < 0.038), Words (F1,46 = 5.4, p < 0.024), Faces (F1,46 = 12.53 p < 0.001) (Fig. 1).
We observed a main effect of the material type (F1,28 = 115.96, p < 0.0001). Words RA (Mean [M] RA = 65.8) was better than faces (M RA = 24.2) in both patient with left and patients with right FLE. However, there was no significant interaction between material type and the side of epilepsy (F1,28 = 2.81 p < 0.14) (Fig. 2).
Seven (22%) of the 32 patients with FLE had significantly impaired RA scores compared to controls: three had a left-sided and four had a right-sided focus. There was no correlation with the side of epilepsy.
Functional MRI results
Consistent with previous memory encoding fMRI studies (Cabeza & Nyberg, 2000; Kim et al., 2010), activation maps for the effect of encoding items involved a number of frontal lobe areas and medial temporal structures comprising dorsolateral and ventrolateral prefrontal cortex, amygdala, hippocampus, and parahippocampal gyrus (Fig. 3, Table S1).
Activations in frontal lobe areas and medial temporal lobes were greater on the left during the encoding of words and pictures. Conversely, during faces encoding, the activation in the frontal and medial temporal lobe was greater on the right. This pattern was observed in controls and for both left and right FLE.
Effect of frontal lobe epilepsy
Patients with FLE demonstrated greater areas of activation within the frontal lobes relative to controls during the encoding across different types of stimuli. Bilateral clusters of increased activation were located in the middle frontal gyrus, perisylvian cortex, inferior frontal gyrus, and supplementary motor area (SMA) (Fig. 4, Table S2). No areas of greater activation were identified in the control group across the whole brain when compared to the FLE group.
Effect of epileptic focus laterality
We explored whether lateralization of the epileptic foci may affect functional activations.
Both left and right FLE showed greater frontal lobe activation than controls. Lateralization of these activations was contralateral to the side of epileptic focus on both groups of patients (Fig. 4B,C). No areas of lesser activation were found in left or right FLE groups relative to controls (Table S3).
Functional correlates of performance
There was a high variability in memory performance among patients with FLE: 7 of 32 patients with FLE had a mean RA score within the impaired ranents with FLE had mean RA scores within normal limits.
Comparing different performance groups showed the following:
1 Patients with FLE with normal memory showed greater activations compared to controls and to FLE patients with impaired memory in the middle and inferior frontal gyrus, bilaterally (Fig. 5A,B, Table S4).
2 Patients with impaired performance had decreased amygdala-hippocampal activation compared to controls and to FLE patients with normal recognition scores (Fig. 5C,D).
3 There was no difference in the frontal activations between controls and those patients with memory impairment.
Patients with left and right FLE had hippocampal volumes measurements within normal limits. Left FLE: mean left hippocampal volumes (SD) 2.67 cm3 (0.28), right hippocampal volumes 2.78 cm3 (0.28). Right FLE: left hippocampal volumes 2.77 cm3 (0.24), right hippocampal volumes 2.89 cm3, (0.3). There was no interaction between the side of epilepsy and the hippocampal volumes F30,1 = 0.46 (n.s.). Hippocampal volumes from FLE patients with impaired recognition memory were not significantly different from those patients with normal memory.
Memory performance and clinical variables
No significant correlations were found between RA scores and age at seizure onset, duration of epilepsy, number of antiepileptic drugs, etiology (cryptogenic vs. lesional), and frequency of seizures.
There was no correlation between IQ scores and RA for neither left nor right FLE patients.
Memory is a highly complex cognitive function and cannot be assigned to a circumscribed structure in the brain. It is known that the medial temporal structures are crucial for long term encoding, but the process of encoding as well as memory retrieval largely depend on a network involving temporal and frontal lobes.
In this study we explored the functional anatomy of memory in patients with FLE and how FLE affects memory networks. In particular, we analyzed the local effect of FLE on the frontal lobe component of the memory network and whether there was evidence for remote dysfunction in the temporal lobe.
Performance scores show that patients with FLE as a group are impaired compared to controls; however, there was high performance variability among patients, with only one fifth of patients with FLE showing significantly impaired recognition. We did not observe material-specific effects related to the lateralization of the epileptic focus.
We showed that patients with FLE recruit wider areas within the frontal lobes during the encoding process compared to controls, suggesting compensatory mechanisms. Effective compensation in patients can only be assessed if there are no differences in in-scanner task performance. The observed increases in activations are likely to represent compensatory mechanisms for two reasons. First analysis subdivided by side of seizure onset revealed that activations are more prominent in the hemisphere contralateral to the epileptic focus. A similar pattern has been reported in the side contralateral to the epileptic focus in patients with TLE in correlation with a maintained memory performance with and without memory impairment (Richardson et al., 2003, 2004; Bonelli et al., 2010). Secondly, increased frontal activations were present in the group of FLE patients with normal memory relative to controls and to FLE patients with memory impairment. The areas of increased activation comprise dorsolateral and ventrolateral prefrontal cortex. These areas have been implicated in successful memory formation, since they are involved in encoding item-specific information and relational processing during the encoding (Blumenfeld et al., 2010).
Medial temporal lobe activation was preserved in the majority of patients with FLE; however, the subgroup of patients with recognition memory impairments showed decreased amygdalar and hippocampal activation, suggesting a possible remote dysfunction in these areas in this subgroup of patients.
Patients with FLE may have a large range of cognitive dysfunction, ranging from severe impairment of attention, executive, and motor coordination skills to subtle personality traits (Helmstaedter, 2011). This heterogeneity has being attributed to the variability in seizure focus localization or differences in etiology or in the course of epilepsy observed among patients. Neuropsychological studies in this population (Helmstaedter et al., 1996; Upton & Thompson, 1996; Exner et al., 2002) have focused on testing frontal lobe functions, showing that patients with FLE are impaired on these domains. Of interest, these deficits have been found not to be specific to FLE, being also present in various degrees in other epilepsy syndromes such as TLE (Helmstaedter et al., 1996; Upton & Thompson, 1996; Exner et al., 2002). In a similar way, some studies reported patients with FLE to be impaired in learning and recall, functions that are supported by temporal lobes (Exner et al., 2002; Nolan et al., 2004; Helmstaedter, 2011).
Evidence for memory dysfunction in FLE patients varies widely between studies. Delaney et al. (1980) and Riva et al. (2002) found no memory impairment in adult subjects with FLE, whereas Nolan et al. (2004) found memory impairment in verbal and nonverbal domains in children with FLE but to a lesser degree than in patients with TLE. Exner et al. (2002) reported memory impairment in patients with FLE of a comparable severity to that in patients with TLE for immediate and delayed recall of visual and verbal items. Memory impairment has also been reported in autosomal dominant nocturnal frontal lobe epilepsy (ADNFLE) where all but one patient were found to be impaired on at least one memory measure, and for four patients the memory impairment was found to be more severe than executive dysfunction (Cho et al., 2008; Picard et al., 2009).
In our FLE sample we found impaired recognition performance in 22% of patients. This indicates that memory dysfunction is not a widespread deficit in this group and offers an explanation to the variable findings of the aforementioned studies. This high variability on memory performance among patients has also been reported in studies on subjects with lesions to the frontal lobes, suggesting that the different location of lesions may play a relevant role for developing memory impairment after damage to the frontal lobes (Bastin et al., 2006). Recent neuropsychological research has raised awareness about the role of the frontal lobes in the long-term memory process. The prefrontal cortex may deal with the organization and control of memory storage that takes place in medial temporal lobe structures (Shimamura, 1995) that contributes to successful memory. Distinct areas within the prefrontal cortex have subspecialized functions during the encoding process: the more ventral areas are the ones involved in processing item-specific information, whereas the dorsal areas deal with the relational memory (Blumenfeld & Ranganath, 2007; Long et al., 2010).
Our fMRI data provide evidence for the involvement of both the frontal and the medial temporal lobe areas in the impairment of memory function in patients with FLE. Normal recognition memory was associated with increased recruitment of frontal areas, contralateral to the epileptic focus, and, conversely, a poor performance was associated with an absence of this increased recruitment and decreased activation in mesial temporal lobe areas.
Frontal lobe activations in patients with FLE with poor memory performance were not significantly different from that in healthy controls, despite the decreased activation observed in the mesial temporal lobe structures. Decreased activations in the epileptogenic area have been reported as a group effect in patients with mesial TLE during memory tasks (Bonelli et al., 2010); however, our analysis did not revealed common areas of decreased activation within the frontal lobes across the whole FLE group or for the subgroup of patients with memory impairment. Activations during memory encoding task are seen within the lateral prefrontal cortex, The great interindividual variability on the location of the epileptogenic focus within the frontal lobes compared to patients with TLE may offer an explanation for this difference. The lack of common areas of decreased activation does not rule out the presence of individual dysfunctional regions that are not captured as group effects. Whereas the medial temporal lobe structures seem to be commonly dysfunctional in FLE patients with memory impairment, there is no common area of dysfunction in the frontal lobes of these patients.
Remote functional and structural changes have been widely reported in patients with TLE (Martin et al., 2000; Bernhardt et al., 2008; Keller et al., 2009). However, only two recent studies have included patients with FLE when exploring the remote effect of focal epilepsies. Vlooswijk et al. (2010) showed dysfunction in the frontotemporal connectivity for verbal tasks in both frontal and temporal focal epilepsy syndromes. Everts et al. (2010) correlated the atypical patterns of language lateralization in focal epilepsy syndromes (frontal and/or temporal epileptic focus), with temporal lobe function finding a correlation between the representation of language and memory performance regardless of the location of epileptic focus. In our study, the decreased activity in medial temporal lobes seen in patients with poor memory provides further evidence of remote dysfunction in FLE.
The impairment in function was not associated with a decrease in the hippocampal volumes. This suggests that the observed decrease of activation is not the result of structural damage but a remote functional effect in the mesial temporal lobe areas of these patients. We hypothesize that different locations of epileptic focus within the frontal lobe may explain why there is a decreased activation in only a subgroup of patients with FLE. It is possible that patients with epileptic foci located in areas with greater connectivity to the limbic system may have a greater degree of remote dysfunction in the medial temporal lobe structures. However, this hypothesis could not be further tested in our patients, since epileptic focus could only be lateralized but not further localized for the majority of the patients.
One limitation of our study is the effect of patient’s motivation and capacity to cope with on-line task demands. Although attention is monitored via the responses during the scanning process, we cannot rule out differences in their attention to and concentration on the task. Decreased fMRI signal has been reported in relation with poor engagement with the task in patients (Price & Friston, 1999), and this may play a role in the observed signal variability.
Our findings provide evidence for functional reorganizational changes in FLE. Frontal lobes are recruited during encoding processes, and compensatory activations within the frontal lobes, contralateral to the epileptic focus, are observed in patients with FLE. This functional reorganization is likely to be effective in the maintenance of memory function in this group.
This work was undertaken at UCLH/UCL, which received a proportion of funding from the Department of Health’s NIHR Biomedical Research Centres’ funding scheme. This project was funded by Wellcome Trust (Project Grant No 079474); The Big Lottery Fund, Wolfson Trust, and the Epilepsy Society (ES MRI scanner). MC thanks the EFNS scientific fellowships program and Caja Madrid postgraduate grants for their support. We are grateful to the radiographers at the National Society for Epilepsy MRI Unit—Philippa Bartlett, Jane Burdett, and Elaine Williams—who scanned the subjects and to all our subjects for their cooperation. The authors acknowledge infrastructure support from the National Institute for Health Research (NIHR) Specialist Biomedical Research Centre for Mental Health at the South London and Maudsley NHS Foundation Trust and the Institute of Psychiatry, King’s College London. This work was undertaken as part of UAB postgraduate program.
M. Centeno, C. Vollmar, J. O’Muircheartaigh, J. Stretton, S.B Bonelli, V. Kumari, and M.R. Symms report no disclosures. Prof. G. J. Barker serves on a scientific advisory board for and has received funding for travel and speaker honoraria from GE Healthcare and receives research support from the Medical Research Council United Kingdom, the Wellcome Trust, Guy’s and St Thomas, Epilepsy Research United Kingdom, and the Baily Thomas Charitable Fund. P. Thompson serves on the editorial board of Seizure and receives research support from the Wellcome Trust. Prof. J.S. Duncan has served on scientific advisory boards for GE Healthcare, Eisai Inc., and Sanofi-Aventis; has received funding for travel from Janssen-Cilag; serves on the editorial boards of Seizure, Epilepsy Research, and Epilepsia; may accrue revenue on a patent regarding a miniaturized wearable apnea detector; receives royalties from the publication of Eyelid Myoclonia and Typical Absences (Libbey, 1995); has received speaker honoraria from UCB and Eisai Inc.; has an active practice in epilepsy surgery; and receives research support from the Medical Research Council United Kingdom and the Wellcome Trust. Prof. M.P. Richardson has served on scientific advisory boards for Schwarz Pharma and UCB; has received funding for travel from Funding Janssen Cilag, UCB, and Eisai Inc.; serves on the editorial board of the Journal of Neurology, Neurosurgery and Psychiatry; and receives research support from the Medical Research Council United Kingdom, the Wellcome Trust, Epilepsy Research United Kingdom, the Charles Sykes Memorial Fund, King’s Medical Research Trust, and the Getty Family Foundation.
Prof. M.J. Koepp has served on scientific advisory boards for GE Healthcare; has received funding for travel from Desitin Pharmaceuticals, GmbH, UCB, and Pfizer Inc.; serves on the editorial boards of Epilepsy Research and Epileptic Disorders; receives research support from MRC, Wellcome Trust, and EU-Framework 7 programme; and he and his spouse own stock in GlaxoSmithKline. 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.