Epilepsy & Behavior 42 (2015) 78–85

Contents lists available at ScienceDirect

Epilepsy & Behavior

journal homepage: www.elsevier.com/locate/yebeh

Incipient preoperative reorganization processes of verbal memory

functions in patients with left temporal lobe epilepsy
Monika Milian a, Lena Zeltner b, Michael Erb c, Uwe Klose d, Kathrin Wagner e, Lars Frings f, Cornelia Veil c,
Sabine Rona a, Holger Lerche g, Silke Klamer g,⁎
a
Department of Neurosurgery, University of Tuebingen, Tuebingen, Germany
b
Department of Vascular Neurology, University of Tuebingen, Tuebingen, Germany
c
Department of Biomedical Magnetic Resonance, University of Tuebingen, Tuebingen, Germany
d
Department of Neuroradiology, University of Tuebingen, Tuebingen, Germany
e
Epilepsy Center, University Hospital Freiburg, Germany
f
Center of Geriatrics and Gerontology Freiburg, University Hospital Freiburg, Germany
g
Department of Neurology and Epileptology, Hertie Institute for Clinical Brain Research, University of Tuebingen, Tuebingen, Germany

a r t i c l e i n f o a b s t r a c t

Article history: We previously reported nonlinear correlations between verbal episodic memory performance and BOLD signal in
Received 14 October 2014 memory fMRI in healthy subjects. The purpose of the present study was to examine this observation in patients
Accepted 21 November 2014 with left mesial temporal lobe epilepsy (mTLE) who often experience memory decline and need reliable predic-
Available online xxxx
tion tools before epilepsy surgery with hippocampectomy. Fifteen patients with left mTLE (18–57 years, nine fe-
males) underwent a verbal memory fMRI paradigm. Correlations between BOLD activity and neuropsychological
Keywords:
Memory fMRI
data were calculated for the i) hippocampus (HC) as well as ii) extrahippocampal mTL structures. Memory per-
Hippocampus formance was systematically associated with activations within the right HC as well as with activations within
Chronic mesial temporal lobe epilepsy the left extrahippocampal mTL regions (amygdala and parahippocampal gyrus). As hypothesized, the analyses
Preoperative reorganization revealed cubic relationships, with one peak in patients with marginal memory performance and another peak
Nonlinear correlations in patients with very good performance. The nonlinear correlations between memory performance and activa-
tions might reflect the compensatory recruitment of neural resources to maintain memory performance in pa-
tients with ongoing memory deterioration. The present data suggest an already incipient preoperative
reorganization process of verbal memory in non-amnesic patients with left mTLE by simultaneously tapping
the resources of the right HC and left extrahippocampal mTL regions. Thus, in the preoperative assessment,
both neuropsychological performance and memory fMRI should be considered together.
© 2014 Elsevier Inc. All rights reserved.

1. Introduction
Surgery within the mesial temporal lobe (mTL) bears the risk of rel-
evant episodic memory decline, typically of verbal memory following
left and of nonverbal memory following right anterior temporal lobe re-
section (ATLR) [1,2]. The prediction of potential postoperative memory
decline is, therefore, of great importance in the clinical setting and the
goal of various memory functional MRI (fMRI) studies [3–7]. There is
agreement that patients with epilepsy with good memory abilities
prior to surgery [2,8,9] and patients whose memory functions are
lateralized to the side of surgery [10] are more likely to have memory
decline.
⁎ Corresponding author at: Department of Neurology and Epileptology, Hertie Institute
for Clinical Brain Research, University of Tuebingen, Hoppe-Seyler-Strasse 3, 72076
Tuebingen, Germany. Tel.: +49 7071 29 87707; fax: +49 7071 29 4488.
E-mail address: silke.klamer@uni-tuebingen.de (S. Klamer).
Several fMRI paradigms have been developed with regard to the
prediction of postoperative memory outcome in operated patients
with mesial temporal lobe epilepsy (mTLE) [3,6,7,11]. To date, only lin-
ear correlations between preoperative fMRI activations and preopera-
tive memory performance [3] or postoperative memory changes [11,
12] were taken into account. In a recently published work, we investi-
gated the nature of correlations between memory performance levels
and fMRI activations within the mTL in healthy subjects [13]. Instead
of linear correlations, we found u-shaped correlations between subjects'
verbal memory performance and mTL activations. The observed hyper-
activation among subjects with marginal memory performance might
reflect a compensatory recruitment of neural resources to maintain
memory performance, as previously reported for patients with mild
cognitive impairment and Alzheimer's disease [14–16].
The purpose of the present study was to investigate the presence of
nonlinear correlations in patients with left mTLE. We used a verbal
memory paradigm, first published by Wagner and colleagues [17],
that has been shown to produce strong left lateralized activation

http://dx.doi.org/10.1016/j.yebeh.2014.11.026
1525-5050/© 2014 Elsevier Inc. All rights reserved.
 M. Milian et al. / Epilepsy & Behavior 42 (2015) 78–85
patterns in healthy subjects [17,18]. We hypothesized there would be a
nonlinear correlation between neuropsychological test scores and fMRI
activations in mesial temporal structures due to compensatory hyperac-
tivation in patients with marginal memory performance.
2. Materials and methods
2.1. Subjects
We examined 15 German-speaking patients with chronic left-sided
mTLE with clear mesiotemporal spikes on EEG and typical temporal
lobe seizure semiology before left ATLR, 12 of them with clear signs of
hippocampal sclerosis (HS) on MRI and 1 with mild and 2 without
clear signs of HS. None of the patients had any extrahippocampal le-
sions. All patients were right-handed (Mhandedness quotient = 0.95, SD =
0.08; [19]) with normal or corrected-to-normal vision (9 females, 6
males, Mage = 36.3 years, SD = 11.7 years, range = 18–57 years,
Meducation = 12.3 years, SD = 2.9 years) and showed left-sided
language dominance as confirmed by fMRI.
The study was approved by the ethics committee of the University
of Tuebingen and is in accordance with the Declaration of Helsinki. All
patients gave written informed consent.
2.2. Neuropsychological tests
As memory performance level decreases with age [20,21], the use of
raw scores in memory assessments appears to be, in our view, not en-
tirely adequate when investigating a sample with a large variance of
participants' age. For our analyses, we therefore considered the stan-
dardized memory performance compared to an age-matched reference
population in the form of percentile ranks as determined by the manual
instead of absolute values (i.e., raw scores).
2.2.1. Verbal memory test
A verbal memory test was performed outside the scanner in which
subjects had to memorize a list of 15 words (Verbaler Lern- und
Merkfähigkeitstest, VLMT, [22]). We assessed three verbal memory
scores: (i) ‘immediate recall’, i.e., the sum of words correctly reproduced
during five learning trials (max. of 75); (ii) ‘delayed recall’, i.e., the num-
ber of correctly remembered words after a 30-minute delay (max. of
15); and (iii) ‘delayed recognition’ using a recognition condition after
the delayed recall condition (max. of 15).
2.2.2. Nonverbal memory test
To assess the functionality of the right mesial temporal lobe regions,
a nonverbal learning and memory performance test, the DCS
(Diagnostikum für Cerebralschädigung [23]), was applied where sub-
jects had to learn 9 geometrical figures. The ‘immediate recall’ score,
i.e., the sum of correctly reproduced figures during five learning trials
(max. of 45), was assessed. Scores were correlated with the VLMT to as-
sess the ability of the nondominant mTL to maintain verbal memory
functions.
2.2.3. Intelligence level
The level of verbal crystallized intelligence in each patient was
assessed using the German multiple-choice vocabulary test to exclude
patients with abnormal neuropsychological performance due to mental
retardation (MWT-B, Mehrfachwahl-Wortschatz-Intelligenztest [24]).
2.3. Magnetic resonance data acquisition
Magnetic resonance imaging (MRI) studies were performed on a
Siemens Magnetom Sonata [Maestro Class] 1.5 T scanner (Siemens
AG, Erlangen, Germany). All data were acquired using an 8-channel
array head coil for reception and the body coil for transmission. In
order to obtain a high-resolution anatomical image of each subject's
79
brain, a sagittal T1-weighted 3D-MPRAGE sequence was used (TR/TI/
TE = 1300/660/3.19 ms, flip angle = 15°, field of view =
256 ∗ 256 mm2, matrix = 256 ∗ 256, 176 slices, voxel size =
1 ∗ 1 ∗ 1 mm3). Additionally, a field map was recorded for later correc-
tion of distortions in the functional images caused by magnetic field in-
homogeneity. For the fMRI tasks, gradient-echo planar T2*-weighted
images covering the whole brain were acquired (TR = 4000 ms,
TE = 64 ms, field of view = 192 ∗ 192 mm2, matrix = 64 ∗ 64, voxel
size = 3 ∗ 3 ∗ 3 mm3, gap = 0.3 mm, 38 interleaved slices). The task
was performed in block design and consisted of 175 acquisitions. The
first two images were discarded in order to reach equilibrium of
magnetization.
The stimuli were visually projected on a translucent screen posi-
tioned at the end of the scanner table using a video projector outside
the magnet room. Subjects saw the presentation via a mirror attached
to the head coil. Outside the scanner room, a Windows laptop using
the software ‘Presentation 0.6’ (http://www.neurobehaviouralsystems.
com) was connected to the video projector. Responses were recorded
by use of a fiber optic button box where patients had to press the correct
one of two buttons with the thumb of the right hand.
2.4. Stimuli and fMRI task design
To investigate verbal memory, patients were presented with 24
word pairs and instructed to memorize and recognize them as de-
scribed previously [17] (Fig. 1). During the blocks of the encoding con-
dition, patients were shown four neither semantically nor
phonematically related word pairs for 7 s (plus 1 s of black screen)
and were asked to memorize them, e.g., “Ananas + Schraube” (“pineap-
ple + screw”). This was alternated with a block of the control condition
in which the names of two weekdays were presented [e.g., “Montag +
Mittwoch” (“Monday + Wednesday”)] for 5 s (plus 1 s of black screen),
and patients had to indicate by button press whether they were identi-
cal or not. During the recognition condition, patients performed a two-
alternative forced-choice test in which one word was presented above
two alternatives (the correct associate and one distractor) for 7 s (plus
1 s of black screen) and patients were asked to select by button press
the memorized associate. Both alternatives had been seen during the
encoding phase, so patients were not able to distinguish between
them on the basis of familiarity alone. Moreover, items were presented
in randomized order, alternating with a block of the control condition.
The fMRI behavioral results obtained in the recognition condition
were not considered as a measure of performance, and the button
presses during the recognition task were used mainly to ensure
cooperation.
2.5. Image processing and fMRI data analysis
Imaging data were analyzed in MATLAB (http://www.mathworks.
com) using Statistical Parametric Mapping (SPM 8) (Wellcome Trust
Centre for Imaging Neuroscience; http://www.fil.ion.ucl.ac.uk/spm).
Functional images were converted into NIFTI-1 format. The imaging
time series of each patient underwent a slice-timing procedure, was
realigned and unwarped based on the estimated field map data, and
then was coregistered to the anatomical reference image and normal-
ized to the MNI (Montreal Neurologic Institute Atlas, MNI) space [25].
The normalized data were smoothed with an isotropic Gaussian kernel
(8-mm full width at half maximum) and temporally filtered with a
high-pass filter with a cutoff time of 128 s.
2.6. Statistics
2.6.1. fMRI single-level statistics
For single-subject analyses, experimental task and control blocks
were convolved with the hemodynamic response function (hrf) in
order to evaluate individual main effects for the (i) encoding vs. control
80 M. Milian et al. / Epilepsy & Behavior 42 (2015) 78–85
Fig. 1. The verbal memory paradigm: one of two experimental cycles showing examples for encoding, recognition, and control conditions. Examples: “Ananas + Schraube” (“pineapple +
screw”), “Montag” and “Mittwoch” (“Monday” and “Wednesday”), and “Melone” (“melon”).
condition and (ii) recognition vs. control condition. To investigate
activations within the mesial temporal lobe (mTL), four regions of inter-
est were defined using an anatomic atlas: the right hippocampus and
left hippocampus (HC) and right and left extrahippocampal mTL
(mTL–HC, consisting of the parahippocampal gyrus and amygdala) re-
gions [26]. The abbreviations HCR and mTLR–HCR stand for the right
and HCL and mTLL–HCL for the left ROIs. The masks were applied to
the normalized images of each patient. We report activations within
these ROIs at a threshold of p b 0.05 (small volume corrected) with an
extent threshold of 5 voxels.
2.6.2. fMRI group-level statistics
In order to test for task-related whole brain activations across pa-
tients, second-level analyses were performed. We tested for the main
effects of encoding and recognition on the word paradigm using
one-sample t-tests. Results are reported at a threshold of p b 0.05
(uncorrected) also with an extent threshold of 5 voxels. A lower thresh-
old or a correction would have led to a lack of activity within the ROIs. As
in the single-subject analyses, the same four ROIs were used.
2.6.3. Behavioral data and correlation analyses
Data were analyzed using SPSS (Statistical Package for Social
Sciences) version 21 for Windows. Descriptive statistics were used to
analyze sociodemographic and neuropsychological characteristics
using minimum, maximum, mean, and standard deviations (SDs) for
parametric data. The percentage of correct answers was calculated for
the fMRI control task and the recognition task. To test for normal distri-
bution, we used Kolmogorov–Smirnov tests. When data showed
non-normal distribution, comparisons between two related groups
were performed with the Wilcoxon signed-rank test. Linear correlations
between nonparametric data are expressed as Spearman's rank correla-
tion coefficients.
To avoid a high number of comparisons, we used a hypothesis-
driven approach by testing for nonlinear (i.e., quadratic and cubic)
correlations between the percentile ranks of the VLMT and the sum
of the t-values of suprathreshold voxels within the ROIs based on our
previous study [13]. The significance level was set at p b 0.05.
2.6.4. Measure of BOLD activation
We chose the sum of the t-values of suprathreshold voxels as the
activation parameter to integrate not only the extension but also the
intensity of the activation in the analysis. This is based on the method
described in the study by Cheung et al. [27] that correlated the amount
of total activated voxels in the mTL with the subjects' neuropsycholog-
ical memory performance. To optimize this quantification method, we
decided to also consider the intensity of activation, i.e., the t-values,
with the aim of obtaining the spatial dimension weighted by the inten-
sity. It did not seem to be adequate to take only the peak t-value in a
region of interest because there could also be a higher peak t-value in
a cluster of, for example, 5 voxels, that in itself does not necessarily
represent a valid activation.
3. Results
3.1. Neuropsychological memory performance
The patients' neuropsychological data are presented in Table 1. Ac-
cording to their percentile ranks in the VLMT immediate recall condi-
tion, they were assigned to three groups: (i) 6 patients showed severe
impairments (percentile rank ≤ 5); (ii) 3 patients showed marginal ver-
bal immediate recall scores (percentile rank around 16, i.e., between 10
and 20); and (iii) 6 patients showed normal immediate recall scores
(percentile rank N 20). A percentile rank of 16 in the Gaussian normal
distribution represents exactly one standard deviation below the
mean and is, therefore, the critical threshold for the classification of
average or below average. Thus, a result around the percentile rank of
16 indicates a marginal performance.
Patients' mean IQ according to the MWT-B was 96.9 ± 11.8
(mean ± SD). The Kolmogorov–Smirnov tests revealed non-normal dis-
tributions for the VLMT immediate and delayed recall conditions as well
as the DCS immediate recall condition (each p b 0.05) and normal distri-
bution for the VLMT delayed recognition task (p = 0.119). There was a
significant correlation between the figural learning in the DCS and the
delayed recognition parameter of the VLMT (delayed recognition: r =
0.523, p b 0.050) and a clear trend towards a significant correlation be-
tween the DCS and the VLMT immediate recall (r = 0.485, p = 0.067)
and delayed recall (r = 0.480, p = 0.070) conditions.
3.2. fMRI behavioral data
Patients' fMRI behavioral data are presented in Table 1. The
Kolmogorov–Smirnov tests revealed non-normal distributions for the
 M. Milian et al. / Epilepsy & Behavior 42 (2015) 78–85 81

Table 1
Demographic, neuropsychological, and fMRI data of the subjects (N = 15).

No. Age/sex Neuropsychology fMRI behavioral data fMRI encoding activation fMRI recognition activation

VLMT PR DCS PR fMRI recognition fMRI control Left HC Left mTL–HC Right HC Right mTL–HC Left HC Left mTL–HC Right HC Right mTL-HC

1 47/M 0 5 33.3 83.0 0.0 0.0 0.0 10.6 0.0 10.1 0.0 51.2
2 49/F 0 12 47.6 72.7 0.0 0.0 0.0 0.0 31.7 31.7 0.0 0.0
3 25/F 0 5 75.0 89.6 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
4 21/M 5 5 79.2 100.0 0.0 0.0 0.0 0.0 27.1 3.0 22.7 0.0
5 46/M 0 30 54.2 85.1 203.4 169.1 156.2 299.4 0.0 13.6 0.0 0.0
6 28/F 0 14 75.0 95.8 116.6 76.3 252.7 260.4 0.0 0.0 29.6 10.6
7 36/M 10 75 83.3 97.9 80.3 94.9 72.9 152.2 9.8 18.0 13.4 108.6
8 57/F 20 47 60.0 65.0 0.0 0.0 33.6 39.0 86.10 29.9 0.0 0.0
9 46/M 20 70 87.5 100.0 0.0 0.0 0.0 0.0 8.0 58.0 8.0 40.0
10 36/F 30 5 75.0 100.0 0.0 0.0 0.0 0.0 0.0 28.0 15.1 59.3
11 30/F 70 92 83.3 87.2 69.0 131.6 91.3 126.5 0.0 0.0 0.0 0.0
12 18/F 85 85 100.0 100.0 147.6 146.7 94.4 110.5 33.8 17.2 82.9 1.9
13a 25/M 70 68 100.0 100.0 33.8 0.7 0.0 0.0 0.0 61.7 0.0 38.6
14a 35/F 85 9 91.7 97.6 125.4 4.5 37.7 4.1 15.9 90.2 22.9 16.3
15a 46/F 70 35 87.5 100.0 10.0 23.5 35.3 78.1 0.0 20.0 0.0 0.0

VLMT: Verbaler Lern- und Merkfähigkeitstest (wordlist learning and memory test); DCS: Diagnostikum für Cerebralschädigung (figure learning and memory test); PR: percentile ranks;
SD: standard deviation; fMRI activation: sum of the t-values of suprathreshold voxels; HC: hippocampus; mTL–HC: mesial temporal lobe–hippocampus.
a
Patients with mild or without clear signs of hippocampal sclerosis.

control condition and the recognition condition (each p b 0.05). Percent
correct recognition performance in the fMRI tasks showed 75.5 ± 19.2
correctly recognized word pairs in the word recognition condition
(Table 1). In the control condition, the mean percentage of correct
responses was 91.6 ± 11.1, demonstrating the good feasibility of the
task and compliance of the patients. Performance was significantly bet-
ter in the control compared to the recognition condition (Z = −3.183,
p b 0.01). There was a significant correlation between the fMRI recogni-
tion behavioral data and all parameters of the VLMT (immediate recall:
r = 0.828, p b 0.001; delayed recall: r = 0.805, p b 0.001; and delayed
recognition: r = 0.831, p b 0.001).
3.3. fMRI data
The sums of the suprathreshold t-values for each ROI and condition
are summarized in Table 1. Activations in the right extrahippocampal
mesial temporal regions, i.e., mTL–HC, were stronger than activations
within the right hippocampal mask (Z = − 2.073, p b 0.05) for the
word-encoding condition. For the other calculations, extrahippocampal
mTL activations were similarly strong as HC activations (each p N 0.05).
3.4. Correlation analyses
Nonlinear correlation analyses were performed between fMRI rec-
ognition data (similar to [13]) and the percentile ranks of the VLMT pa-
rameters. Only the cubic correlation analyses produced statistically
significant results. Correlations between the behavioral fMRI data and
fMRI activations were not calculated, since for this parameter, no age-
adjusted normative values are available.
3.4.1. Hippocampal (HC) ROI
Significant cubic correlations between fMRI activations and neuro-
psychological memory scores (VLMT immediate recall: R2 = 0.568,
p b 0.05; VLMT delayed recall: R2 = 0.577, p b 0.05; VLMT delayed rec-
ognition: R2 = 0.828, p b 0.0001) were only found within the right

3.3.1. Group analyses

In the group analysis for the word-encoding condition, we found sig-
nificant activation within the right HC mask (Fig. 2). The peak activation
was observed in the right hippocampus (21, −39, 3). No significant ac-
tivation was found in extrahippocampal mTL regions. The recognition
condition produced no significant activations within both the HC and
mTL–HC masks in the group analysis.

3.3.2. Single-subject analyses — comparison of patients with marginal and

average memory performance
Exemplarily, Fig. 3A shows significant clusters of fMRI activation in
two patients with marginal memory performance during the recogni-
tion condition of the word paradigm. In the two illustrated patients
with normal memory performance (Fig. 3B), activity within the ROI is
less present. The coordinates of the activated clusters are different for
all patients, pointing to a great variability in the anatomical distribution
of activity within the ROI which results in the absence of significant
clusters in the group analysis.
Fig. 2. Significant clusters of fMRI activation in the group analysis (N = 15) in the word-
3.3.3. Lateralization encoding condition within the hippocampus mask (p b 0.05, uncorrected; HC = hippo-
campus). Colored voxels exceeded the statistical threshold and are superimposed on the
Statistical comparisons between activations in the left and right ROIs corresponding anatomical image. The left side of each image corresponds to the left side
(i.e., HC and mTL–HC) revealed no significant differences (each p N 0.05), of the brain. (For interpretation of the references to color in this figure legend, the reader
indicating no lateralization of BOLD signal. is referred to the web version of this article.)
82 M. Milian et al. / Epilepsy & Behavior 42 (2015) 78–85

Fig. 3. Significant clusters of fMRI activation on the single-subject level. (A) Patients with marginal memory performance during the recognition condition (p b 0.05, small volume
corrected) and (B) patients with normal memory performance during the recognition condition (p b 0.05, small volume corrected). HC = hippocampus, left and right; mTL = mesial tem-
poral lobe, left and right. The slice of the brain corresponds to the peak voxel within the ROI (presupposed significant activation). Colored voxels exceeded the statistical threshold and are
superimposed on corresponding normalized anatomical images. The left side of each image corresponds to the left side of the brain. (For interpretation of the references to color in this
figure legend, the reader is referred to the web version of this article.)

hippocampal ROI (Table 2, Fig. 4A). Activations in the left hippocam- 3.4.2. Extrahippocampal mesial temporal lobe regions (mTL–HC)
pal region were not correlated with verbal memory performance Activations in the left extrahippocampal mTL region (mTLL–HCL),
(each p N 0.05). including the left amygdala and left parahippocampal gyrus, were
 M. Milian et al. / Epilepsy & Behavior 42 (2015) 78–85 83

Table 2 significantly related to the percentile ranks of the VLMT delayed recog-
Cubic correlations between fMRI word recognition activation and VLMT immediate recall, nition condition (R2 = 0.541, p b 0.05), explaining a variance of 54.1%.
delayed recall, and delayed recognition conditions (N = 15).
Similarly, there was a clear trend towards a significant cubic correlation
Recognition paradigm VLMT (percentile ranks) between the fMRI activation in the left extrahippocampal mTL region
Immediate Delayed Delayed and the VLMT immediate recall (R2 = 0.441, p = 0.084) (Fig. 4B) and
recall recall recognition delayed recall (R2 = 0.476, p = 0.060) conditions (Table 2). Activations
R2 p R2 p R2 p in the right extrahippocampal mTL region were not correlated with ver-
bal memory performance (each p N 0.05).
HC HCR 0.568 0.022 0.577 0.020 0.828 b0.001
HCL 0.279 0.289 0.111 0.718 0.113 0.710
mTL–HC mTLR–HCR 0.205 0.452 0.206 0.450 0.057 0.880 4. Discussion
mTLL–HCL 0.441 0.084 0.476 0.060 0.541 0.030

VLMT: Verbaler Lern- und Merkfähigkeitstest (wordlist learning and memory test); mTL:
mesial temporal lobe; HC: hippocampus; R: right; L: left. p b 0.05, statistical significance.
Significant results are in bold, while those marginally failed the significance level are in
italics.
The aim of the present work was to evaluate the nature of nonlinear
correlations between memory performance levels and BOLD activity in
a group of patients suffering from chronic left mTLE, a group that is well
known to have memory problems and need reliable prediction tools for
postoperative outcome.

4.1. Activation pattern and lateralization of activity

The encoding condition produced stronger activations within mesial

temporal lobe regions compared to the recognition condition, indicating
that less effort was required to recognize information [28,29]. In the
group analysis, the encoding condition produced significant right hip-
pocampal activation. We assume that the absence of activated clusters
in the group analysis during recognition was due to a large variability
in the anatomical distribution of activity between patients within the
mesial temporal lobe. Overall, activations appeared bilaterally within
the ROIs, without the strong left lateralizations that have been observed
in healthy controls.

4.2. Inefficient compensatory mTL activation may explain higher activations

in subjects with marginal memory performance

As hypothesized, nonlinear correlation models explained the rela-

tion between patients' verbal neuropsychological performance and acti-
vations in the fMRI word recognition task, with one peak in patients
with marginal memory performance and another peak in very good
performing patients. Schwarze et al. [30] already demonstrated in pa-
tients with right TLE that impaired memory performance showed in-
creased functional activity compared to healthy controls. The authors
interpreted this hyperactivation of the hippocampus as a compensatory
but inefficient process. Also in patients with mild cognitive impairment,
the preliminary phase of Alzheimer's disease, a mesial temporal hyper-
activation was found, while in patients with Alzheimer's disease, a
hypoactivation is present [14–16,31–33].

4.3. Strong preoperative activation — great postoperative decline?

The nonlinear relationship highlights that a strong activation does

Fig. 4. Scatterplots demonstrating the correlations between the behavioral data scores and
activations in the ROIs in the 15 patients with left mTLE: (A) word paradigm recognition
HCR; (B) word paradigm recognition mTLL–HCL. No linear correlations were observable
between the neuropsychological data and the sum of the t-values of suprathreshold
voxels. The solid lines show the cubic fits, and the R2 values illustrate the explained vari-
ance of the fits. The black dots represent the impaired group, the gray dots the patients
with marginal memory performance, and the white dots the patients with normal mem-
ory performance.
not necessarily go hand in hand with good memory performance. The
assumption ‘the greater the mTL activation, the greater the memory de-
cline after ATLR’ seems to apply only to a group of patients, i.e., patients
with normal or very poor memory performance in the presurgical neu-
ropsychological evaluation. In patients with marginal memory perfor-
mance, the inefficient compensatory activity might lead to wrong
conclusions. Since there is this gap in quantification of memory, fMRI
is suitable for indicating the asymmetry of mTL activation rather than
quantification. Asymmetry of memory functions was already shown to
be more valid than absolute activity for the prediction of postoperative
memory outcome [3,6,7,34,35].
4.4. Incipient preoperative reorganization of episodic memory in patients
with mTLE
All parameters of the verbal learning memory test correlated signif-
icantly with the BOLD activity within the contralateral hippocampus
84 M. Milian et al. / Epilepsy & Behavior 42 (2015) 78–85
during the recognition condition. On the other hand, when regarding
the ipsilateral side, activities within the extrahippocampal mTL regions
were related to memory performance. The found correlations with the
contralateral hippocampus as well as the ipsilateral extrahippocampal
mTL regions in this study indicate an already incipient preoperative re-
organization process of memory functions in non-amnesic patients
even before epilepsy surgery is performed. Similar to our observation,
Richardson et al. [36] assume that the contralateral activation in non-
amnesic patients with left hippocampal sclerosis represents adaptive
functional reorganization of retained verbal memory function in the
presence of left mTL pathology. This aspect is also supported by our
finding that verbal memory was correlated with nonverbal memory.
As nonverbal memory is usually processed by the nondominant, i.e.,
the contralateral hemisphere, good nonverbal memory functions point
to the functionality of the contralateral hippocampus. As a consequence,
the better verbal memory performance points to the capability of the
nonaffected hippocampus to also maintain parts of verbal memory.
Moreover, the integrity of the ipsilateral extrahippocampal mTL regions
in this study is in accordance with the results of previous research,
suggesting that patients with unilateral HS use adjacent mTL regions
(e.g., parahippocampal gyrus) rather than the affected hippocampus
to maintain memory functions [37,38]. In general, it is believed that
patients with left HS show more widespread areas of activation than
controls in terms of functional reorganization [39].
Based on our results, we propose to consider activations within the
subregions of the mTL separately, especially the contralateral HC and
the ipsilateral extrahippocampal mTL structures, to account for differen-
tial activation patterns in these regions during preoperative memory
reorganization.
4.5. Methodological aspects and limitations
This study's sample size was relatively small, but the cohort was
homogeneous concerning the pathology, i.e., patients with chronic left
mTLE with clear mesiotemporal spikes and no extrahippocampal
pathology. Moreover, all possible memory performance levels were rep-
resented, although the number of patients with impaired and normal
memory performance was overrepresented compared to those with
marginal memory.
5. Conclusions
Similar to the healthy subjects, patients with marginal memory per-
formance demonstrated a stronger activity than patients with average
memory scores, which we interpret as the effort of the patient to com-
pensate during ongoing memory deterioration. Thus, when considering
a whole continuum of all possible memory performances, the distribution
of the BOLD activity along the continuum can be described by a cubic re-
lationship. Further, we were able to demonstrate that in non-amnesic pa-
tients with chronic left mTLE but no extrahippocampal lesions, verbal
memory shifts already preoperatively to ipsilateral extrahippocampal
mTL structures as well as the contralateral HC. Thus, when interpreting
memory fMRI in the preoperative assessment, it is indispensable to take
into account firstly the neuropsychological performance and secondly ac-
tivations in the contralateral HC and ipsilateral extrahippocampal mTL
regions.
Funding
This work was supported by funding from the AKF Program of the
University of Tuebingen with the project number 275-0-0.
Conflict of interest statement
None of the authors has any conflict of interest to disclose.
References
[1] Lee TM, Yip JT, Jones-Gotman M. Memory deficits after resection from left or right an-
terior temporal lobe in humans: a meta-analytic review. Epilepsia 2002;43:283–91.
[2] Helmstaedter C, Elger CE. Cognitive consequences of two-thirds anterior temporal
lobectomy on verbal memory in 144 patients: a three-month follow-up study.
Epilepsia 1996;37:171–80.
[3] Bonelli SB, Powell RH, Yogarajah M, et al. Imaging memory in temporal lobe epilepsy:
predicting the effects of temporal lobe resection. Brain 2010;133:1186–99.
[4] Dupont S, Duron E, Samson S, et al. Functional MR imaging or Wada test: which is
the better predictor of individual postoperative memory outcome? Radiology
2010;255:128–34.
[5] Binder JR, Sabsevitz DS, Swanson SJ, Hammeke TA, Raghavan M, Mueller WM. Use of
preoperative functional MRI to predict verbal memory decline after temporal lobe
epilepsy surgery. Epilepsia 2008;49:1377–94.
[6] Powell HW, Richardson MP, Symms MR, et al. Preoperative fMRI predicts memory
decline following anterior temporal lobe resection. J Neurol Neurosurg Psychiatry
2008;79:686–93.
[7] Richardson MP, Strange BA, Thompson PJ, Baxendale SA, Duncan JS, Dolan RJ. Pre-
operative verbal memory fMRI predicts post-operative memory decline after left
temporal lobe resection. Brain 2004;127:2419–26.
[8] Gleissner U, Helmstaedter C, Schramm J, Elger CE. Memory outcome after selective
amygdalohippocampectomy in patients with temporal lobe epilepsy: one-year
follow-up. Epilepsia 2004;45:960–2.
[9] Baxendale S, Thompson P, Harkness W, Duncan J. Predicting memory decline follow-
ing epilepsy surgery: a multivariate approach. Epilepsia 2006;47:1887–94.
[10] Sabsevitz DS, Swanson SJ, Morris GL, Mueller WM, Seidenberg M. Memory outcome
after left anterior temporal lobectomy in patients with expected and reversed Wada
memory asymmetry scores. Epilepsia 2001;42:1408–15.
[11] Richardson MP, Strange BA, Duncan JS, Dolan RJ. Memory fMRI in left hippocampal
sclerosis: optimizing the approach to predicting postsurgical memory. Neurology
2006;66:699–705.
[12] Rabin ML, Narayan VM, Kimberg DY, et al. Functional MRI predicts post-surgical
memory following temporal lobectomy. Brain 2004;127:2286–98.
[13] Klamer S, Zeltner L, Erb M, et al. Nonlinear correlations impair quantification of
episodic memory by mesial temporal BOLD activity. Neuropsychology 2013;27:
402–16.
[14] Celone KA, Calhoun VD, Dickerson BC, et al. Alterations in memory networks in mild
cognitive impairment and Alzheimer's disease: an independent component analysis.
J Neurosci 2006;26:10222–31.
[15] O'Brien JL, O'Keefe KM, LaViolette PS, et al. Longitudinal fMRI in elderly reveals loss
of hippocampal activation with clinical decline. Neurology 2010;74:1969–76.
[16] Dickerson BC, Salat DH, Bates JF, et al. Medial temporal lobe function and structure in
mild cognitive impairment. Ann Neurol 2004;56:27–35.
[17] Wagner K, Frings L, Quiske A, et al. The reliability of fMRI activations in the medial
temporal lobes in a verbal episodic memory task. Neuroimage 2005;28:122–31.
[18] Milian M, Zeltner L, Klamer S, Klose U, Rona S, Erb M. BOLD signal in memory para-
digms in hippocampal region depends on echo time. J Magn Reson Imaging 2013;
37:1064–71.
[19] Oldfield RC. The assessment and analysis of handedness: the Edinburgh inventory.
Neuropsychologia 1971;9:97–113.
[20] Jenkins L, Myerson J, Joerding JA, Hale S. Converging evidence that visuospatial cog-
nition is more age-sensitive than verbal cognition. Psychol Aging 2000;15:157–75.
[21] Park DC, Lautenschlager G, Hedden T, Davidson NS, Smith AD, Smith PK. Models of
visuospatial and verbal memory across the adult life span. Psychol Aging 2002;17:
299–320.
[22] Helmstaedter C, Lendt M, Lux S. VLMT: Verbaler Lern- und Merkfähigkeitstest,
Testhandbuch. Göttingen: Hogrefe; 2000.
[23] Lamberti G, Weidlich S. DCS: a visual learning and memory test for neuropsycholog-
ical assessment. Göttingen: Hogrefe and Huber Publishers; 1999.
[24] Lehrl S. Mehrfachwahl-Wortschatz-Intelligenztest MWT-B. 5. unveränderte Aufl.
Balingen: Spitta Verlag; 2005.
[25] Mazziotta JC, Toga AW, Evans A, Fox P, Lancaster J. A probabilistic atlas of the human
brain: theory and rationale for its development. The International Consortium for
Brain Mapping (ICBM). Neuroimage 1995;2:89–101.
[26] Tzourio-Mazoyer N, Landeau B, Papathanassiou D, et al. Automated anatomical la-
beling of activations in SPM using a macroscopic anatomical parcellation of the
MNI MRI single-subject brain. Neuroimage 2002;15:273–89.
[27] Cheung MC, Chan AS, Lam JM, Chan YL. Pre- and postoperative fMRI and clinical
memory performance in temporal lobe epilepsy. J Neurol Neurosurg Psychiatry
2009;80:1099–106.
[28] Broadbent NJ, Squire LR, Clark RE. Spatial memory, recognition memory, and the
hippocampus. Proc Natl Acad Sci 2004;101:14515–20.
[29] Moser MB, Moser EI, Forrest E, Andersen P, Morris RG. Spatial learning with a
minislab in the dorsal hippocampus. Proc Natl Acad Sci 1995;92:9697–701.
[30] Schwarze U, Hahn C, Bengner T, Stodieck S, Buchel C, Sommer T. Enhanced activity
during associative encoding in the affected hippocampus in right temporal lobe
epilepsy patients. Brain Res 2009;1297:112–7.
[31] Dickerson BC, Salat DH, Greve DN, et al. Increased hippocampal activation in mild
cognitive impairment compared to normal aging and AD. Neurology 2005;65:
404–11.
[32] Dickerson BC, Sperling RA. Functional abnormalities of the medial temporal lobe
memory system in mild cognitive impairment and Alzheimer's disease: insights
from functional MRI studies. Neuropsychologia 2008;46:1624–35.
[33] Hamalainen A, Pihlajamaki M, Tanila H, et al. Increased fMRI responses during
encoding in mild cognitive impairment. Neurobiol Aging 2007;28:1889–903.
 M. Milian et al. / Epilepsy & Behavior 42 (2015) 78–85
[34] Janszky J, Jokeit H, Kontopoulou K, et al. Functional MRI predicts memory perfor-
mance after right mesiotemporal epilepsy surgery. Epilepsia 2005;46:244–50.
[35] Frings L, Wagner K, Halsband U, Schwarzwald R, Zentner J, Schulze-Bonhage A.
Lateralization of hippocampal activation differs between left and right temporal
lobe epilepsy patients and correlates with postsurgical verbal learning decrement.
Epilepsy Res 2008;78:161–70.
[36] Richardson MP, Strange BA, Duncan JS, Dolan RJ. Preserved verbal memory function
in left medial temporal pathology involves reorganisation of function to right medial
temporal lobe. Neuroimage 2003;20(Suppl. 1):S112–9.
85
[37] Banks SJ, Sziklas V, Sodums DJ, Jones-Gotman M. fMRI of verbal and nonverbal mem-
ory processes in healthy and epileptogenic medial temporal lobes. Epilepsy Behav
2012;25:42–9.
[38] Bonnici HM, Sidhu M, Chadwick MJ, Duncan JS, Maguire EA. Assessing hippocampal
functional reserve in temporal lobe epilepsy: a multi-voxel pattern analysis of fMRI
data. Epilepsy Res 2013;105:140–9.
[39] Alessio A, Pereira FR, Sercheli MS, et al. Brain plasticity for verbal and visual memo-
ries in patients with mesial temporal lobe epilepsy and hippocampal sclerosis: an
fMRI study. Hum Brain Mapp 2013;34:186–99.