YEBEH-04269; No of Pages 12 Epilepsy & Behavior xxx (2015) xxx–xxx

Contents lists available at ScienceDirect

Epilepsy & Behavior journal homepage: www.elsevier.com/locate/yebeh

The effect of medial temporal lobe epilepsy on visual memory encoding A.M. Gregory a,c,⁎, R. Nenert a, J.B. Allendorfer a,d, R. Martin a,c,d, R.K. Kana c, J.P. Szaflarski a,b,d,e a

Department of Neurology, University of Alabama at Birmingham, Birmingham, AL, USA Department of Neurobiology, University of Alabama at Birmingham, Birmingham, AL, USA c Department of Psychology, University of Alabama at Birmingham, Birmingham, AL, USA d Department of UAB Epilepsy Center, University of Alabama at Birmingham, Birmingham, AL, USA e Department of Neurology, University of Cincinnati Academic Health Center, Cincinnati, OH, USA b

a r t i c l e

i n f o

Article history: Received 10 September 2014 Revised 3 March 2015 Accepted 7 March 2015 Available online xxxx Keywords: Epilepsy Temporal lobe Memory Neuronal plasticity fMRI Scene encoding

a b s t r a c t Effective visual memory encoding, a function important for everyday functioning, relies on episodic and semantic memory processes. In patients with medial temporal lobe epilepsy (MTLE), memory deficits are common as the structures typically involved in seizure generation are also involved in acquisition, maintenance, and retrieval of episodic memories. In this study, we used group independent component analysis (GICA) combined with Granger causality analysis to investigate the neuronal networks involved in visual memory encoding during a complex fMRI scene-encoding task in patients with left MTLE (LMTLE; N = 28) and in patients with right MTLE (RMTLE; N = 18). Additionally, we built models of memory encoding in LMTLE and RMTLE and compared them with a model of healthy memory encoding (Nenert et al., 2014). For those with LMTLE, we identified and retained for further analyses and model generation 7 ICA task-related components that were attributed to four different networks: the frontal and posterior components of the DMN, visual network, auditory-insular network, and an “other” network. For those with RMTLE, ICA produced 9 task-related components that were attributed to the somatosensory and cerebellar networks in addition to the same networks as in patients with LMTLE. Granger causality analysis revealed group differences in causality relations within the visual memory network and MTLErelated deviations from normal network function. Our results demonstrate differences in the networks for visual memory encoding between those with LMTLE and those with RMTLE. Consistent with previous studies, the organization of memory encoding is dependent on laterality of seizure focus and may be mediated by functional reorganization in chronic epilepsy. These differences may underlie the observed differences in memory abilities between patients with LMTLE and patients with RMTLE and highlight the modulating effects of epilepsy on the network for memory encoding. © 2015 Elsevier Inc. All rights reserved.

1. Introduction The importance of effective visual memory encoding for everyday functioning is undeniable. The visual memory encoding process requires both episodic memory and semantic memory to adequately encode visual stimuli and visual sensory information for later recall [1]. In addition to the hippocampus, other temporal lobe structures such as the amygdala and entorhinal cortex are involved in visual memory encoding [2], and the function of these structures may be impaired in patients with medial temporal lobe epilepsy (MTLE). Further, visual attention mechanisms required to enhance accurate visual encoding of complex scenes involve frontal and parietal networks that form

⁎ Corresponding author at: Department of Psychology and UAB Epilepsy Center, 312 Civitan International Research Center, 1719 6th Ave. South, University of Alabama at Birmingham, Birmingham, AL 35294, USA. Tel.: +1 619 681 3534. E-mail addresses: [email protected] (A.M. Gregory), [email protected] (R. Nenert), [email protected] (J.B. Allendorfer), [email protected] (R. Martin), [email protected] (R.K. Kana), szafl[email protected] (J.P. Szaflarski).

feedback loops with primary visual areas [3,4]. According to the nociferous cortex hypothesis, in patients with MTLE, extratemporal regions including the frontal and parietal attention networks are also affected [5]. These local and distributed effects may negatively influence the visual memory encoding networks in MTLE. Further, visual memory encoding involves semantic memory of visual stimuli [6] which previous studies have found to be differentially affected in patients with left MTLE and in patients with right MTLE [7,8] as the functional adequacy of the hippocampus ipsilateral to the seizure focus is often compromised and the capacity or reserve of the contralateral homologues may augment memory function [9]. Medial temporal lobe epilepsy is frequently reported to cause prominent memory deficits [10,11]. With seizures typically originating from the hippocampus and, less frequently, from other medial temporal structures, epilepsy-related damage in MTLE is thought to produce functional reorganization of memory processes [12–14], as these structures are involved in the acquisition, maintenance, and retrieval of episodic memories [15]. However, the extent of damage is confined not only to the medial temporal lobes [5]. The distributed nature of

http://dx.doi.org/10.1016/j.yebeh.2015.03.006 1525-5050/© 2015 Elsevier Inc. All rights reserved.

Please cite this article as: Gregory AM, et al, The effect of medial temporal lobe epilepsy on visual memory encoding, Epilepsy Behav (2015), http://dx.doi.org/10.1016/j.yebeh.2015.03.006

2

A.M. Gregory et al. / Epilepsy & Behavior xxx (2015) xxx–xxx

functional and anatomical abnormalities seen in MTLE is paralleled by widespread cognitive impairments affecting not only memory but also executive, sensorimotor, visuospatial, and language functions [16,17]. In neuroimaging studies, MTLE has been shown to differentially affect activation patterns in those whose seizures originate from the left or the right temporal lobe [6,12–14]. For example, LMTLE is usually associated with deficits in verbal functions, while patients with RMTLE typically exhibit impairments in visuospatial memory functioning [8]; both types of epilepsy are often associated with lateralization of memory encoding to the unaffected hemisphere [6,18–20]. Further, LMTLE with hippocampal sclerosis is commonly characterized by atypical language lateralization [21], and, in one study, left lateralization of a scene-encoding task was demonstrated in patients with RMTLE when compared to patients with LMTLE and healthy controls, while subject performance was similar, indicating plasticity of memory functions [6]. In line with the distributed nature of the functional damage caused by TLE, reorganization in the resting state networks has been observed in fMRI studies employing region of interest (ROI) approaches [22] as well as whole-brain network analyses including independent component analysis (ICA) [23–25]. In addition to the disruptions in the motor and sensory networks, numerous studies have demonstrated TLErelated abnormalities in the default mode [26–28] and alertness [29] networks, as well as networks subserving higher brain functions such as attention [30,31], executive control [27], and language [7]. For instance, the spatial extent of components included in the DMN was reduced in patients with MTLE, particularly in the prefrontal rather than temporoparietal cortices when compared to healthy controls [26]. Similarly, decreases in the functional connectivity of the frontal and parietal areas primarily included in the DMN and dorsal attention network (DAN) have been described in temporal lobe epilepsy (TLE) [32]. Another study reported impairments in the alertness network of patients with TLE when compared to healthy controls and demonstrated an association between the reduced overall area of activation and the patients' performance on a behavioral attention test [29]. Further, impairments in executive functions such as working memory are frequently observed in patients with TLE [16,33]. Finally, a recent fMRI study correlated decreases in working memory to lower functional connectivity between the prefrontal cortex, anterior cingulate, and inferior frontal gyrus, indicating widespread damage in focal-onset epilepsy [23]. The primary aim of the present study was to investigate the neuronal networks involved in visual memory encoding during a complex fMRI scene-encoding task in patients with LMTLE and in patients with RMTLE using advanced methods of image processing — group independent component analysis (GICA) and Granger causality analysis (GCA). Based on findings from the existing literature, we hypothesized that there would be differences in the visual memory encoding network between those with LMTLE and those with RMTLE [6,7,12,13,20,34–36]. The secondary goal included the following: (1) to build a model of memory encoding in LMTLE and RMTLE and (2) to compare these results with the model of healthy visual scene encoding [36]. The hypothesis guiding this portion of the study was that LMTLE and RMTLE would have differential effects on the neural underpinnings of the visual memory encoding while not causing specific differences in memory performance (i.e., memory plasticity).

Twenty-eight patients with LMTLE and 18 patients with RMTLE were identified and kept for subsequent analyses. The majority of subjects included in this study were previously reported in Bigras et al. [6]. Patients were 19 to 66 years of age (M = 38 years, SD = 12), 70% were male, with a mean age at seizure onset of 20 years and a mean duration of epilepsy of 19 years (Table 1). Independent samples t-test and chisquare test of independence revealed no significant differences between groups, both p N 0.05. Results of this study were compared to previously reported findings of 40 healthy controls whose demographics are reported elsewhere [6,36]. Briefly, healthy control participants were 19– 59 years of age (M = 33), with no history of neurological disorders. This study was approved by the University of Cincinnati and University of Alabama at Birmingham Institutional Review Boards, and all participants provided written informed consent prior to enrollment. 2.2. Visual scene-encoding task For the purpose of this study, we employed a previously wellcharacterized block-design fMRI scene-encoding task [13]. Briefly, during the active blocks, the participants were presented with stimuli that represented a balanced mixture of indoor (50%) and outdoor (50%) scenes including images of inanimate objects, people, and faces [13]. To assess memory encoding, we explicitly instructed participants to memorize all scenes for later (postscan) test of memory retrieval. Attention to the task was monitored by participants indicating whether the scene was indoor or outdoor via a corresponding left or right button press. The response box was held in the right hand. In the control blocks, participants viewed pairs of scrambled images presented side-by-side and were instructed to indicate (using the same response box) whether the images in each pair were the same or not (50% of pairs contained images that were the same). Scrambled pairs with different images were similar in brightness and hue. Scene encoding involves several cognitive processes with the target one being encoding of visually presented stimuli. The control condition allows for the subtraction of visuoperceptual, decision-making, and motor aspects of the task. The task was composed of 14 blocks (10 images per block). Blocks alternated between images of indoor or outdoor scenes (i.e., 7 active blocks) and images of a pair of scrambled pictures (i.e., 7 control blocks) for a total of 70 target scenes and 70 control pairs. The paradigm lasted 7 min and 15 s, and each image was presented for 2.5 s, followed by 0.5 s of a white blank screen. To ensure comprehension of the task, participants completed a practice run before entering the scanner. Practice items consisted of five indoor and outdoor scenes as well as five side-byside scrambled image pairs. Participants were required to accurately respond to all practice items before entering the scanner. Directly after completing the in-scanner task, participants were administered a postscan recognition test that included 60 indoor and outdoor scenes, with a balanced content of target and foil pictures. Foil pictures matched the content and parameters of those presented in the scanner. Participants were instructed to indicate whether they remembered Table 1 Demographic characteristics of the patients with left medial temporal lobe epilepsy (LMTLE) and the patients with right medial temporal lobe epilepsy (RMLTE) (AEDs — antiepileptic drugs).

2. Methods 2.1. Participants Sixty-eight patients with epilepsy were recruited as part of a larger study [6,35]. Of those patients, 13 were excluded from analyses because they were administered a different version of the task [34]. Additionally, patients with extratemporal or lesional epilepsy were excluded; diagnosis of LMTLE or RMTLE for all patients was determined based on video-EEG monitoring, seizure semiology, and neuroimaging results.

Age Male Education in years Age at onset Duration of epilepsy Right-handedness Total number of current AEDs

LMTLE (N = 28)

RMTLE (N = 18)

36.25 (10.16) 61% 14.14 (2.70) 18.36 (13.39) 17.89 (13.05) 89% 2.21 (.83)

42.00 (12.64) 83% 13.44 (2.28) 25.39 (17.30) 16.22 (12.52) 89% 1.89 (.68)

Note: Independent samples t-test an chi-square test of independence revealed no significant differences between groups, all p N 0.05. Demographic characteristics of the healthy controls were previously reported [36].

Please cite this article as: Gregory AM, et al, The effect of medial temporal lobe epilepsy on visual memory encoding, Epilepsy Behav (2015), http://dx.doi.org/10.1016/j.yebeh.2015.03.006

A.M. Gregory et al. / Epilepsy & Behavior xxx (2015) xxx–xxx

seeing the picture in the scanner by pressing “Y” or “N” on a standard laptop keyboard for “yes” and “no” answers, respectively. 2.3. Functional MRI All images were collected on a 4-Tesla Varian MRI scanner. Head movement was minimized with the use of head restraints and foam padding. Participants were provided a button box to record responses and to alert the MRI technologist of any problems if needed. An anatomical T1 scan was collected first, TR = 13 ms, TE = 6 ms, FOV = 25.6 × 19.2 × 15.0, flip angle array = 3:22/90/180, voxel size = 1 × 1 × 1 mm. Next, manual shimming was performed, followed by echo planar imaging (EPI) which was completed in thirty 4-mm thick contiguous planes sufficient to encompass the apex of the cerebrum to the inferior aspect of the cerebellum. The protocol for EPI scans was as follows: TR/TE = 3000/25 ms, FOV = 25.6 × 25.6 cm, matrix = 64 × 64 pixels, slice thickness = 4 mm, flip angle array = 85/180/ 180/90. Task stimuli were presented using Psyscope [37] running on an Apple Macintosh G3 computer. 2.4. Functional MRI — data analysis Spatial preprocessing and statistical analysis were performed using MATLAB toolbox SPM8 (http://www.fil.ion.ucl.ac.uk/spm/software/ spm8/). Functional images were corrected for time discrepancy between slices (slice timing in interleaved mode with the second to the last slice used as reference), corrected for motion, normalized using the EPI-weighted template, with trilinear interpolation and 2 × 2 × 2-mm voxel size, and spatially smoothed with an 8-mm kernel full-width half-maximum. To allow for magnetic equilibration, we removed the first five volumes from further analysis. 2.5. General linear model First-level analysis using standard GLM methods was conducted in order to determine whether activations obtained with this task were comparable with the results of our previous studies that used the same fMRI task [6,35] and with the results of similar investigations from the literature [38,39]. Specifically, for the analyses, we used TR of 3000 ms, with 30-slice resolution and with slice 28 serving as a reference for slice timing. Slow drifts in the fMRI data were corrected using a high-pass filter set at 128 s. Two regressors were modeled: 1) active block (indoor/outdoor scenes) and 2) control block (scrambled scene pairs). The duration of each block was set at 30 s. To control for brain activation of no interest, we subtracted the control task from the active task. Visual comparison of the GLM maps indicated the results to be similar to the previously published data; these comparisons are not presented here [6]. 2.6. Independent component analysis (ICA) 2.6.1. Combined group ICA of participants with MTLE Group ICA is commonly used for making group inferences from fMRI data of multiple subjects. Independent component analysis is a datadriven technique that does not rely on a priori hypotheses about brain activity and is, therefore, particularly useful when the temporal synchrony of brain activity is altered by neuropathology [40]. For example, several studies have shown varying patterns of blood oxygenation level-dependent responses related to interictal epileptiform discharges [41–43] as well as age-dependent differences in patients with epilepsy [44]. Thus, for an exploratory analysis, we performed GICA of all 46 subjects to examine the presence of the hypothesized differences between groups. Group ICA was executed using the ICA approach [40]. The GIFT toolbox (http://icatb.sourceforge.net) for group ICA was used for estimating individual spatial patterns and for investigating group

3

differences. The preprocessed individual datasets were concatenated and reduced through three stages of principal component analysis to obtain the final dataset. Next, estimation of independent sources was performed using the Infomax algorithm [45]. In accordance with the common settings in previous ICA–GCA studies [46–50], the number of components was estimated using the minimum description length criteria [51]. Thirty-three spatially independent components were computed. The Infomax algorithm was repeated twenty times with randomly initialized decomposition matrices and the same convergence threshold using the ICASSO approach [52] in the GIFT toolbox. After clustering the obtained components, centrotype-based components were selected and considered a stable result of the decomposition. Finally, the individual spatial patterns and time courses were obtained through GICA 3 back reconstruction algorithm [53]. Visual inspection of the temporal models identified eight components correlated with the task, and only those were retained for further analyses. Group independent component analysis data for the retained components were then entered into one-sample t-tests in SPM12 to create a mask for analysis of group differences of each component. To directly compare groups, we conducted F-tests on four of the most taskrelated components to determine whether there were any differences between the two groups across the task-related components. Results of the exploratory analyses of GICA data showed differences between the entire group on three of the four most task-related components. For example, there was a significant main effect for group F(1,43) = 18.31, p b 0.001, on IC 09 (medial temporal region). These results demonstrate a difference in visual memory encoding between those with LMTLE and those with RMTLE; thus, further group ICAs were performed separately.

2.6.2. Group ICA of participants with left MTLE and participants with right MTLE Group ICA was performed separately on each group to investigate group differences under the same study condition. Thirty-three spatially independent components were computed for the group with LMTLE and 34 for the group with RMTLE. The spatial and temporal models of the group components were first visually inspected in the composite viewer of GIFT. Visual inspection of the spatial models identified 18 components representing signal noise for LMTLE and 15 for RMTLE. Visual inspection of the temporal models identified eight components for LMTLE and ten components for RMTLE as uncorrelated with the task. These results were then confirmed using the correlation criteria (r) within the temporal sorting tool in GIFT and were subsequently discarded.

2.7. Granger causality analysis (GCA) Granger causality analysis is a test of predictive causality that is based on the notion that if information about the past of a time series X adds information to the predictive value of the likelihood of time series Y occurring, above and beyond information about the past of time series Y alone, then X is, therefore, Granger-causal [54]. Granger causality analysis has been shown to be a viable technique to analyze fMRI data [55,56] and has been used to investigate effective connectivity in patients with epilepsy [27,57]. Granger causality analysis was performed using a MATLAB toolbox [58], the implementation of which is based on vector autoregressive modeling. Specifically, the time series of the GICA components were modeled as weighted sums of their past values to produce models of one directional causality among the multiple series [59]. The number of past observations to use in the regression model, that is, the best model order to use, is estimated using the Bayesian Information Criterion (BIC) [60]. Statistical testing of causal interactions was corrected for multiple comparisons with the false discovery rate (FDR), with the significance level set at p b 0.05.

Please cite this article as: Gregory AM, et al, The effect of medial temporal lobe epilepsy on visual memory encoding, Epilepsy Behav (2015), http://dx.doi.org/10.1016/j.yebeh.2015.03.006

4

A.M. Gregory et al. / Epilepsy & Behavior xxx (2015) xxx–xxx

3. Results

higher compared to patients with RMTLE (mean rank = 17.17) on the postscan recognition test, U = 119, z = −4.809, p b 0.001. Distributions of the performance scores for both patients with LMTLE and patients with RMTLE were similar across all conditions (active, control, and postscan recognition). No significant differences were observed between groups in the active (indoor/outdoor scenes), U = 198, z = − 0.942, p = 0.346, and control (scrambled scenes), U = 230, z = − 0.188, p = 0.851, conditions. No significant differences were observed between patients with LMTLE and patients with RMTLE on the postscan recognition testing, U = 208.50, z = − 0.692, p = 0.489.

3.1. Behavioral analyses

3.1.2. Within group

3.1.1. Between groups There were no differences in demographics between the groups with MTLE (Table 1). Further, we compared the groups with MTLE with the healthy control group [36] on age, education, and handedness using independent samples t-test and chi-square test of independence. There was no age difference found between healthy controls (33.33 ± 12.37) and patients with LMTLE (36.25 ± 10.16), t (66) = − 1.031, p = 0.306. However, an age difference was found between healthy controls and patients with RMTLE (42.00 ± 12.64), t (56) = − 2.45, p = 0.017. Healthy controls were more educated (15.78 ± 2.74) than both left (14.14 ± 2.70) and right MTLE groups (13.44 ± 2.28), a statistically significant difference of 1.63 (95% CI, 0.29 to 2.97) years for LMTLE, t (66) = 2.43, p = 0.018, and 2.33 (95% CI, 0.85 to 3.81) years for RMTLE, t (56) = 3.15, p = 0.003. Chi-square test of independence indicated that there was no significant difference between left or right MTLE groups and healthy controls across gender (χ2 (4) = 4.041, p = .401) and handedness (χ2 (2) = 4.145, p = 0.126). The assumption of cell frequency was violated for handedness; however, an examination of the likelihood ratio confirmed that the groups were not statistically significantly different, p = 0.13. Because of the non-Gaussian distribution of performance scores, Mann–Whitney U test was performed to determine the differences between the groups, with the unit of measurement being mean percent of scenes correctly answered. Table 2 highlights the task performances of the groups. Distributions of the performance scores were different between the control group and the group with LMTLE and the group with RMTLE across all conditions (active, control, and postscan recognition). Healthy controls' median performance score was higher than that of patients with LMTLE in the active (indoor/outdoor scenes), U = 423, z = − 2.308, p = 0.021, and control (scrambled scenes), U = 214.50, z = − 4.716, p b 0.001, conditions. Differences were also observed between healthy controls and patients with LMTLE on the postscan recognition testing, U = 209, z = −4.683, p b 0.001. When compared to the group with RMTLE, healthy controls' median performance score was higher in both the active, U = 192.50, z = − 3.872, p b 0.001, and control, U = 149.50, z = −4.450, p b 0.001, conditions. Finally, healthy controls (mean rank = 39.95) performed significantly

3.1.2.1. Left medial temporal lobe epilepsy. An exact sign test with continuity correction was used to compare the differences in task performance, mean percent of scenes answered correctly, between the active (encoding) and control conditions. Of the 28 patients with LMTLE, 18 had better mean performance scores on the active sceneencoding task than on the control task, 9 had worse performance scores on the active task than on the control task, and one performed equally on the active and control tasks. Overall, patients with LMTLE performed better on the active task than on the control task (89% vs. 85% correct), but this difference was not significant (p = 0.124).

2.8. Statistical analysis Demographic and behavioral data were analyzed using SPSS (version 22). Independent samples t-tests (two-tailed) were used to assess differences between the group with left MTLE and the group with right MTLE. Performance on the encoding task and the postscanning performance were assessed using independent samples t-tests (twotailed) and, when appropriate, nonparametric statistical test (Mann– Whitney U test). The significance level was set at 0.05.

Table 2 Median percent of correct responses across tasks for the group with left medial temporal lobe epilepsy (LMTLE) and the group with right medial temporal lobe epilepsy (RMLTE) and for the healthy controls. fMRI task

LMTLE (N = 28)

RMTLE (N = 17)

Control (N = 40)

Active Control Recognition

89 85 78

89 87 76

91 94 86

Note: Active task (indoor vs. outdoor); Control task (same vs. different); Recognition task (post-scan testing of recognized scenes). Distributions of performance scores of both left and right MTLE groups were significantly different from healthy controls, p b 0.05. Distributions of performance scores were not significantly different between those with left and those with right MTLE, p b 0.05. Performance data were missing for one subject with RMTLE due to technical reasons.

3.1.2.2. Right medial temporal lobe epilepsy. An exact sign test with continuity correction was also used here, where units of measurement are mean percent of scenes answered correctly. Eight had better mean performance scores on the active task than on the control task, 6 had worse performance scores, and three performed equally on both tasks. Overall, patients with RMTLE performed better on the active task than on the control task (89% vs. 84% correct), but this difference was not significant (p = 0.791). 3.2. Independent component analysis 3.2.1. Left medial temporal lobe epilepsy Thirty-three components were produced by GICA. Of those, seven were identified as correlated with the task and retained for further analyses and model generation (see Section 2.6). Cortical localizations of the task-related components are depicted in Table 3. The retained components were attributed to four different networks: the frontal and posterior components of the DMN (ICs 01 and 02, respectively), visual network (ICs 03, 14, and 09), auditory-insular network (IC 08), and an “other” network (IC 33). Independent component (IC) 01 is located in the superior and medial frontal regions of the DMN, and IC 02 is located in the posterior/retrosplenial region of DMN. Independent component (ICs) 03 and 14 are primarily located in the cuneus, lingual gyrus, declive, posterior cingulate, middle and inferior occipital lobes, and the fusiform gyrus. IC 09 is attributed to the ventral stream of visual processing starting in V1 and ending in the inferior temporal cortex. IC 08 was identified as part of the auditory-insular network and is primarily located within the insula and primary and association auditory cortices. IC 33 is located in several brain regions including the posterior cingulate, parahippocampal gyrus, frontal and parietal lobes, lingual gyrus, and superior temporal gyrus. While posterior cingulate has been implicated in numerous memory-related processes [23], IC 33 is hypothesized to be a compensatory mechanism of impaired visual attention and auditory processes and is identified as the “other” network (see the Discussion section). 3.2.2. Right medial temporal lobe epilepsy Group independent component analysis produced 34 components, of which nine were identified as correlated with the task and retained for further analyses (see Section 2.6). Cortical localizations of the taskrelated components are depicted in Table 4. These nine ICs were

Please cite this article as: Gregory AM, et al, The effect of medial temporal lobe epilepsy on visual memory encoding, Epilepsy Behav (2015), http://dx.doi.org/10.1016/j.yebeh.2015.03.006

A.M. Gregory et al. / Epilepsy & Behavior xxx (2015) xxx–xxx

5

Table 3 Cortical localizations of task-related independent components in patients with LMTLE (– indicates that this independent component was not observed on one side). Component ID

Area

Brodmann area

Peak z-score and its spatial (x, y, z) L/R correlates

1

Superior frontal gyrus Middle frontal gyrus Medial frontal gyrus Posterior cingulate Cingulate gyrus Cuneus Precuneus Inferior parietal lobule Angular gyrus Middle temporal gyrus Superior temporal gyrus Lingual gyrus Cuneus Declive Middle occipital gyrus Inferior occipital gyrus Fusiform gyrus Culmen Declive of vermis Insula Superior temporal gyrus Transverse temporal gyrus Middle temporal gyrus Precentral gyrus Postcentral gyrus Culmen Declive Fusiform gyrus Parahippocampal gyrus Lingual gyrus Cuneus Posterior cingulate Lingual gyrus Culmen of vermis Culmen Precuneus Parahippocampal gyrus Culmen Posterior cingulate Parahippocampal gyrus Precuneus Postcentral gyrus Thalamus Medial frontal gyrus Culmen of vermis Superior frontal gyrus Inferior frontal gyrus Superior parietal lobule Paracentral lobule Lingual gyrus Superior temporal gyrus Middle frontal gyrus Cuneus

6, 8 6, 8 6, 8 23, 29, 30, 31 23, 24, 31 7, 18, 19 7, 19, 23, 31, 39 7, 39, 40 39 39 39 17, 18 17, 18, 23, 30

11.5 (−2, 30, 56)/11.1 (6, 26, 56) 6.4 (−18, 18, 56)/5.8 (26, 26, 54) 4.1 (−2, 47, 42)/4.8 (6, 47, 42) 7.7 (0, −45, 24)/6.5 (4, −47, 23) 7.3 (0, −45, 28)/6.7 (4, −47, 26) 7.2 (0, −66, 31)/6.8 (4, −68, 33) 6.9 (0, −66, 35)/6.8 (2, −70, 37) –/3.3 (44, −66, 40) –/3.2 (50, −63, 31) –/3.0 (51, −61, 27) –/2.7 (51, −57, 27) 9.5 (0, −87, −1)/9.9 (0, −84, −3) 8.4 (0, −91, 3)/7.9 (4, −91, 3) 5.8 (−2, −80, −11)/5.5 (12, −80, −13) 5.8 (−8, −93, 14)/3.2 (12, −94, 14) 3.9 (−16, −90, −7)/4.9 (14, −92, −9) 4.0 (−20, −80, −14)/4.8 (20, −86, −13) –/3.1 (14, −68, −8) 2.6 (0, −72, −12)/– 6.9 (−44, −17, 3)/5.1 (44, −21, −1) 6.5 (−44, −19, 6)/5.4 (46, −17, 1) 5.3 (−44, −21, 10)/3.5 (42, −25, 10) 4.8 (−51, −16, −4)/4.7 (55, −12, −3) 4.4 (−46, −11, 8)/3.0 (48, −9, 6) 3.3 (−51, −26, 14)/– 8.4 (−28, −51, −16)/7.8 (26, −49, −14) 8.1 (−26, −55, −14)/7.5 (26, −53, −14) 5.3 (−26, −51, −9)/5.5 (30, −51, −9) 4.4 (−22, −47, −9)/5.2 (24, −45, −10) 3.7 (−26, −70, −10)/3.2 (20, −80, −13) 7.2 (−2, −64, 7)/7.5 (6, −64, 7) 6.9 (−2, −60, 7)/7.4 (2, −60, 7) 6.3 (−6, −64, 3)/7.2 (6, −64, 3) 5.7 (−2, −62, 0)/7.0 (6, −60, 1) 5.6 (−6, −58, 0)/6.5 (6, −56, −1) 4.5 (0, −67, 18)/4.4 (2, −63, 18) 2.7 (−18, −51, −4)/4.0 (18, −53, −4) 10.8 (0, −47, 1)/9.4 (4, −45, 1) 7.3 (−4, −52, 6)/6.9 (4, −52, 6) 6.2 (−8, −43, 2)/5.7 (8, −39, 2) 4.5 (−2, −55, 62)/5.6 (6, −59, 62) 3.6 (−6, −53, 67)/5.3 (6, −51, 65) 4.3 (−4, −21, 7)/3.9 (4, −21, 7) 2.6 (−4, 56, 34)/4.2 (6, 51, 42) 4.1 (−4, −60, 1)/3.8 (4, −60, 1) –/4.0 (10, 51, 42) 3.6 (−57, 27, 0)/– –/3.5 (6, −65, 57) –/3.4 (4, −44, 61) 3.1 (−8, −58, 5)/2.6 (8, −58, 3) 2.7 (−57, 7, −9)/– –/2.7 (42, 39, 37) 2.6 (−8, −58, 8)/–

2

3

8

9

14

33

18 17, 18 18, 19

13, 22, 41 13, 21, 22, 38, 41, 42 41, 42 21, 22 6 40, 43

19, 20, 37 19, 36, 37 18 17, 18, 23, 30 23, 29, 30, 31 18, 19

23, 31 18, 19, 30 29, 30 27, 30 7 4, 5, 7 8, 9 8, 9 45, 47 7 5 18, 19 38 8, 9 30

Note: Table includes the anatomical location, the corresponding Brodmann area(s), and the maximum z-score with its Talairach coordinates for each IC.

attributed to six different networks: the frontal and posterior components of the DMN (ICs 17 and 01, respectively), visual network (ICs 20 and 29), somatosensory network (IC 24), auditory-insular network (ICs 9 and 31), cerebellar network (IC 22), and an “other” network (IC 33). Similar to the two components identified as part of the DMN in LMTLE (ICs 01 and 02), IC 17 is located in the superior and medial frontal regions of the DMN, and IC 01 is located in the posterior/retrosplenial region of DMN. As part of the visual network, IC 20 is primarily located within the cuneus, lingual gyrus, fusiform gyrus, and middle and inferior occipital lobes. As in LMTLE, GICA for RMTLE produced one visual network component, IC 29, that is attributed to the ventral stream of visual processing, or the “what” pathway, starting in V1 and ending in the inferior temporal cortex. Attributed to the somatosensory network, IC 24 is primarily located within the precentral gyrus, postcentral gyrus, and inferior parietal lobule. Two components were identified as part of the auditory-insular network, ICs 09 and 31. Similar to IC 08 of the same

network in LMTLE, ICs 09 and 31 are primarily located within the insula and primary and association auditory cortices. Although not located in the same regions as IC 33 of the group with LMTLE, IC 12 was also attributed to an “other” network. Specifically, this component is primarily located within the culmen, posterior cingulate, parahippocampal gyrus, lingual gyrus, and thalamus. Finally, IC 22 was attributed to the cerebellar network as it is primarily located in the cerebellar regions. 3.3. Granger causality analysis 3.3.1. Left medial temporal lobe epilepsy The components are grouped into four networks that are involved in visual memory encoding processes: default mode, visual, auditoryinsular, and attention. Significant causality relations were observed between six of the seven components (p b 0.05; corrected for FDR) of which the auditory-insular network was the single component not

Please cite this article as: Gregory AM, et al, The effect of medial temporal lobe epilepsy on visual memory encoding, Epilepsy Behav (2015), http://dx.doi.org/10.1016/j.yebeh.2015.03.006

6

A.M. Gregory et al. / Epilepsy & Behavior xxx (2015) xxx–xxx

Table 4 Cortical localizations of task-related independent components in patients with RMTLE (– indicates that this independent component was not observed on one side). Component ID

Area

Brodmann area

Peak Z-score and its spatial (x, y, z) L/R correlates

1

Precuneus Cuneus Cingulate gyrus Posterior cingulate Angular gyrus Inferior parietal lobule Superior temporal gyrus Transverse temporal gyrus Insula Precentral gyrus Middle temporal gyrus Inferior frontal gyrus Postcentral gyrus Claustrum Culmen Posterior cingulate Parahippocampal gyrus Lingual gyrus Thalamus Cuneus Culmen of vermis Precuneus Postcentral gyrus Superior frontal gyrus Medial frontal gyrus Middle frontal gyrus Cuneus Lingual gyrus Fusiform gyrus Middle occipital gyrus Inferior occipital gyrus Declive Middle frontal gyrus Culmen Superior frontal gyrus Uncus Declive of vermis Culmen Cerebellar lingual Culmen of vermis Declive Declive of vermis Lingual gyrus Parahippocampal gyrus Cuneus Pyramis Precentral gyrus Postcentral gyrus Inferior parietal lobule Paracentral lobule Superior temporal gyrus Medial frontal gyrus Middle temporal gyrus Uncus Parahippocampal gyrus Cingulate gyrus Superior parietal lobule Precuneus Superior frontal gyrus Middle frontal gyrus Culmen Declive Fusiform gyrus Parahippocampal gyrus Lingual gyrus Posterior cingulate Superior temporal gyrus Middle temporal gyrus Insula Middle frontal gyrus Claustrum Uncus Fusiform gyrus Parahippocampal gyrus Caudate

7, 23, 31 7, 18, 19 23, 24, 31 23, 29, 30, 31 39

6.3 (−2, −63, 29)/7.0 (2, −65, 29) 6.4 (−2, −66, 31)/6.9 (2, −68, 31) 6.2 (−2, −57, 27)/6.8 (2, −61, 27) 5.9 (−2, −55, 23)/6.3 (2, −55, 23) --/3.2 (48, −62, 36) --/3.0 (46, −62, 40) 4.8 (−53, −12, 2)/5.9 (42, −25, 9) 4.7 (−42, −23, 10)/5.5 (42, −29, 11) 4.9 (−44, −13, 8)/5.3 (40, −21, 8) 4.4 (−48, −9, 8)/4.9 (48, −13, 8) 3.3 (−55, −8, −3)/4.0 (55, −10, −3) 3.5 (−42, 17, −4)/3.7 (46, 17, −9) 3.5 (−50, −23, 14)/-2.7 (−36, −10, 4)/3.4 (36, −21, 5) 15.3 (0, −45, 2)/13.6 (4, −47, 2) 12.5 (−2, −44, 6)/13.5 (2, −44, 6) 8.9 (−8, −41, 2)/10.4 (8, −39, 2) 6.2 (−8, −56, 5)/5.3 (12, −49, 1) 4.4 (−8, −29, 3)/5.0 (8, −29, 5) 4.7 (−2, −64, 9)/3.6 (6, −64, 7) 4.4 (0, −60, 0)/3.8 (4, −60, 0) 2.9 (−4, −61, 62)/3.6 (4, −57, 62) --/2.7 (6, −49, 65) 9.6 (−10, 47, 42)/11.4 (8, 51, 40) 9.4 (−6, 49, 42)/11.2 (8, 47, 42) 3.9 (−26, 35, 46)/4.3 (32, 45, 36) 8.6 (−4, −93, 3)/8.5 (4, −95, 1) 8.8 (−4, −89, 3)/8.7 (4, −91, 1) 5.4 (−18, −86, −13)/4.1 (20, −88, −12) 5.1 (−8, −94, 14)/4.2 (8, −96, 14) 5.0 (−18, −92, −7)/5.1 (16, −90, −9) 4.6 (−20, −82, −16)/3.5 (24, −82, −16) 3.1 (−32, 59, 10)/-2.9 (−24, −51, −14)/2.7 (26, −49, −14) 2.9 (−30, 59, 14)/-2.7 (−26, 3, −22)/-2.7 (−2, −74, −10)/-12.1 (0, −49, −8)/10.9 (4, −51, −6) 9.8 (−4, −47, −11)/9.4 (4, −47, −11) 9.5 (0, −64, −5)/9.3 (4, −62, −5) 6.8 (−4, −70, −10)/6.8 (4, −70, −10) 6.8 (0, −72, −10)/5.5 (4, −69, −13) 6.3 (−8, −64, −2)/6.2 (4, −66, 0) 3.3 (−26, −48, 4)/-2.9 (−2, −62, 9)/-2.7 (−2, −77, −23)/-5.2 (−32, −24, 64)/-4.7 (−50, −19, 49)/-4.5 (−44, −38, 52)/-4.4 (0, −15, 45)/4.1 (2, −27, 46) 4.1 (−32, 10, −31)/3.0 (46, 14, −29) 3.8 (0, −15, 49)/3.3 (4, −15, 49) 2.6 (−42, 4, −29)/3.7 (51, 10, −29) 3.5 (−18, −5, −20)/-3.5 (−26, 1, −24)/-3.2 (0, −7, 45)/3.4 (2, −17, 41) 3.1 (−24, −46, 59)/-3.0 (0, −35, 44)/2.7 (4, −35, 44) 2.9 (−18, −12, 67)/-2.7 (−53, 2, 42)/-7.2 (−28, −49, −13)/7.7 (26, −47, −13) 7.1 (−28, −53, −12)/6.6 (30, −51, −13) 7.1 (−28, −45, −13)/7.1 (28, −47, −9) 5.4 (−24, −45, −10)/6.7 (24, −47, −9) 3.1 (−18, −47, −1)/4.3 (18, −49, 1) 3.0 (−12, −56, 10)/3.1 (12, −54, 6) 8.0 (−44, −1, −13)/5.7 (42, 3, −20) 7.5 (−48, −1, −13)/5.8 (50, −1, −18) 6.3 (−42, −14, −6)/3.5 (42, −18, −6) 3.6 (−34, 50, −13)/-3.6 (−38, −23, 1)/-3.4 (−30, 5, −20)/---/3.2 (46, −5, −23) 3.0 (−24, −50, 3)/---/3.0 (6, 4, 7)

9

12

17

20

22

24

29

31

13, 21, 22, 38, 41, 42 41, 42 13, 47 6, 13, 43, 44 21, 22 47 40, 43

23, 29, 30 27, 30, 35 18, 19 30 7 7 6, 8, 9, 10 6, 8, 9, 10 8, 9 17, 18, 19, 23 17, 18 18, 19 18 17, 18 10 10

18, 19 19, 30 30 4, 6 1, 2, 3, 5, 40 40 5, 6, 31 38 6 21 28 34 24, 31 7 7 6 6

19, 20, 37 19, 30, 35, 36, 37 18, 19 29, 30 13, 21, 22, 38 21 13, 22

28 20 19, 30

Please cite this article as: Gregory AM, et al, The effect of medial temporal lobe epilepsy on visual memory encoding, Epilepsy Behav (2015), http://dx.doi.org/10.1016/j.yebeh.2015.03.006

A.M. Gregory et al. / Epilepsy & Behavior xxx (2015) xxx–xxx

7

Table 4 (continued) Component ID

Area

Brodmann area

Peak Z-score and its spatial (x, y, z) L/R correlates

31

Inferior temporal gyrus Lingual gyrus Cerebellar tonsil

20, 21

2.9 (−53, −20, −16)/2.6 (53, −7, −20) 2.9 (−24, −54, 1)/-2.7 (−2, −54, −31)/--

Note: Table includes the anatomical location, the corresponding Brodmann area(s), and the maximum z-score with its Talairach coordinates for each IC.

identified as having significant causal relations (Fig. 1A). Fig. 1B depicts the memory encoding model for LMTLE created from these networks. 3.3.2. Right medial temporal lobe epilepsy Significant causality relations were observed between each of the nine components (p b 0.05; corrected for FDR) and are depicted in Fig. 2A. These components are grouped into six networks related to visual memory encoding: default mode, visual, somatosensory, auditory-insular, attention, and cerebellar. As in LMTLE, a model of memory encoding was created based on these networks, and the relative contributions of each of the specific networks in RMTLE are depicted in Fig. 2B. 4. Discussion The present study used group ICA and GCA to investigate differences in the networks involved in visual memory encoding between patients with LMTLE and patients with RMTLE. Based on the findings from the existing literature on the neuropsychological performance of those with MTLE, as well as fMRI data, we hypothesized that there would be differences in the visual memory encoding network between those with LMTLE and those with RMTLE. Exploratory analyses of the network revealed significant differences between the two groups. Thus, we performed group ICA on the group with LMTLE and the group with RMTLE separately and built models for visual memory encoding separately for each of the groups. Although task performance between the two groups was not statistically different, both patient groups performed significantly worse than healthy controls across all tasks, and the number of components correlated with the process of scene encoding varied between the groups with epilepsy (Figs. 1 and 2), indicating plasticity within the systems. In LMTLE, the nodes involved in the encoding process were attributed to four different networks and included the default mode, visual, auditory-insular, and “other” network. In RMTLE, the nodes involved in this process were attributed to six different networks and included the aforementioned networks as well as somatosensory and cerebellar networks. The existence of these differences was confirmed by exploratory statistical comparisons between three of the four ICs tested (Section 2.6.1.), demonstrating significant differences in the networks underlying visual memory encoding between those with LMTLE and those with RMTLE [61–65]. Thus, the GICA data are complementary to the previously obtained unidimensional GLM results for patients with TLE. In contrast to the previous hypothesis-driven GLM analyses comparing the activations associated with the control and active conditions convolved with a fixed shape hemodynamic response [6,34,35], the current GICA approach is data-driven and, therefore, has the ability to detect additional activated regions involved in the process of scene encoding when compared to standard GLM results. Additionally, in using the GICA approach, we were able to utilize both spatial information and temporal information provided by GICA to examine the specific components that contribute to visual scene encoding and to determine whether there were network differences between those with LMTLE and those with RMTLE. Further, to examine the specific contributions of the components involved in LMTLE and RMTLE, we used GCA to add directionality to the GICA data, allowing us to build a more comprehensive model of the networks involved in scene encoding for LMTLE and RMTLE. Group differences in the causal relations provided by GCA were observed. The relative contributions of the components to the visual memory network in MTLE are discussed below, followed by a

comparison with a model of visual memory encoding based on data obtained from a large sample of healthy control subjects [36]. 4.1. Default mode network Two ICs in both LMTLE (ICs 01 and 02) and RMTLE (ICs 17 and 01) were identified as task-positive (e.g., there is an increase in the BOLD signal while patients are performing the task), but, in this case, they were attributed to the default mode network (DMN) rather than the task itself (for more details, see [36]). This is because while this network was initially identified as more active during rest when compared to a specific cognitive task [66], evidence suggests that resting state networks are modulated by internal and external stimuli, decreasing and increasing in response to specific cues [67]. While the causal links were different in all three groups, the locations of these components are similar to the superior frontal component of the DMN found in healthy controls [36]. However, the involvement of the anterior cingulate cortex (ACC) is absent in both groups with MTLE. Previous studies have suggested that impairment in a single node may affect the function of an entire network [26], with such effect being consistent with the nociferous cortex hypothesis [5]. Several studies have demonstrated DMN abnormalities in TLE as assessed by resting state fMRI (for a review, see [22]). For example, one such study found that seizure-onset lateralization differentially affects the DMN with reduced connectivity between the anterior DMN and the posterior DMN in both LMTLE and RMTLE compared to healthy controls [61]. However, patients with LMTLE showed an increase in connectivity of the anterior DMN to fronto-centro-parietal areas and posterior DMN to bilateral opercular areas whereas patients with RMTLE showed reduced connectivity of the posterior DMN to adjacent areas and the right antero-medial temporal region. A similar study concluded that changes in hippocampal network connectivity may be the cause of increased activation of extrahippocampal regions whose function may be a compensatory mechanism for a TLE-related reduction in hippocampal connectivity elsewhere [68]. Another study found bilateral decreases in the mesial temporal lobes of patients with RMTLE to be related to duration of epilepsy, suggesting that seizure lateralization not only differentially affects network function [26] but also may contribute to network differences. From a network perspective, these studies highlight the notion that impairment in one node of a network typically causes widespread network abnormalities, which are reflected in the performance scores of the patients with epilepsy in the current study. 4.2. Visual network Similar visual network components were found among both groups with MTLE and healthy controls. Such activation patterns have been observed with the same and similar fMRI tasks [13,34] and are expected, given the nature of the scene-encoding task. Causal relations between visual network components and other networks were different across groups. Differences in visual perception may explain performance differences between groups with MTLE and healthy controls on the scrambled image (control) condition of the scene-encoding task; however, differences in visual perception between the groups with MTLE are unlikely as they performed equally well on the control condition. A study by Grant et al. (2008) supports this notion. As an indireect measure of occipital lobe function, they used two low-level visual tasks

Please cite this article as: Gregory AM, et al, The effect of medial temporal lobe epilepsy on visual memory encoding, Epilepsy Behav (2015), http://dx.doi.org/10.1016/j.yebeh.2015.03.006

8

A.M. Gregory et al. / Epilepsy & Behavior xxx (2015) xxx–xxx

A. Left MTLE Granger Causality Relations

B. Left MTLE Memory Encoding Model

Fig. 1. (A) Relations and directionality of the information flow between task-related independent components (ICs) for LMTLE. Details regarding each component are provided in Table 1. Each IC was attributed to a particular network (see the Discussion section). Arrows depict a significant (p b 0.05, corrected for FDR) causality relation between components. Component representations are in neurological convention (left hemisphere is on the left side of the image). (B) Proposed models for LMTLE visual memory encoding based on the results obtained in (A).

under easy and difficult conditions and found no significant differences between those with left and those with right TLE [69]. Therefore, the differences in causal relations found in the current study likely reflect plasticity within the visual network and demonstrate

the negative effect epilepsy has on extratemporal regions [5]. Further, the variance seen between the two groups with epilepsy highlights the differential effect lateralization of seizures has on network function.

Please cite this article as: Gregory AM, et al, The effect of medial temporal lobe epilepsy on visual memory encoding, Epilepsy Behav (2015), http://dx.doi.org/10.1016/j.yebeh.2015.03.006

A.M. Gregory et al. / Epilepsy & Behavior xxx (2015) xxx–xxx

9

A. Right MTLE Granger Causality Relations

B. Right MTLE Memory Encoding Model

Fig. 2. (A) Relations and directionality of the information flow between task-related independent components (ICs) for RMTLE. Details regarding each component are provided in Table 2. Each IC was attributed to a particular network (see the Discussion section). Arrows depict a significant (p b 0.05, corrected for FDR) causality relation between components. Component representations are in neurological convention (left hemisphere is on the left side of the image). (B) Proposed models for RMTLE visual memory encoding based on the results obtained in (A).

Please cite this article as: Gregory AM, et al, The effect of medial temporal lobe epilepsy on visual memory encoding, Epilepsy Behav (2015), http://dx.doi.org/10.1016/j.yebeh.2015.03.006

10

A.M. Gregory et al. / Epilepsy & Behavior xxx (2015) xxx–xxx

4.3. Auditory-insular network Part of the auditory-insular network, these components correspond to the auditory-insular component identified in healthy controls [36] and are primarily composed of insula and primary and association auditory cortices which correspond to an ICA-derived auditory network previously described [70]. The most notable difference is the involvement of two components in this network for RMTLE. The temporal divergence between the two components may suggest a need for additional recruitment to support the underlying processes. Functional reorganization in MTLE that is seizure onset-dependent may further explain the presence of the additional auditory-insular component in RMTLE. This additional component can be viewed in part as compensatory, as no significant differences in behavioral performances were found between the group with left MTLE and the group with right MTLE despite the presence of network differences. 4.4. Somatosensory network In this study, the somatosensory network was composed of one component and was observed only in patients with RMTLE. In the visual memory model based on results obtained with a healthy control group [36], neither this component nor any form of this network was determined to elicit more BOLD activity during the task compared with the control block. Interestingly, no causal links were received by this network in RMTLE. Anatomically, this component predominantly includes the premotor, primary motor, and primary and secondary somatosensory cortices. This component likely corresponds to the action–execution and perception somesthesis paradigms [70] and reflects the activation seen during manual motor tasks [71]. In fMRI task paradigms, activation of the left M1 area corresponding to right hand motor functioning has been observed during action verb generation [72,73]. As previously mentioned, the scene-encoding task involved two conditions: presence of indoor or outdoor scenes and two scrambled images that were the same or different. Participants were directed to indicate their choices during the two conditions using the same button box in the right hand. It is possible that in the group with RMTLE, internal rehearsal of the directions was required to maintain an accurate motor response to each of the conditions presented. This is consistent with one animal study that demonstrated activation in premotor regions preceding activation in the prefrontal cortex and striatum [74], and this activation was observed during abstract instructions such as “same” or “different”, as in the control condition of the present study's task paradigm. Therefore, it is possible that the presence of this network in RMTLE, but not in LMTLE, is a reflection of strategic differences between the two groups, and further studies may need to investigate the different pathways for memory endcoding and the supporting processes in patients with different locations of seizure onset. Further, the presence of this network in right, but not in left, MTLE group may reflect the compensatory recruitment of additional strategies to maintain response accuracy during the task, which is consistent with the previously observed increased BOLD signal in this area that was associated with improved semantic performance [75]. 4.5. Cerebellar network A single cerebellar component (IC 22) was identified in RMTLE, and it corresponds well to the cerebellar component observed in the same study of healthy controls [36]. Patients with cerebellar damage often experience language impairments [25], and recent studies have documented right cerebellar involvement in language and verbal memory processes [13,16,25]. This may partly explain why this network was not revealed by GICA in LMTLE and is likely the result of network dysfunction related to left medial temporal focal seizures that typically involve left dominant language areas. This is in agreement with the impaired verbal memory functioning seen in LMTLE [8] and the

reversed hemispheric pattern of associations in poststroke aphasia recovery [76]. 4.6. “Other network” Both IC 33 for LMTLE and IC 12 for RMTLE are located within a broad range of brain regions and have, thus, been termed in this study the “other” network. The feedback loop from this network to the visual network as well as the inclusion of the lingual, parahippocampal, and inferior frontal gyri indicate that these ICs may be involved in visual processing. Additionally, the “other” network receives causal influence from the anterior DMN and the auditory-insular network in RMTLE. It is, therefore, possible that this component reflects plasticity of visual-auditory processing and integration in MTLE and may in part explain the worse performance of visual memory functioning in both groups with epilepsy when compared to the healthy controls' task performance. As previously noted, visual attention mechanisms are needed to more accurately encode complex scenes [4], and studies have demonstrated that these mechanisms involve a wide range of brain regions in healthy controls, including the frontal and parietal lobes, that form feedback loops with the visual network [3,4]. While the same analyses conducted on healthy controls did not reveal ICs with a similar range in location [36], the healthy control model demonstrated an attention network that is subdivided into a frontoparietal network that is described in terms of the hemispheric encoding/retrieval asymmetry or ‘HERA’ model [77], which posits that the left and right frontal lobes have different involvement in the memory encoding and retrieval processes, explaining the temporal divergence of the frontoparietal ICs. Many studies have demonstrated that TLE causes wide range network disruptions [22,27,28,63,65,68]. Behavioral data also provide evidence of reduced attention in TLE [78], and it has been suggested that the effects on the attention network are the mechanism of cognitive impairments in MTLE [79]. Our results are consistent with these findings. However, significant compensatory mechanisms must account for this impairment in attentional control. It is, therefore, likely that the ICs within the “other” network represent the hypothesized compensatory mechanism of an impaired visual attention network that also integrates sensory information in MTLE. The presence of the parietal cortex as a node in this network further supports this notion, as it is well known as an integrative area of sensory processing [80,81]. 4.7. Network for visual memory encoding in temporal lobe epilepsy Using group ICA and GCA to investigate visual memory encoding in MTLE, we built a visual memory encoding model for the two groups, each representing the relative effects of LMTLE and RMTLE. Although no differences in behavioral performance were observed between groups with epilepsy, the effect of LMTLE on the network involved in visual encoding was different when compared to the effect of RMTLE. This is indicative of the presence of compensatory mechanisms involved in visual memory encoding for both MTLE groups, but particularly for those with LMTLE and may be related to language lateralization, as visual encoding involves not only episodic but also semantic memory processes [1,6]. When compared to the visual memory encoding model previously proposed in healthy controls [36], the representation of this process in each patient group was different, supporting our second hypothesis. A reduced number of nodes contributing to the scene-encoding network were observed when compared to controls, as well as less significant causal links between the nodes. These network differences could underlie the previously reported behavioral differences in memory encoding between patients with MTLE and healthy controls [6]. Considering epilepsy as a neural network disease, it is hypothesized that neuronal reorganization occurs as a result of epilepsy-related neuronal abnormalities [82], and evidence suggests that intrahemispheric

Please cite this article as: Gregory AM, et al, The effect of medial temporal lobe epilepsy on visual memory encoding, Epilepsy Behav (2015), http://dx.doi.org/10.1016/j.yebeh.2015.03.006

A.M. Gregory et al. / Epilepsy & Behavior xxx (2015) xxx–xxx

shifts in cognitive functions are not always efficient [6]. For instance, Bigras et al. investigated whether lateralization of seizure focus in TLE affected lateralization of memory encoding in the same sceneencoding task described here and correlated those results with neuropsychological testing. These authors found left lateralization of scene encoding in patients with RMTLE differed significantly from patients with LMTLE and healthy controls. Additionally, patients with LMTLE who exhibited lateralization of scene encoding to the right hemisphere performed worse on verbal memory measures, and left lateralization of scene encoding in RMTLE was associated with worse visual memory performance but improved verbal memory performance. They concluded that visual memory encoding is better supported by the hemisphere ipsilateral to the seizure focus than contralateral. While lateralization of the task was not directly assessed in the present study, these findings may explain the differences observed between patients with LMTLE and patients with RMTLE in the visual memory networks described here. 4.8. Limitations Network differences and performance deficits in the groups with epilepsy may in part be related to a medication effect as the healthy controls were not on AEDs and the patients with epilepsy were taking, on average, two AEDs at the time of their participation in the study. Additionally, many of our patients had frequent, uncontrolled seizures; therefore, we cannot determine whether the observed differences are due to seizures and the location of their onset or medication effects. While the effects of each of the available AEDs on the resting state networks (RSNs) have not been examined in detail, at least two studies showed that topiramate has profound effects on the RSNs [83,84]; the effects of other medications on the fMRI signals may be different [85]. Further studies are needed to address this question. While not directly assessed, our findings of increased BOLD activity in the somatosensory cortex of patients with RMTLE, but not those with LMTLE, during scene encoding suggest that strategic differences may exist between the two patient groups. Further work is needed to examine the effect of strategic differences on network function in MTLE. This study was also limited by the implementation of GICA to assess network differences as this test necessitates the use of PCA as a data reduction technique. In this process, important components could have been lost. 4.9. Conclusion The effects of LMTLE and RMTLE on memory encoding networks differ. Consistent with previous studies, the organization of memory encoding is dependent on laterality of seizure focus and may be mediated by functional reorganization in chronic epilepsy. These findings may explain the differences in memory abilities between patients with LMTLE and patients with RMTLE and highlight the modulating effects of epilepsy on the visual memory encoding network. Acknowledgments Christi Banks, CCRC and Kristina Bigras, MA helped with data collection. This study was presented in part at the American Epilepsy Society Meeting in Washington, DC, 12/2013. This study was supported in part by the UC Neuroscience Institute (JPS) and in part by R01 NS048281 (JPS). Disclosure The authors have no conflicts of interest in conjunction with this manuscript to report. The study was supported in part by The Neuroscience Institute in Cincinnati, OH, and in part by the Department of Neurology at the University of Alabama at Birmingham. This work was

11

performed in partial fulfillment of the master's thesis requirement by the first author. References [1] Jonides J, Lewis RL, Nee DE, Lustig CA, Berman MG, Moore KS. The mind and brain of short-term memory. Annu Rev Psychol 2008;59:193–224. [2] Pihlajamaki M, Tanila H, Hanninen T, Kononen M, Mikkonen M, Jalkanen V, et al. Encoding of novel picture pairs activates the perirhinal cortex: an fMRI study. Hippocampus 2003;13:67–80. [3] Nenert R, Viswanathan S, Dubuc DM, Visscher KM. Modulations of ongoing alpha oscillations predict successful short-term visual memory encoding. Front Hum Neurosci 2012;6:127. [4] Petersen SE, Posner MI. The attention system of the human brain: 20 years after. Annu Rev Neurosci 2012;35:73–89. [5] Hermann B, Seidenberg M. Executive system dysfunction in temporal lobe epilepsy: effects of nociferous cortex versus hippocampal pathology. J Clin Exp Neuropsychol 1995;17:809–19. [6] Bigras C, Shear PK, Vannest J, Allendorfer JB, Szaflarski JP. The effects of temporal lobe epilepsy on scene encoding. Epilepsy Behav 2013;26:11–21. [7] Karunanayaka P, Kim KK, Holland SK, Szaflarski JP. The effects of left or right hemispheric epilepsy on language networks investigated with semantic decision fMRI task and independent component analysis. Epilepsy Behav 2011;20:623–32. [8] Delaney RC, Rosen AJ, Mattson RH, Novelly RA. Memory function in focal epilepsy: a comparison of non-surgical, unilateral temporal lobe and frontal lobe samples. Cortex 1980;16:103–17. [9] Chelune GJ. Hippocampal adequacy versus functional reserve: predicting memory functions following temporal lobectomy. Arch Clin Neuropsychol 1995;10:413–32. [10] Quintas R, Raggi A, Giovannetti AM, Pagani M, Sabariego C, Cieza A, et al. Psychosocial difficulties in people with epilepsy: a systematic review of literature from 2005 until 2010. Epilepsy Behav 2012;25:60–7. [11] Tang V, Kwan P, Poon WS. Neurocognitive and psychological profiles of adult patients with epilepsy in Hong Kong. Epilepsy Behav 2013;29:337–43. [12] Sidhu MK, Stretton J, Winston GP, Bonelli S, Centeno M, Vollmar C, et al. A functional magnetic resonance imaging study mapping the episodic memory encoding network in temporal lobe epilepsy. Brain 2013;136:1868–88. [13] Detre JA, Maccotta L, King D, Alsop DC, Glosser G, D'Esposito M, et al. Functional MRI lateralization of memory in temporal lobe epilepsy. Neurology 1998;50:926–32. [14] Springer JA, Binder JR, Hammeke TA, Swanson SJ, Frost JA, Bellgowan PS, et al. Language dominance in neurologically normal and epilepsy subjects: a functional MRI study. Brain 1999;122(Pt 11):2033–46. [15] Alessio A, Damasceno BP, Camargo CH, Kobayashi E, Guerreiro CA, Cendes F. Differences in memory performance and other clinical characteristics in patients with mesial temporal lobe epilepsy with and without hippocampal atrophy. Epilepsy Behav 2004;5:22–7. [16] Black LC, Schefft BK, Howe SR, Szaflarski JP, Yeh HS, Privitera MD. The effect of seizures on working memory and executive functioning performance. Epilepsy Behav 2010;17:412–9. [17] Oyegbile TO, Dow C, Jones J, Bell B, Rutecki P, Sheth R, et al. The nature and course of neuropsychological morbidity in chronic temporal lobe epilepsy. Neurology 2004; 62:1736–42. [18] Golby AJ, Poldrack RA, Illes J, Chen D, Desmond JE, Gabrieli JD. Memory lateralization in medial temporal lobe epilepsy assessed by functional MRI. Epilepsia 2002;43:855–63. [19] Rabin ML, Narayan VM, Kimberg DY, Casasanto DJ, Glosser G, Tracy JI, et al. Functional MRI predicts post-surgical memory following temporal lobectomy. Brain 2004; 127:2286–98. [20] Bellgowan PS, Binder JR, Swanson SJ, Hammeke TA, Springer JA, Frost JA, et al. Side of seizure focus predicts left medial temporal lobe activation during verbal encoding. Neurology 1998;51:479–84. [21] Hamberger MJ, Cole J. Language organization and reorganization in epilepsy. Neuropsychol Rev 2011;21:240–51. [22] Cataldi M, Avoli M, de Villers-Sidani E. Resting state networks in temporal lobe epilepsy. Epilepsia 2013;54:2048–59. [23] Vlooswijk MC, Jansen JF, Jeukens CR, Majoie HJ, Hofman PA, de Krom MC, et al. Memory processes and prefrontal network dysfunction in cryptogenic epilepsy. Epilepsia 2011;52:1467–75. [24] Zhang Z, Lu G, Zhong Y, Tan Q, Yang Z, Liao W, et al. Impaired attention network in temporal lobe epilepsy: a resting FMRI study. Neurosci Lett 2009;458:97–101. [25] Zhang Z, Lu G, Zhong Y, Tan Q, Liao W, Chen Z, et al. Impaired perceptual networks in temporal lobe epilepsy revealed by resting fMRI. J Neurol 2009;256:1705–13. [26] Zhang Z, Lu G, Zhong Y, Tan Q, Liao W, Wang Z, et al. Altered spontaneous neuronal activity of the default-mode network in mesial temporal lobe epilepsy. Brain Res 2010;1323:152–60. [27] Ji GJ, Zhang Z, Zhang H, Wang J, Liu DQ, Zang YF, et al. Disrupted causal connectivity in mesial temporal lobe epilepsy. PLoS One 2013;8:e63183. [28] Desalvo MN, Douw L, Tanaka N, Reinsberger C, Stufflebeam SM. Altered structural connectome in temporal lobe epilepsy. Radiology 2013:131044. [29] Zheng J, Qin B, Dang C, Ye W, Chen Z, Yu L. Alertness network in patients with temporal lobe epilepsy: a fMRI study. Epilepsy Res 2012;100:67–73. [30] Yang Z, Wang H, Zhang Z, Zhong Y, Chen Z, Lu G. Study based on ICA of “dorsal attention network” in patients with temporal lobe epilepsy. Sheng Wu Yi Xue Gong Cheng Xue Za Zhi 2010;27:10–5. [31] Vanderhasselt MA, De Raedt R, Leyman L, Baeken C. Role of the left DLPFC in endogenous task preparation: experimental repetitive transcranial magnetic stimulation study. Neuropsychobiology 2010;61:162–8.

Please cite this article as: Gregory AM, et al, The effect of medial temporal lobe epilepsy on visual memory encoding, Epilepsy Behav (2015), http://dx.doi.org/10.1016/j.yebeh.2015.03.006

12

A.M. Gregory et al. / Epilepsy & Behavior xxx (2015) xxx–xxx

[32] Liao W, Zhang Z, Pan Z, Mantini D, Ding J, Duan X, et al. Altered functional connectivity and small-world in mesial temporal lobe epilepsy. PLoS One 2010;5:e8525. [33] Martin RC, Sawrie SM, Gilliam FG, Palmer CA, Faught E, Morawetz RB, et al. Wisconsin Card Sorting performance in patients with temporal lobe epilepsy: clinical and neuroanatomical correlates. Epilepsia 2000;41:1626–32. [34] Szaflarski JP, Holland SK, Schmithorst VJ, Dunn RS, Privitera MD. High-resolution functional MRI at 3 T in healthy and epilepsy subjects: hippocampal activation with picture encoding task. Epilepsy Behav 2004;5:244–52. [35] Vannest J, Szaflarski JP, Privitera MD, Schefft BK, Holland SK. Medial temporal fMRI activation reflects memory lateralization and memory performance in patients with epilepsy. Epilepsy Behav 2008;12:410–8. [36] Nenert R, Allendorfer JB, Szaflarski JP. A model for visual memory encoding. PLoS One 2014;9:e107761. [37] Macwhinney B, Cohen J, Provost J. The PsyScope experiment-building system. Spat Vis 1997;11:99–101. [38] Binder JR, Bellgowan PS, Hammeke TA, Possing ET, Frost JA. A comparison of two FMRI protocols for eliciting hippocampal activation. Epilepsia 2005;46:1061–70. [39] Mechanic-Hamilton D, Korczykowski M, Yushkevich PA, Lawler K, Pluta J, Glynn S, et al. Hippocampal volumetry and functional MRI of memory in temporal lobe epilepsy. Epilepsy Behav 2009;16:128–38. [40] Calhoun VD, Adali T, Pearlson GD, Pekar JJ. A method for making group inferences from functional MRI data using independent component analysis. Hum Brain Mapp 2001;14:140–51. [41] Federico P, Archer JS, Abbott DF, Jackson GD. Cortical/subcortical BOLD changes associated with epileptic discharges: an EEG–fMRI study at 3 T. Neurology 2005;64: 1125–30. [42] Salek-Haddadi A, Diehl B, Hamandi K, Merschhemke M, Liston A, Friston K, et al. Hemodynamic correlates of epileptiform discharges: an EEG–fMRI study of 63 patients with focal epilepsy. Brain Res 2006;1088:148–66. [43] Kobayashi E, Bagshaw AP, Grova C, Dubeau F, Gotman J. Negative BOLD responses to epileptic spikes. Hum Brain Mapp 2006;27:488–97. [44] Jacobs J, Hawco C, Kobayashi E, Boor R, LeVan P, Stephani U, et al. Variability of the hemodynamic response as a function of age and frequency of epileptic discharge in children with epilepsy. Neuroimage 2008;40:601–14. [45] Bell AJ, Sejnowski TJ. An information-maximization approach to blind separation and blind deconvolution. Neural Comput 1995;7:1129–59. [46] Demirci O, Stevens MC, Andreasen NC, Michael A, Liu J, White T, et al. Investigation of relationships between fMRI brain networks in the spectral domain using ICA and Granger causality reveals distinct differences between schizophrenia patients and healthy controls. Neuroimage 2009;46:419–31. [47] Ding X, Lee SW. Changes of functional and effective connectivity in smoking replenishment on deprived heavy smokers: a resting-state FMRI study. PLoS One 2013;8:e59331. [48] Jiao Q, Lu G, Zhang Z, Zhong Y, Wang Z, Guo Y, et al. Granger causal influence predicts BOLD activity levels in the default mode network. Hum Brain Mapp 2011;32: 154–61. [49] Sridharan D, Levitin DJ, Menon V. A critical role for the right fronto-insular cortex in switching between central-executive and default-mode networks. Proc Natl Acad Sci U S A 2008;105:12569–74. [50] Stevens MC, Pearlson GD, Calhoun VD. Changes in the interaction of resting-state neural networks from adolescence to adulthood. Hum Brain Mapp 2009;30:2356–66. [51] Li YO, Adali T, Calhoun VD. Estimating the number of independent components for functional magnetic resonance imaging data. Hum Brain Mapp 2007;28:1251–66. [52] Himberg J, Hyvarinen A, Esposito F. Validating the independent components of neuroimaging time series via clustering and visualization. Neuroimage 2004;22: 1214–22. [53] Erhardt EB, Rachakonda S, Bedrick EJ, Allen EA, Adali T, Calhoun VD. Comparison of multi-subject ICA methods for analysis of fMRI data. Hum Brain Mapp 2011;32: 2075–95. [54] Granger CWJ. Investigating causal relations by econometric models and crossspectral methods. Econometrica (pre-1986) 1969;37:424. [55] Wen X, Rangarajan G, Ding M. Is Granger causality a viable technique for analyzing fMRI data? PLoS One 2013;8:e67428. [56] Seth AK, Chorley P, Barnett LC. Granger causality analysis of fMRI BOLD signals is invariant to hemodynamic convolution but not downsampling. Neuroimage 2013;65: 540–55. [57] Szaflarski JP, DiFrancesco M, Hirschauer T, Banks C, Privitera MD, Gotman J, et al. Cortical and subcortical contributions to absence seizure onset examined with EEG/fMRI. Epilepsy Behav 2010;18:404–13.

[58] Seth AK. A MATLAB toolbox for Granger causal connectivity analysis. J Neurosci Methods 2010;186:262–73. [59] Seth AK. Causal connectivity of evolved neural networks during behavior. Network 2005;16:35–54. [60] Gideon S. Estimating the dimension of a model. Ann Stat 1978;6:461–4. [61] Haneef Z, Lenartowicz A, Yeh HJ, Engel Jr J, Stern JM. Effect of lateralized temporal lobe epilepsy on the default mode network. Epilepsy Behav 2012;25:350–7. [62] Pereira FR, Alessio A, Sercheli MS, Pedro T, Bilevicius E, Rondina JM, et al. Asymmetrical hippocampal connectivity in mesial temporal lobe epilepsy: evidence from resting state fMRI. BMC Neurosci 2010;11:66. [63] Pittau F, Grova C, Moeller F, Dubeau F, Gotman J. Patterns of altered functional connectivity in mesial temporal lobe epilepsy. Epilepsia 2012;53:1013–23. [64] James GA, Tripathi SP, Ojemann JG, Gross RE, Drane DL. Diminished default mode network recruitment of the hippocampus and parahippocampus in temporal lobe epilepsy. J Neurosurg 2013;119:288–300. [65] Maccotta L, He BJ, Snyder AZ, Eisenman LN, Benzinger TL, Ances BM, et al. Impaired and facilitated functional networks in temporal lobe epilepsy. Neuroimage Clin 2013;2:862–72. [66] Raichle ME, MacLeod AM, Snyder AZ, Powers WJ, Gusnard DA, Shulman GL. A default mode of brain function. Proc Natl Acad Sci U S A 2001;98:676–82. [67] Fox MD, Snyder AZ, Vincent JL, Corbetta M, Van Essen DC, Raichle ME. The human brain is intrinsically organized into dynamic, anticorrelated functional networks. Proc Natl Acad Sci U S A 2005;102:9673–8. [68] Haneef Z, Lenartowicz A, Yeh HJ, Levin HS, Engel Jr J, Stern JM. Functional connectivity of hippocampal networks in temporal lobe epilepsy. Epilepsia 2014;55:137–45. [69] Grant AC, Donnelly KM, Chubb C, Barr WB, Kuzniecky R, Devinsky O. Temporal lobe epilepsy does not impair visual perception. Epilepsia 2008;49:710–3. [70] Smith SM, Fox PT, Miller KL, Glahn DC, Fox PM, Mackay CE, et al. Correspondence of the brain's functional architecture during activation and rest. Proc Natl Acad Sci U S A 2009;106:13040–5. [71] Biswal B, Yetkin FZ, Haughton VM, Hyde JS. Functional connectivity in the motor cortex of resting human brain using echo-planar MRI. Magn Reson Med 1995;34: 537–41. [72] Kana RK, Blum ER, Ladden SL, Ver Hoef LW. “How to do things with words”: role of motor cortex in semantic representation of action words. Neuropsychologia 2012; 50:3403–9. [73] Repetto C, Colombo B, Cipresso P, Riva G. The effects of rTMS over the primary motor cortex: the link between action and language. Neuropsychologia 2013;51:8–13. [74] Muhammad R, Wallis JD, Miller EK. A comparison of abstract rules in the prefrontal cortex, premotor cortex, inferior temporal cortex, and striatum. J Cogn Neurosci 2006;18:974–89. [75] Donnelly KM, Allendorfer JB, Szaflarski JP. Right hemispheric participation in semantic decision improves performance. Brain Res 2011;1419:105–16. [76] Szaflarski JP, Allendorfer JB, Banks C, Vannest J, Holland SK. Recovered vs. notrecovered from post-stroke aphasia: the contributions from the dominant and non-dominant hemispheres. Restor Neurol Neurosci 2013;31:347–60. [77] Tulving E, Kapur S, Craik FI, Moscovitch M, Houle S. Hemispheric encoding/retrieval asymmetry in episodic memory: positron emission tomography findings. Proc Natl Acad Sci U S A 1994;91:2016–20. [78] Takaya S, Hanakawa T, Hashikawa K, Ikeda A, Sawamoto N, Nagamine T, et al. Prefrontal hypofunction in patients with intractable mesial temporal lobe epilepsy. Neurology 2006;67:1674–6. [79] Bocquillon P, Dujardin K, Betrouni N, Phalempin V, Houdayer E, Bourriez JL, et al. Attention impairment in temporal lobe epilepsy: a neurophysiological approach via analysis of the P300 wave. Hum Brain Mapp 2009;30:2267–77. [80] Brang D, Taich ZJ, Hillyard SA, Grabowecky M, Ramachandran VS. Parietal connectivity mediates multisensory facilitation. Neuroimage 2013;78:396–401. [81] Wolpert DM, Goodbody SJ, Husain M. Maintaining internal representations: the role of the human superior parietal lobe. Nat Neurosci 1998;1:529–33. [82] Powell GE, Polkey CE, McMillan T. The new Maudsley series of temporal lobectomy. I: short-term cognitive effects. Br J Clin Psychol 1985;24(Pt 2):109–24. [83] Szaflarski JP, Allendorfer JB. Topiramate and its effect on fMRI of language in patients with right or left temporal lobe epilepsy. Epilepsy Behav 2012;24:74–80. [84] Yasuda CL, Centeno M, Vollmar C, Stretton J, Symms M, Cendes F, et al. The effect of topiramate on cognitive fMRI. Epilepsy Res 2013;105:250–5. [85] Wandschneider B, Stretton J, Sidhu M, Centeno M, Kozak LR, Symms M, et al. Levetiracetam reduces abnormal network activations in temporal lobe epilepsy. Neurology 2014;83:1508–12.

Please cite this article as: Gregory AM, et al, The effect of medial temporal lobe epilepsy on visual memory encoding, Epilepsy Behav (2015), http://dx.doi.org/10.1016/j.yebeh.2015.03.006

The effect of medial temporal lobe epilepsy on visual memory encoding.

Effective visual memory encoding, a function important for everyday functioning, relies on episodic and semantic memory processes. In patients with me...
2MB Sizes 2 Downloads 15 Views