Hf

MFPa

b

c

d

e

f

g

a

ARRAA

KBfHHMP

1

c

B

0d

Neuropsychologia 49 (2011) 2060–2070

Contents lists available at ScienceDirect

Neuropsychologia

journa l homepage: www.e lsev ier .com/ locate /neuropsychologia

ippocampal dysfunction in patients with mild cognitive impairment: Aunctional neuroimaging study of a visuospatial paired associates learning task

ischa de Rovera,b,c,f, Valentino A. Pironti a,c, Jonathan A. McCabea,c, Julio Acosta-Cabronerod,. Sergio Aranaa,c, Sharon Morein-Zamira,c, John R. Hodgesd,e,g, Trevor W. Robbinsa,b,aul C. Fletchera,c, Peter J. Nestord, Barbara J. Sahakiana,c,∗

MRC/Wellcome Trust Behavioural and Clinical Neuroscience Institute, University of Cambridge, Cambridge, UKDepartment of Experimental Psychology, University of Cambridge, Cambridge, UKDepartment of Psychiatry, University of Cambridge, Cambridge, UKDepartment of Clinical Neuroscience, Addenbrooke’s Hospital, Cambridge, UKMRC-Cognition and Brain Sciences Unit, Cambridge, UKCognitive Psychology Unit, Leiden University, Leiden, The NetherlandsNeuroscience Research Australia and University of New South Wales, Sydney

r t i c l e i n f o

rticle history:eceived 14 September 2010eceived in revised form 23 March 2011ccepted 27 March 2011vailable online 6 April 2011

eywords:iomarkers

MRIippocampusumansCI

aired-associates learning

a b s t r a c t

Mild cognitive impairment (MCI) patients report memory problems greater than those normally expectedwith ageing, but do not fulfil criteria for clinically probable Alzheimer’s disease. Accumulating evidencedemonstrates that impaired performance on the Paired Associates Learning (PAL) test from the Cam-bridge Neuropsychological Test Automated Battery (CANTAB) may be sensitive and specific for early anddifferential diagnosis of Alzheimer’s disease. We adapted the basic CANTAB PAL task for functional mag-netic resonance imaging (fMRI) in order to examine the functional brain deficits, at encoding and retrievalseparately, in patients with MCI compared to healthy matched volunteers. As well as investigating themain effects of encoding and retrieval, we characterized neural responses in the two groups to increasingmemory load. We focused on changes in BOLD response in the hippocampus and related structures, asan a priori region of interest based on what is known about the neuropathology of the early stages ofAlzheimer’s disease and previous information on the neural substrates of the PAL task. We also usedstructural MRI in the same patients to assess accompanying structural brain abnormalities associatedwith MCI.

In terms of the BOLD response, the bilateral hippocampal activation in the MCI and control groupsdepended upon load, the MCI patients activating significantly more than controls at low loads and sig-nificantly less at higher loads. There were no other differences between MCI patients and controls interms of the neural networks activated during either encoding or retrieval of the PAL task, including theprefrontal, cingulate and temporal cortex. The functional deficit in hippocampal activation in the MCIpatients was accompanied by structural differences in the same location, suggesting that the decrease

in hippocampal activation may be caused by a decrease in the amount of grey matter. This is one of thefirst studies to have used both encoding and retrieval phases of a memory paradigm for fMRI in MCIpatients, and to have shown that the BOLD response in MCI patients can show both hyperactivation andhypoactivation in the same individuals as a function of memory load and encoding/retrieval. The findingssuggest that performance on PAL might be a useful cognitive biomarker for early detection of Alzheimer’s

used

disease, especially when

. Introduction

As populations age worldwide, prevalence rates of age-relatedognitive disorders are rapidly increasing. Alzheimer’s disease (AD)

∗ Corresponding author at: Department of Psychiatry, Addenbrooke’s Hospital,ox 189, Cambridge CB2 2QQ, UK. Tel.: +44 1223 768009; fax: +44 1223 764760.

E-mail address: bjs-sec@medschl.cam.ac.uk (B.J. Sahakian).

028-3932/$ – see front matter © 2011 Elsevier Ltd. All rights reserved.oi:10.1016/j.neuropsychologia.2011.03.037

in conjunction with neuroimaging.© 2011 Elsevier Ltd. All rights reserved.

accounts for about 60% of all dementia cases and thus the urgencyof this problem is fast increasing (Beddington et al., 2008). Drugsthat are hoped will modify disease progression are currently underdevelopment, creating an urgent need for the early detection ofAD to enable preventative treatment to avoid any further neu-

ropathology and ensuing cognitive decline (Geerts, 2007; Masters& Beyreuther, 2006; Vellas, Andrieu, Sampaio, & Wilcock, 2007).The focus to identify individuals in the prodrome of AD has primar-ily been on patients who report memory problems greater than

M. de Rover et al. / Neuropsychologia 49 (2011) 2060–2070 2061

Table 1Demographic details of the patients with mild cognitive impairment and controls.

ControlsMean (SD)

MCIMean (SD)

Age 65.69 (5.64) 69.33 (4.17) n.s.Gender (female/male) 4/12 3/12MMSE 28.16 (1.36)

Range = 25–3025.93 (1.33)Range = 24–38

<0.001

CDR sum of boxes 1.4 (0.9)Range = 0.5–3

NART 116.25 (4.46)Range = 105–122

113.40 (5.79)Range = 116–122

n.s.

Geriatric Depression Scale 1.63 (1.45)Range = 0–7

2.27 (2.69)Range = 0–9

n.s.

.7)10–2

ncdkc(

bfchf(gSitdZsCdp2(

abrdt1tiHaFNio2

saMt1b(r

Digit span (combined forwards and backwards scores) 15.8 (2Range =

ormally expected with ageing, but who do not fulfil criteria forlinically probable AD, as their non-memory cognitive faculties andaily living activities are preserved. This condition is commonlynown as mild cognitive impairment (MCI) and is defined as alinical state in between healthy cognitive ageing and dementiaPetersen et al., 1999).

However, AD can be very difficult to diagnose at the MCI stageecause the associated cognitive decline is not easily discriminatedrom the decline found in healthy ageing and in other neuropsy-hological disorders (Petersen, 2004). Recently, important progressas been made using the Paired Associates Learning (PAL) test

rom the Cambridge Neuropsychological Test Automated BatteryCANTAB) in discriminating between those MCI patients who willo on to develop AD and those who will not (Blackwell et al., 2004;wainson et al., 2001). Accumulating evidence demonstrates thatmpaired performance on the CANTAB PAL, which is a cued recallest of memory, may be sensitive and specific for the early andifferential diagnosis of AD (Blackwell et al., 2004; de Jager, Lesk,hi, Marsico, & Chandler, 2008). Cited values for sensitivity andpecificity indices were 0.94 and 0.91, respectively, and thereforeANTAB PAL may have value as a cognitive biomarker for earlyetection of AD (Beddington et al., 2008). CANTAB PAL may alsolay a role in evaluating new anti-dementia therapies (Greig et al.,005) since it was shown to be sensitive to disease progressionAhmed, Mitchell, Arnold, Nestor, & Hodges, 2008).

Although neuropathological damage underlying AD eventuallyffects the entire brain volume (Fox, Warrington, & Rossor, 1999),rain regions are differentially affected in histological terms. Neu-ofibrillary tangles, made up of hyperphosphorylated tau, firsteposit pathology in the transentorhinal cortex, spreading earlyo the entorhinal cortex and hippocampus proper (Braak & Braak,991). This observation accords with numerous imaging studieshat have found mesial temporal atrophy in early AD includ-ng at the MCI stage (Jack et al., 2008; Pennanen et al., 2005).ypometabolism and atrophy of the posterior cingulate cortex arelso present in the MCI-stage of AD (Minoshima et al., 1997; Nestor,ryer, Smielewski, & Hodges, 2003; Pengas, Hodges, Watson, &estor, 2010); the recent development of in vivo amyloid ligand

maging has highlighted this region as one of the most intense sitesf amyloid deposition in early AD (Rowe et al., 2007; Ziolko et al.,006).

CANTAB PAL is a visuospatial associative learning test that is sen-itive and specific for the diagnosis of AD (Sahakian et al., 1988),nd is also sensitive to the MCI-stage (Swainson et al., 2001).emorizing objects in space is a function well-known to engage

he hippocampus (Maguire, Frith, Burgess, Donnett, & O’Keefe,

998; O’Keefe & Nadel, 1978; Smith & Milner, 1981), presumablyecause it is a site at which spatial and object processing convergesJones & Powell, 1970). Lesions of the hippocampus and adjacentegions including the parahippocampal gyrus in rhesus monkeys

014.2 (3.0)Range = 11–19

n.s.

impairs responding on a delayed-matching-to-sample task thatrequires object-location memory (Parkinson, Murray, & Mishkin,1988). Functional imaging evidence has tended to support a rolefor the hippocampus and the parahippocampal gyrus in associativeaspects of memory (Stark, 2007, chap. 12), including the encodingof object location (Maguire et al., 1998; Owen, Miller, Petrides, &Evans, 1996).

Consistent with its utility in the assessment of patients with MCIand Alzheimer’s disease, performance on CANTAB PAL is impairedby lesions of the temporal lobe and amygdala-hippocampectomyin human subjects, although it is also susceptible to frontal lobeinjury (Owen, Sahakian, Semple, Polkey, & Robbins, 1995). There-fore, extra-hippocampal areas are almost certainly also involved inPAL performance and the neural basis of the sensitivity of CANTABPAL for AD and MCI remains to be defined.

We converted the CANTAB PAL for use in neuroimaging stud-ies and performed a functional MRI (fMRI) study to investigatethe functional brain deficits probed by CANTAB PAL in MCI. Theconverted CANTAB PAL task has distinct encoding and retrievalphases and therefore the functional neuroimaging version maybe used to discriminate between deficits involved in encodingversus retrieval processes. Functional neuroimaging approacheshave enabled the delineation of discrete neural networks impli-cated in encoding and retrieval processes that include sectors ofthe prefrontal cortex (Fletcher & Henson, 2001; Lee, Robbins, &Owen, 2000) but there have been relatively few studies of thisissue for object-location associate learning using functional res-onance imaging (Gould, Arroyo, et al., 2006; Gould, Brown, Owen,Bullmore, & Howard, 2006; Gould et al., 2005; Gould, Brown, Owen,Ffytche, & Howard, 2003). It should also be noted that the majorityof fMRI studies of MCI have used encoding paradigms only, and toour knowledge, only one (Petrella et al., 2006) appears to have com-bined encoding and retrieval phases in the same study (Dickerson& Sperling, 2008).

Further, we collected structural neuroimaging data from thesame subjects to assess changes in grey matter density withvoxel-based morphometry (VBM), enabling us to integrate find-ings across imaging modalities. Such integration is important forthe development of biomarkers since different modalities can pro-vide complementary information and thereby superior diagnosis(Jack et al., 2008). It is also important for the elucidation of par-ticularly profound neuropsychological test impairments in MCI interms of underlying functional brain deficits and in turn for relat-ing the functional brain impairments to underlying structural brainchanges. Finally, this approach also enables a determination ofwhether MCI patients compensate for neuropathological damage,

either by recruiting additional brain areas, or by increasing acti-vation of the same brain areas, or both (Dickerson et al., 2004;Dickerson & Sperling, 2008; Hamalainen et al., 2006). In order toassess these issues more fully, memory load was explicitly manip-

2062 M. de Rover et al. / Neuropsychologia 49 (2011) 2060–2070

Fig. 1. CANTAB PAL task adapted for fMRI. (a) Structure of the experiment. Per subject there are fMRI 2 sessions with a short break in between during which the subjectrests but does not leave the scanner. Both sessions contain 2 sets of patterns for each memory load (2 times PAL4, 2 times PAL6 and 2 times PAL8) and 1 set of each controlcondition (1 control condition for PAL4, 1 for PAL6 and 1 control condition for PAL8). These nine blocks appear in pseudorandom order. Block order was counterbalancedbetween subject-groups. Each block consists of 4 phases: the encoding phase of the first repetition, the retrieval phase of the first repetition, the encoding phase of the secondrepetition and the retrieval phase of the second repetition. There is an 8.1 s delay after every phase. (b) Structure of an encoding phase. Example stimuli from the encodingphase of the task. 4, 6 or 8 patterns are displayed (2 example patterns are shown) sequentially in the available boxes (in the 8-pattern stage 2 more boxes are added). (c)Structure of a retrieval phase. In the retrieval phase each pattern is then presented in the centre of the display in random order and the subject is required to press a buttonto move a red square around over the available boxes and then press another button to confirm the choice for the box in which the pattern was previously seen. Regardlessof the number of errors all the patterns will be redisplayed in their original locations, followed by another retrieval phase (repetition 2, see Fig. 1a). (For interpretation of thereferences to colour in this figure legend, the reader is referred to the web version of the article.)

ycholo

uia

2

2

Hut1oaMnDct1eperVwawlwIfpfttvig

2

sbtrWctaeorppirepicaspwstmbwwlsoewaf

M. de Rover et al. / Neurops

lated in the functional imaging task by varying the number oftem-location pairings that participants were required to encodend retrieve.

. Materials and methods

.1. Participants

Twenty MCI patients were recruited from the Memory Clinic at Addenbrooke’sospital in Cambridge (UK). The selection procedure for MCI patients was basedpon Petersen criteria (Petersen et al., 1999) as follows: subjective memory symp-oms corroborated by an informant; objective memory impairment (lower than.5 standard deviations relative to controls on measures of delayed recall) basedn the Rey Auditory Verbal Learning Test; preserved Activities of Daily Living asscertained by the Cambridge Behavioural Inventory (Nagahama, Okina, Suzuki, &atsuda, 2006); and no evidence of dementia (based on Mini-Mental State Exami-

ation (MMSE), Folstein, Folstein, & McHugh, 1975) score of at least 24 and Clinicalementia Rating (Morris, 1993) score of 0.5. Seven of the MCI sample have pro-eeded to a diagnosis of Alzheimer’s disease (after an average of about 2 yr followinghese scans) and the others had stable MCI, although some were lost to follow up at2 months. Importantly, none of these patients improved over time. Twenty healthylderly volunteers were recruited from the general public and matched by age andre-morbid IQ to the MCI patients (Table 1). Exclusion criteria were restricted toxclusions for MRI imaging (metallic foreign body, claustrophobia, etc.), concur-ent medication targeting the central nervous system and psychiatric diagnosis.ision was normal or corrected-to-normal in every participant. All participants gaveritten informed consent according to the Helsinki Declaration and the study was

pproved by the Ethics Committee of the University of Cambridge. Control subjectsere excluded if they had had a psychiatric illness, epilepsy, a serious head injury,

oss of consciousness, surgery requiring extended hospitalisation, medication thatould interfere with cognitive performance or medication for depression or anxiety.

n the control group two participants were excluded because of equipment mal-unctioning, one participant was excluded because of excessive movement, and onearticipant was excluded because his recall performance was not significantly dif-erent from chance level. In the MCI group two participants were excluded becausehey were unable to finish the task and three participants were excluded becauseheir recall performance was not significantly different from chance level. Sixteenolunteers were finally included in the control group and fifteen volunteers werencluded in the MCI group (age of the control group: range 60–76; age of the MCIroup: range 63–77; n.s.).

.2. Experimental procedure

In the CANTAB PAL test, visual abstract, coloured patterns (‘objects’) are pre-ented (and extinguished) randomly and sequentially, one by one in six or eightoxes around the edge of a touch screen computer. After a brief delay, the same pat-erns are presented in the middle of the screen in random order and the subject isequired to touch the box in which they saw that pattern appear (www.cantab.com).

e converted the CANTAB PAL for use in neuroimaging studies, but made as fewhanges as possible to optimize the comparability between the neuroimaging andhe CANTAB PAL. Thus, in the neuroimaging PAL: 4 (PAL4), 6 (PAL6) or 8 (PAL8)bstract coloured patterns were presented one-by-one in random order in six oright boxes (encoding phase, Fig. 1a and b) back-projected via an LCD-projectornto a translucent screen that subjects viewed while in the scanner through a mir-or mounted on the head coil. Subjects were instructed to try to remember whichattern belonged in which box. After a brief delay (8.1 s), the same patterns wereresented one by one in the middle of the screen and subjects were required to

ndicate the box in which the patterns belonged by (retrieval phase, Fig. 1c). Theetrieval phase was followed by another brief delay (8.1 s) after which the nextncoding phase (i.e. for 6 or 8 item loads) started. Because of the 8.1 s delay, eachhase started at a different point on the BOLD response. No further jittering was

ncluded because the fMRI data were analyzed in a block design. The followinghanges were made during the conversion between the neuropsychological testnd the imaging paradigm: instead of touching a box on a touchscreen computer,ubjects’ responses were registered by button presses. In the retrieval phase, eachattern was presented in the centre of the display in random order and subjectsere instructed to use button presses to indicate the box in which they previously

aw the item appear. There were three buttons which subjects could press withheir right index, middle and ring finger. Subjects moved a red square, using their

iddle finger (clockwise) and/or ring finger (counter clockwise) and then press theutton under their index finger to confirm the choice for the box in which the objectas previously seen, after which the red square would turn green (Fig. 1). Further,e standardized durations of the retrieval phase, such that every item of which the

ocation was to be retrieved was presented for 7 s, regardless of the time it took for aubject to respond (Fig. 1). In the CANTAB PAL the encoding and retrieval of a set of

bject-location associations would be repeated until the subject correctly retrievedvery object-location association in that particular set. For neuroimaging purposes,e standardized the number of repetitions, such that every set of object-location

ssociations would be encoded and retrieved twice, regardless of behavioural per-ormance. Within groups we pseudorandomized the order of the blocks such that

gia 49 (2011) 2060–2070 2063

there was no relationship between memory load and (scanning) time (to avoid aninteraction between scanner drift and memory load). The order of the blocks wascounterbalanced between groups, such that the same number of participants wouldstart with memory load 4 (PAL4) in the control group as in the MCI group and soon for the other loads and the second and later blocks. As a final change from theoriginal task, we introduced control conditions: a visual control condition for theencoding phase and a visuo-motor control condition for the retrieval phase. Thevisual control condition consisted of the same six or eight boxes as presented in theencoding condition, but this time, instead of the visual patterns, grey squares werepresented. Subjects were instructed to rest (but not close their eyes) when they sawthe grey squares and not to try to remember anything about the grey squares. Thevisuo-motor control condition consisted of the same six or eight boxes as presentedin the retrieval phase, but this time, instead of the patterns, an arrow was presentedin the middle of the screen, pointing to one of the boxes. Subjects were instructedto select the box to which the arrow was pointing by button presses, such that sub-jects would make the same motor responses as in the retrieval phase but withoutretrieval.

2.3. MRI data acquisition

MRI data were acquired at the MRC-Cognition and Brain Sciences Unit inCambridge. Whole head T2*-weighted EPI-BOLD fMRI data were acquired with aSiemens Trio 3T MR scanner with Tim (total imaging matrix technology) using adescending slice acquisition sequence (EPI; volume TR = 2000 ms, TE = 30 ms, 78◦

flip-angle, 32 axial slices, slice-matrix size = 64 × 64, slice thickness = 3 mm, 0.75 mmgap, FOV = 192 mm, isotropic voxel-size = 3 mm × 3 mm × 3 mm). Task presentationand recording of the behavioural responses was performed using Visual Basic 2005.Functional data were acquired in two runs of 20 m each.

Additionally, high-resolution structural MR images were acquired with a T1-weighted MP-RAGE sequence (volume TR = 2.25 s, TE = 3.93 ms, 15◦ flip-angle, 176sagittal slices, slice-matrix size = 256 × 256, slice thickness = 1 mm, no slice gap,voxel-size = 1 mm × 1 mm × 1 mm).

2.4. MR image preprocessing and statistical analysis

Image preprocessing and statistical analysis was executed with the StatisticalParametric Mapping 5 software (SPM5, Wellcome Department of Imaging Neuro-science, University College London, London, UK; www.fil.ion.ucl.ac.uk). The first siximages were discarded in order to reduce T1 saturation effects. The functional EPI-BOLD images were realigned and the subject-mean functional MR images wereco-registered with the corresponding structural MR images using mutual infor-mation optimization. The structural MR images were processed with SPM5, whichenables spatial normalization – i.e. warping to MNI152 standard space (MontrealNeurological Institute and Hospital, McGill University, Montreal, Canada) – tis-sue classification and radio-frequency bias correction to be combined within itsmodel (Ashburner & Friston, 2005). Preprocessing steps such as skull-stripping andbias-field correction are known to improve the performance of warping algorithms(Acosta-Cabronero, Williams, Pereira, Pengas, & Nestor, 2008; Fein et al., 2006);hence raw whole-head scans were pre-processed prior to SPM5 with the automatedpipeline described elsewhere (Acosta-Cabronero et al., 2008; Pereira et al., 2010).Briefly, skull-stripping was performed using the hybrid watershed algorithm, HWA(Segonne et al., 2004). Stripped volumes were then bias-corrected using the non-parametric non-uniform intensity normalization or N3 v.1.10 (Sled, Zijdenbos, &Evans, 1998) and finally, fine brain extractions were performed using BET v.2.1(Smith, 2002). To evaluate brain morphometry using voxel-based morphometry(VBM; Ashburner & Friston, 2000), grey matter segments were modulated to pre-serve the total amount of tissue, and spatially filtered with an isotropic, 8-mmFWHM Gaussian kernel (Hayasaka & Nichols, 2003; Petersen et al., 1999). A generallinear model (GLM) was then fit at each voxel, with one variable of interest (group),and a nuisance covariate: total intracranial volume (TIV) calculated using a previ-ously described method (Pengas, Pereira, Williams, & Nestor, 2009). Grey matterdifferences between groups were assessed using a two population t-test (patientsworse than controls); the resulting statistical parametric map was thresholded atuncorrected p < 0.001. We found no differences in total intracranial volume (TIV)or in overall grey matter density between our healthy control group and our MCIgroup (average TIV in the control group: 1.7 ± 0.2 in the MCI group: 1.8 ± 0.2; t-testp > 0.05). A post hoc analysis of the reverse contrast (patients better than controls)showed no brain matter density loss in controls relative to patients (p > 0.05).

To spatially normalize the functional MR images, the transformation matri-ces generated from the unified segmentation process were applied; the resultingwarped images were then smoothed with an 8-mm FWHM kernel. The fMRI datawere proportionally scaled to account for global effects and analyzed statisti-cally using the general linear model and statistical parametric mapping (Friston,Holmes, Worsley, Poline, & Frackowiak, 1995). The linear model included convolvedexplanatory variables (box-car regressors) modeling the experimental conditions in

a blocked fMRI design. The 24 experimental conditions modeled included a factorialdesign of memory load (PAL4, PAL6, PAL8) by memory stage (encoding, encodingcontrol, recall and recall control) by repetition (first and second repetition). Theseexplanatory variables were temporally convolved with the canonical hemodynamicresponse function provided by SPM5. In addition, the linear model included as

2064 M. de Rover et al. / Neuropsychologia 49 (2011) 2060–2070

Table 2Addenbrooke’s Cognitive Examination Revised (ACE-R) scores of MCI patients and control population.

ControlsMean (SD)

MCIMean (SD)

Orientation/attention (Max = 18) 17.87 (0.34)Range = 17–18

17.20 (1.08)Range = 14–18

<0.001

Memory (Max = 26) 23.73 (2.31)Range = 17–26

16.40 (4.39)Range = 8–24

<0.001

Verbal fluency (Max = 14) 11.58 (1.86)Range = 6–14

9.73 (2.09)Range = 4–12

<0.001

Language (Max = 26) 24.65 (1.37)Range = 19–26

24.47 (1.30)Range = 22–26

n.s.

Visuospatial (Max = 16) 15.59 (0.83)Range = 12–16

15.33 (0.90)Range = 14–16

n.s.

Total (Max = 100) 93.42 (4.20) 83.13 (6.08)Range = 69–92

<0.001

epatPsyf

gIo22tNiemoti

apgcpgM

3

msoMpR

0102030405060708090

100

Perc

enta

ge c

orre

ct

8 LAP4 LAP PAL 6

** ****

Con MCI Con MCI Con MCI

Fig. 2. Overall cued recall performance (percentage correct answers ± SEM) inthe control and the MCI group. White bar Con = control group, n = 16. Grey bar

Fo

Range = 78–100

ffects of no-interest: session/subject-effects, realignment parameters, and a tem-oral high-pass filter to account for various low-frequency effects. In the statisticalnalysis, all relevant contrasts were set up: every memory-related condition minushe corresponding control condition as a baseline (for example PAL4 encoding minusAL4 encoding control condition); and used to generate contrast images for eachubject, which were subsequently subjected to a second-level random effect anal-sis, using ANOVA. The terms of activation and deactivation are used as synonymsor a relative increase and decrease in BOLD signal respectively.

To test for the specific role of the hippocampus and parahippocampalyrus during encoding and retrieval, based on the findings reviewed in thentroduction, the hippocampus and the parahippocampal gyrus were a pri-ri defined as regions of interest (ROI) using the pick atlas tool (version.4: http://fmri.wfubmc.edu/cms/software) (Maldjian, Laurienti, Kraft, & Burdette,003). To test null hypotheses in ROI analysis, a small volume correction for multipleesting was applied using the false discovery rate (FDR) method (Genovese, Lazar, &ichols, 2002). For whole brain random effect analysis correction for multiple non-

ndependent comparisons was based on the family-wise error (FWE) rate (Worsleyt al., 1996). In the second-level random effect analysis, the following contrasts wereade both for the encoding and the retrieval phase: the main effects irrespective

f load, group differences in these main effects, the main effects of deactivation,he main effect of load and group differences in load, as well as the load x groupnteractions, and their subsequent post hoc tests.

We investigated deactivations in the contrast ‘control condition versus PAL4’nd checked for group effects. To investigate regional attenuated deactivations inatients we masked this contrast with the main effect of deactivation in the controlroup and then looked for brain areas in which the deactivations were larger in theontrol than in the MCI group. To investigate regional augmented deactivations inatients we masked this contrast with the main effect of deactivation in the MCIroup and then looked for brain areas in which the deactivations were larger in theCI than in the control group.

. Results

The MCI and control groups were well matched for age, pre-orbid verbal IQ and gender distribution. There was also no

ignificant difference in scores on the Geriatric Depression Scale

r digit span although, as expected, they differed significantly inMSE scores [t(29) = 4.52, p < 0.001] (Table 1). In addition, MCI

atients were given the Addenbrooke’s Cognitive Examinationevised (ACE-R, Mioshi, Dawson, Mitchell, Arnold, & Hodges, 2006).

ig. 3. ROI analysis revealed significant bilateral activations in the hippocampus in all partf an individual high-resolution T1-weighted volume. Only significant clusters are shown

MCI = MCI group, n = 15. Healthy volunteers (Con) performed significantly betterthan MCI patients (MCI) in every PAL-stage tested (PAL4, PAL6 and PAL8: p < 0.001).**p < 0.001.

Their scores are reported in Table 2 along with the scores of areference control population.

3.1. Behavioural results

As can be seen in Fig. 2, at all three difficulty levels thepercentage of correctly recalled object-location associations wassignificantly lower in the MCI group compared to the controlgroup [PAL4: t(29) = 6.33, p < 0.001; PAL6: t(29) = 3.74, p < 0.001;PAL8: t(29) = 5.32, p < 0.001]. The effect sizes were 2.28, 1.34 and1.92 for PAL4, PAL6 and PAL8, respectively. In addition to themain effect of group, an Analysis of Variance (ANOVA) with

group, difficulty and repetition as independent factors revealedmain effects of difficulty [F(2,58) = 29.00, p < 0.001] and repetition[F(1,29) = 150.10, p < 0.001]. The interaction between difficulty andgroup did not reach significance, but was marginal [F(2,58) = 2.88,

icipants during encoding. Activations are shown superimposed onto selected slices(p < 0.05 FDR corrected).

M. de Rover et al. / Neuropsychologia 49 (2011) 2060–2070 2065

-1.500

-1.000

-0.500

0.000

0.500

1.000

4 6 8

Load

a

b

Bet

a Va

lue

Control

MCI

4 6 80.000

0.200

0.400

0.600

0.800

1.000

1.200

1.400

Load

Bet

a V

alue

ControlMCI

Fig. 4. (a) Parameter estimates of bilateral hippocampal activation, shown for MCIand control groups and for the three difficulty levels (PAL4, PAL6 and PAL8 on thexsP

piwbtPtfsbtd

3

33plrrwpsec

ahit

Fig. 5. Brain regions activated in a whole brain random effect analysis comparing the

dle frontal gyri, superior temporal gyrus, extending into parietalcortex, middle occipital and cingulate gyri (p < 0.05, FWE corrected,Fig. 5 and Table 3). There were no significant effects of group, mem-ory load or repetition in this analysis. The whole brain contrast of

Table 3Brain regions differentially activated during the encoding and during the retrievalphases of the PAL task. Not including extranuclear local maxima and brain stemareas.

Side (L/R) Z-score Local maxima

EncodingMiddle occipital gyrus R Inf 30 −90 0Superior temporal gyrus L 5.81 −48 −48 12

R 5.07 46 −28 0Middle cingulate gyrus R 5.51 2 −30 30Middle frontal gyrus R 5.28 24 38 −14Anterior cingulate gyrus R 5.04 0 6 28Lingual gyrus L 4.64 −16 −48 −4Precuneus R 4.60 8 -42 46

-axis). (b) Parameter estimates of parahippocampal peaks during retrieval stage,hown for control and MCI groups for the three difficulty levels (PAL4, PAL6 andAL8 on the x-axis).

< 0.07], which stemmed from a slightly smaller group differencen PAL6 compared to PAL4 and PAL8. Cued recall performance

as significantly above chance at all three levels of difficulty inoth groups [Control group: PAL4: t(15) = 16.50 p < 0.001; PAL6:(15) = 12.73, p < 0.001; PAL8: t(15) = 10.14, p < 0.001; MCI group:AL4: t(14) = 4.26, p < 0.001; PAL6: t(14) = 6.23, p < 0.001; PAL8:(14) = 5.72, p < 0.001]. For PAL4 and PAL6, chance level is 16.7% andor PAL8, chance level is 12.5%. No other interactions approachedignificance (p > 0.05). The number of non-responses was low inoth groups (total average: MCI group: 2.47, (range = 0–8); con-rol group: 0.93 (range = 0–7)) but was nonetheless significantlyifferent [F(1,29) = 12.6, p < 0.01].

.2. Neuroimaging results

.2.1. Functional imaging

.2.1.1. Encoding stage. ROI analyses in the hippocampus for allarticipants examining the effects of encoding of the object-

ocation associations as compared to the visual control conditions,evealed significant bilateral activations (Fig. 3, p < 0.05, FDR cor-ected). Interrogating this finding further revealed that this effectas primarily due to significant bilateral activations in the hip-ocampus in control participants (p < 0.05, FDR corrected) as noignificant effect was found for the MCI group alone. No significantffects were found for the parahippocampal gyrus ROI (p > 0.05 FDRorrected).

ROI analyses found no significant main effects of memory load,

lthough there was a significant group by load interaction for theippocampus bilaterally. As shown in Fig. 4 this is driven by an

ncrease in hippocampal activation with increasing load in the con-rol group and a decreasing hippocampal activation with increasing

encoding phases versus the visual control conditions in all participants. Activationsare shown on an individual brain rendered in 3D. Only significant clusters are shown(FWE p < 0.05).

memory load in the MCI group (Fig. 5, p < 0.05 FDR corrected, MNIcoordinates were −30, −14, −20 and 32, −12, −20 for left and righthippocampus respectively). Post hoc analyses revealed a significantdifference between the control group and the MCI group during thePAL8, t(29) = 2.8, p = 0.009; a trend toward significant results in thePAL4, t(29) = −1.942, p = 0.062; and no significant difference duringthe PAL6, t(29) = 0.603, p = 0.551.

In a whole brain, random effect analysis, we investigatedwhether there were effects of encoding of the object-location asso-ciations in other parts of the brain. We examined encoding ascompared to the visual control conditions and found significantactivation of several brain areas, including the inferior and mid-

Superior frontal gyrus R 4.53 24 52 −2RetrievalPosterior cingulate L 5.09 −8 −54 12Cuneus L 4.57 2 −78 30

2066 M. de Rover et al. / Neuropsychologia 49 (2011) 2060–2070

Fig. 6. ROI analysis revealed significant bilateral activations in the parahippocampal gyrus in all participants during retrieval. Activations are shown superimposed ontoselected slices of an individual high-resolution T1-weighted volume. Only significant clusters are shown (p < 0.05 FDR corrected).

F retriea eighte

l(wi

3tcartgpcfnco

wccipob

ig. 7. Brain regions activated in a whole brain random effect analysis comparing there shown superimposed onto selected slices of an individual high-resolution T1-w

oad x group showed a significant cluster in the left hippocampusp < 0.001, uncorrected; MNI coordinates: −30, −14, −20). Thereere no regional deactivations in the control group that are signif-

cantly attenuated or augmented in the patients.

.2.1.2. Retrieval stage. ROI analyses of the effects of retrieval ofhe pattern-location associations as compared to the visuomotorontrol conditions in all participants revealed significant bilateralctivations in the parahippocampal gyrus (Fig. 6, p < 0.05, FDR cor-ected) but no effect was found in the hippocampus. Investigatinghis finding further suggested that it was driven by the controlroup, given significant bilateral activations in the parahippocam-al gyrus for the controls but not for the MCI group (p < 0.05, FDRorrected; MNI coordinates were −18, −40, −8 and 16, −40, −6or the left and right parahippocampal gyrus respectively). No sig-ificant effects were found for the hippocampus ROI (p > 0.05 FDRorrected). ROI analyses examining the interaction between mem-ry load and group did not reveal any significant results.

In a whole brain random effect analysis, we investigatedhether there were effects of retrieval of the object-location asso-

iations in other parts of the brain. We examined retrieval asompared to the visuomotor control condition. There was signif-

cant activation of the posterior cingulate and left cuneus (Fig. 7,< 0.05, FWE corrected, Table 3). There were no significant effectsf group, memory load or repetition in this analysis. The wholerain contrast of load x group showed significant activation in the

val phases versus the visuo-motor control conditions in all participants. Activationsd volume. Only significant clusters are shown (p < 0.05: FWE corrected).

left hippocampus (p < 0.001, uncorrected; MNI co-ordinates: −18,−8, −16).

3.2.2. Structural imaging: voxel-based morphometryWhole brain analysis showed that, compared with the healthy

control group, the MCI group had significant reductions of greymatter density in the left and right hippocampus (MNI coordi-nates: −26, −14, −12 and 28, −10, −12 respectively) and in theleft precuneus (MNI coordinates: −4, −74, 54) (Fig. 8).

4. Discussion

This study has shown that the neural network activated bya commonly used and highly sensitive neuropsychological testfor MCI and early AD, requiring visuospatial paired associateslearning, includes the hippocampus during encoding and theparahippocampal gyrus during retrieval. MCI patients perform-ing the task exhibited both hyper-activation and hypo-activationof the hippocampus as a function of memory load during encod-ing. In addition, we found corresponding structural deficits in thehippocampus of MCI patients compared to healthy controls.

Medial temporal lobe degeneration has long been assumed to be

the main neural substrate for the early memory deficits observed inAlzheimer’s disease, including prodromal amnestic MCI-stage AD,as revealed by structural or metabolic neuroimaging (Apostolovaet al., 2006; Chetelat et al., 2003; Jack et al., 2008, 2009; Whitwell

M. de Rover et al. / Neuropsychologia 49 (2011) 2060–2070 2067

Fig. 8. Voxel-based morphometry results. The SPM map overlaid onto the Colin-27 template represents significant grey matter density loss in 15 MCI subjects compared to1 t thresw s (top

eyeDD

oaAthchDebado

(mhscsStfdirA

lp

6 controls. Results are reported at an uncorrected p < .001 threshold and an extenas applied to exclude false positives. The two areas affected were the hippocampu

t al., 2007). In contrast, fMRI has been less employed in the anal-sis of memory decline in MCI, although its use is growing (Bait al., 2009; Dannhauser et al., 2008; Dickerson et al., 2004, 2005;ickerson & Sperling, 2008; Machulda et al., 2003; Small, Perera,eLaPaz, Mayeux, & Stern, 1999; Trivedi et al., 2008).

The early fMRI studies tended in general to reflect the findingsf other modalities. Thus, Small et al. (1999), Machulda et al. (2003)nd Johnson et al. (2004) all found that patients with MCI or earlylzheimer’s disease exhibited less activation during a memory test

han healthy controls. Bai et al. (2009) found altered patterns ofippocampal connectivity during episodic memory retrieval. Byontrast, a number of more recent studies have found greaterippocampal activation in MCI during fMRI (Celone et al., 2006;ickerson et al., 2004, 2005; Hamalainen et al., 2006) under a vari-ty of testing conditions. A recent example of this is the studyy Trivedi et al. (2008) which found greater right hippocampalctivation (but less frontal cortex activation) relative to controlsuring certain test conditions, specifically the intentional encodingf items subsequently recognized.

In their review of fMRI studies of MCI, Dickerson and Sperling2008) conclude that early in the course of this disorder, when

emory deficits and hippocampal atrophy are less prominent,yperactivation of medial temporal lobe circuits possibly repre-ents compensatory activity in response to the task employed. Thisompensatory hyperactivity may be analogous to that reported ineveral other brain disorders or disease, although Dickerson andperling discuss alternative possibilities linked more specifically tohe Alzheimer disease process. Later in the course of the transitionrom MCI to clinical Alzheimer’s disease, functioning of the MTLeteriorates further to an extent that such compensatory activity

s no longer possible. The hyperactivity in early MCI might thenepresent a possible predictor or biomarker of the progression to

lzheimer’s disease.

The present study confirmed the importance of medial temporalobe structures, either when specifically focusing on the hippocam-us (during encoding, Fig. 4a) and parahippocampal gyrus (during

hold of 20. TIV was entered as a nuisance covariate and a relative threshold of 0.1) and the precuneus (bottom).

retrieval, Fig. 4b) as regions of interest, specified either a priori oremerging from the whole brain, random effect analysis (for encod-ing, the left hippocampus, as a function of memory load). Whereascontrol subjects exhibited increases in activity from the lighter tothe heavier loads during the encoding stage of the visual object-location memory task, MCI patients exhibited the opposite trend;significant increases over control subjects at light memory loads,but significant reductions for the heaviest memory load (Fig. 4a).These changes were not directly reflected in memory performance,which was significantly impaired at all levels of load, althoughthere was a marginally significant tendency for the deficit in MCIpatients to be less at the intermediate memory load. One possibleinterpretation of this finding is that the MCI patients attemptedto recruit additional processing resources within the hippocam-pus when the task was still tractable for them, but could not doso when the task was at its most challenging, perhaps becauseof their failure to encode efficiently. In the present study we thussaw both hippocampal hyperactivity and hypoactivity in the samepatients as a function of memory load. By comparison with theMCI patients tested by Dickerson and Sperling (e.g. 2005) our MCIpatients were not particularly mild, with MMSE average scores ofabout 26 (compared with 29.6 + 0.5SD); this might explain whywe were able to observe both hypoactivity and hyperactivity, asa function of task difficulty. O’Brien et al. (2010) have noted thathippocampal hyperactivity may herald the prodromal stages ofAlzheimer’s disease and be an indicator of impending hippocampalfailure.

The memory material for the present study has most in commonwith that used by Gould et al. (2005), Gould, Arroyo, et al. (2006) andGould, Brown, et al. (2006) who adopted the CANTAB PAL designand mode of stimulus presentation, but used drawings of everydayobjects rather than abstract stimuli and tested patients with mild

Alzheimer’s disease rather than MCI. In these studies hippocam-pal deficits were not found but may have been underestimatedbecause their use of everyday objects may have enabled strategiesfor encoding that changed the nature of the task. Moreover, they

2 sychol

uitp

iACc&bfiboiidi

uepatosFossGromtArme

icnrwmmeahvtici(wcrttptti

068 M. de Rover et al. / Neurop

sed a force choice recognition paradigm to test retrieval, whichs clearly very different from the standard PAL approach in whichhe participant is presented with a stimulus and must recollect itsrior position.

Previous neuropsychological studies using PAL have shownmpairments in MCI patients that predict the future diagnosis oflzheimer’s disease up to 32 months later, indicating that theANTAB PAL task may be a neuropsychological ‘marker’ of pre-linical AD (Blackwell et al., 2004; Fowler, Saling, Conway, Semple,

Louis, 1995, 1997, 2002; Swainson et al., 2001). The deficits inehavioural performance of the present version of PAL adapted forMRI were quite profound and found at easier stages of the task thann the previous neuropsychological studies. However, this proba-ly reflected the fact that these MCI patients had already receivedbjective memory tests such as the Rey Auditory-Verbal Learn-ng Task and therefore had been previously screened for memorympairment rather than in the earlier studies when PAL effectivelyiscriminated patients with subjective memory disorders present-

ng for the first time at a memory clinic.It is a possible limitation of the present study that we were

nable to define a level of performance of the PAL task that wasquivalent between MCI patients and controls, although this hasroven difficult (though not impossible) to achieve in other studieslso (Dickerson & Sperling, 2008). Careful matching of behaviouralest memory scores would have provided the most powerful testf whether a functional neuroimaging biomarker could be a moreensitive index of deficit than a behavioural or cognitive one.uture studies may achieve behavioural matching with a PAL2r PAL3 version of the task. It is nevertheless quite controver-ial whether a functional neuroimaging biomarker would be moreensitive, as a recent meta-analysis (Schmand, Huizenga, & vanool, 2010) concluded that memory impairment is a more accu-

ate early predictor of AD than atrophy of the medial temporal loben magnetic resonance imaging (whereas CSF abnormalities andemory impairment are about equally predictive). This means that

he possibility of a cognitive biomarker for the detection of earlylzheimer’s disease may be as plausible as a biochemical or neu-oimaging biomarker, although it is the case that such a biomarkeray well have problems of specificity, unless tied to other evidence,

.g. progressive decline.The current study confirmed that different neural networks are

mplicated in encoding and retrieval, at least for this visuospatialued recall task. The encoding stage involved an extensive neuraletwork including different sectors of the prefrontal cortex, ante-ior cingulate, temporal–parietal and occipital cortex, consistentith many other studies (Fletcher & Henson, 2001; Lee et al., 2000),ost of which however utilised verbal rather than visual memoryaterial. However, we could find no evidence to support a differ-

ntial hemispheric asymmetry in encoding (nor indeed in retrieval)s suggested by the hemispheric encoding/retrieval asymmetryypothesis (Tulving et al., 1994) – possibly because of the non-erbal nature of the memory material used in this study in contrasto much of the literature (Lee et al., 2000). By contrast, retrievalmplicated a much more restricted network comprising mainly theuneus and posterior cingulate, consistent with numerous stud-es highlighting activation in this region with memory retrievalCabeza & Nyberg, 2000; Svoboda, McKinnon, & Levine, 2006). Itas interesting to see that the hippocampus proper was impli-

ated in the encoding network, but the parahippocampal gyrus inetrieval. Presumably, this difference reflects the changing roles ofhe hippocampus during encoding and retrieval processes. Of par-icular significance for this study is that differences between MCI

atients and controls were limited to the hippocampal formation,here being no evidence of differential activation of the other struc-ures, such as the prefrontal cortex and posterior cingulate cortex,n these memory networks.

ogia 49 (2011) 2060–2070

The posterior cingulate cortex is often described as part of a so-called “resting state network”, exhibiting ongoing activity duringrest and deactivation during task performance (Greicius, Krasnow,Reiss, & Menon, 2003; Gusnard, Raichle, & Raichle, 2001; Mazoyeret al., 2001; McKiernan, Kaufman, Kucera-Thompson, & Binder,2003; Raichle et al., 2001). Severe hypometabolism and focal atro-phy in the posterior cingulate cortex have been previously shownto be associated with MCI, suggesting that the posterior cingulatecortex is as vulnerable to neurodegeneration as the hippocampus(Nestor et al., 2003; Pengas et al., 2010). It is thus significant thatwe found the posterior cingulate cortex to be activated during theretrieval phase of the CANTAB PAL task. However, we did not findany differences in this activation between MCI patients and healthycontrols. Further, although recent studies have shown that the pos-terior cingulate cortex is highly atrophic in incipient Alzheimer’sdisease (Choo et al., 2010; Pengas et al., 2010) our VBM analysisdid not reveal any significant structural differences in the posteriorcingulate area between this sample of MCI patients and healthycontrols.

There were no significant differences in intra-cranial volumebetween the MCI patients and controls and grey matter densityreductions were detected only in the hippocampus, bilaterally andthe left precuneus. Given the previous evidence of hippocampalinvolvement of PAL performance (Owen et al., 1995) we assumethat the present hippocampal structural deficit underlies, at leastin part, the deficient performance of these patients on the PAL task.However, it is apparent that this structural deficit can be asso-ciated with both hyper- and hypo-activation of this structure, asmeasured using the BOLD response. The use of fMRI thus providesimportant ancillary information as to the functional status of thiscompromised structure in MCI and the present study underlines theimportance of using several imaging modalities, in association withneuropsychological tests, to assess MCI with a view to predictingthe progression to Alzheimer’s disease.

It is important to note that we observed both structural andfunctional changes in the patient group, but a limitation of thestudy is that the group size precludes detailed analysis of inter-actions between these variables. The relationship, for instance,between areas of functional activation and atrophic areas is likely tobe important. Although localised differently (the functional groupdifference in the posterior hippocampal/parahippocampal region,the structural change maximal in the anterior hippocampus), itremains possible that the two observations are intimately relatedbut that the limited numbers studied here do not allow us to elu-cidate this relationship further.

This study provides new information about the neural basis ofthe deficits in MCI patients on the CANTAB PAL task. As the CANTABPAL task is evidently a sensitive test for predicting progression toAlzheimer’s disease in patients with MCI it could also be potentiallyuseful for evaluating its treatment. The PAL task is, moreover, a cuedrecall task, as specifically recommended for early detection in arecent influential revision of diagnostic criteria for Alzheimer’s dis-ease (Dubois et al., 2007), albeit of delayed visuospatial, as distinctfrom delayed verbal, recall. One important translational advantageof the human PAL task, with its visuospatial learning characteristicsis that, unlike the many verbal memory paradigms commonly usedto assess MCI, there are analogues of the test that can be used indrug development in experimental animals, whether non-humanprimates (Parkinson et al., 1988; Taffe, Weed, Gutierrez, Davis, &Gold, 2002; Weed et al., 1999) or rodents (Talpos, Winters, Dias,Saksida, & Bussey, 2009).

Ultimately, it is anticipated that a better characterization of MCI

will contribute to a sensitive and specific early detection of AD andenable preventative treatment. Most importantly, this study hasshown that hippocampal dysfunction can be detected at a very earlystage in MCI patients at least 2 yr (on average, thus far) prior to AD

ycholo

dapdm

A

fiCfFgo0piW

aA

R

A

A

A

A

AB

B

B

B

C

C

C

C

D

d

D

D

D

M. de Rover et al. / Neurops

iagnosis. This emphasizes the importance of sensitive cognitivend imaging measures that can be used in screening healthy elderlyeople so that preventative treatments, including neuroprotectiverugs, can be administered before there is a significant decline inental capital and well-being (Beddington et al., 2008).

cknowledgements

We thank an anonymous donor to Cambridge in America fornancial support for this research and the staff at the MRC-ognition and Brain Sciences Unit for excellent assistance with

MRI scanning. MdR was funded by a Marie Curie Intra-Europeanellowship within the 6th European Community Framework Pro-ramme. VAP is funded by an MRC studentship. SM-Z is supportedn a Programme Grant from the Wellcome Trust (WT Ref. No.89589/Z/09/Z). JA-C is sponsored by ARUK. The study was com-leted within the Behavioural and Clinical Neuroscience Institute

n Cambridge (UK) supported by a joint-grant from the MRC andelcome Trust.SA is on the NIH-Cambridge University Scholars’ Program. BJS

nd TWR consult for Cambridge Cognition (CANTAB). We thankndrew Blackwell and ET Bullmore for discussion.

eferences

costa-Cabronero, J., Williams, G. B., Pereira, J. M., Pengas, G., & Nestor, P. J.(2008). The impact of skull-stripping and radio-frequency bias correctionon grey-matter segmentation for voxel-based morphometry. Neuroimage, 39,1654–1665.

hmed, S., Mitchell, J., Arnold, R., Nestor, P. J., & Hodges, J. R. (2008). Predictingrapid clinical progression in amnestic mild cognitive impairment. Dementia andGeriatric Cognitive Disorders, 25, 170–177.

postolova, L. G., Dinov, I. D., Dutton, R. A., Hayashi, K. M., Toga, A. W., Cummings, J. L.,et al. (2006). 3D comparison of hippocampal atrophy in amnestic mild cognitiveimpairment and Alzheimer’s disease. Brain, 129, 2867–2873.

shburner, J., & Friston, K. J. (2000). Voxel-based morphometry – the methods.Neuroimage, 11, 805–821.

shburner, J., & Friston, K. J. (2005). Unified segmentation. Neuroimage, 26, 839–851.ai, F., Zhang, Z., Watson, D. R., Yu, H., Shi, Y., Yuan, Y., et al. (2009). Abnormal

functional connectivity of hippocampus during episodic memory retrieval pro-cessing network in amnestic mild cognitive impairment. Biological Psychiatry,65, 951–958.

eddington, J., Cooper, C. L., Field, J., Goswami, U., Huppert, F. A., Jenkins, R., et al.(2008). The mental wealth of nations. Nature, 455, 1057–1060.

lackwell, A. D., Sahakian, B. J., Vesey, R., Semple, J. M., Robbins, T. W., & Hodges, J.R. (2004). Detecting dementia: Novel neuropsychological markers of preclinicalAlzheimer’s disease. Dementia and Geriatric Cognitive Disorders, 17, 42–48.

raak, H., & Braak, E. (1991). Neuropathological staging of Alzheimer-relatedchanges. Acta Neuropathologica (Berl), 82, 239–259.

abeza, R., & Nyberg, L. (2000). Imaging cognition II: An empirical review of 275 PETand fMRI studies. Journal of Cognitive Neuroscience, 12, 1–47.

elone, K. A., Calhoun, V. D., Dickerson, B. C., Atri, A., Chua, E. F., Miller, S. L.,et al. (2006). Alterations in memory networks in mild cognitive impairmentand Alzheimer’s disease: An independent component analysis. Journal of Neuro-science, 26, 10222–10231.

hetelat, G., Desgranges, B., de la Sayette, V., Viader, F., Berkouk, K., Landeau, B., et al.(2003). Dissociating atrophy and hypometabolism impact on episodic memoryin mild cognitive impairment. Brain, 126, 1955–1967.

hoo, I. H., Lee, D. Y., Oh, J. S., Lee, J. S., Lee, D. S., Song, I. C., et al. (2010). Posteriorcingulate cortex atrophy and regional cingulum disruption in mild cognitiveimpairment and Alzheimer’s disease. Neurobiology of Aging, 31, 772–779.

annhauser, T. M., Shergill, S. S., Stevens, T., Lee, L., Seal, M., Walker, R. W., et al.(2008). An fMRI study of verbal episodic memory encoding in amnestic mildcognitive impairment. Cortex, 44, 869–880.

e Jager, C., Lesk, V., Zhi, X., Marsico, M., & Chandler, J. (2008). Episodic memorytest constructs affect discrimination among health elderly and cases with mildcognitive impairment and Alzheimer’s disease. In International conference onAlzheimer’s disease Chicago,

ickerson, B. C., Salat, D. H., Bates, J. F., Atiya, M., Killiany, R. J., Greve, D. N., et al.(2004). Medial temporal lobe function and structure in mild cognitive impair-ment. Annals of Neurology, 56, 27–35.

ickerson, B. C., Salat, D. H., Greve, D. N., Chua, E. F., Rand-Giovannetti, E., Rentz, D.

M., et al. (2005). Increased hippocampal activation in mild cognitive impairmentcompared to normal aging and AD. Neurology, 65, 404–411.

ickerson, B. C., & Sperling, R. A. (2008). Functional abnormalities of the medialtemporal lobe memory system in mild cognitive impairment and Alzheimer’sdisease: Insights from functional MRI studies. Neuropsychologia, 46, 1624–1635.

gia 49 (2011) 2060–2070 2069

Dubois, B., Feldman, H. H., Jacova, C., Dekosky, S. T., Barberger-Gateau, P., Cum-mings, J., et al. (2007). Research criteria for the diagnosis of Alzheimer’s disease:Revising the NINCDS-ADRDA criteria. Lancet Neurology, 6, 734–746.

Fein, G., Landman, B., Tran, H., Barakos, J., Moon, K., Di Sclafani, V., et al. (2006).Statistical parametric mapping of brain morphology: Sensitivity is dramat-ically increased by using brain-extracted images as inputs. Neuroimage, 30,1187–1195.

Fletcher, P. C., & Henson, R. N. (2001). Frontal lobes and human memory: Insightsfrom functional neuroimaging. Brain, 124, 849–881.

Folstein, M. F., Folstein, S. E., & McHugh, P. R. (1975). Mini-mental state. A practicalmethod for grading the cognitive state of patients for the clinician. Journal ofPsychiatric Research, 12, 189–198.

Fowler, K. S., Saling, M. M., Conway, E. L., Semple, J. M., & Louis, W. J. (1995). Com-puterized delayed matching to sample and paired associate performance in theearly detection of dementia. Applied Neuropsychology, 2, 72–78.

Fowler, K. S., Saling, M. M., Conway, E. L., Semple, J. M., & Louis, W. J. (1997). Comput-erized neuropsychological tests in the early detection of dementia: Prospectivefindings. Journal of the International Neuropsychological Society, 3, 139–146.

Fowler, K. S., Saling, M. M., Conway, E. L., Semple, J. M., & Louis, W. J. (2002). Pairedassociate performance in the early detection of DAT. Journal of the InternationalNeuropsychological Society, 8, 58–71.

Fox, N. C., Warrington, E. K., & Rossor, M. N. (1999). Serial magnetic resonanceimaging of cerebral atrophy in preclinical Alzheimer’s disease. Lancet, 353, 2125.

Friston, K. J., Holmes, A. P., Worsley, K. J., Poline, J. B., & Frackowiak, R. S. (1995).Statistical parametric maps in functional imaging: A general linear approach.Human Brain Mapping, 2, 189–210.

Geerts, H. (2007). Drug evaluation: (R)-flurbiprofen – an enantiomer of flurbiprofenfor the treatment of Alzheimer’s disease. IDrugs, 10, 121–133.

Genovese, C. R., Lazar, N. A., & Nichols, T. (2002). Thresholding of statistical mapsin functional neuroimaging using the false discovery rate. Neuroimage, 15,870–878.

Gould, R. L., Arroyo, B., Brown, R. G., Owen, A. M., Bullmore, E. T., & Howard, R.J. (2006). Brain mechanisms of successful compensation during learning inAlzheimer disease. Neurology, 67, 1011–1017.

Gould, R. L., Brown, R. G., Owen, A. M., Bullmore, E. T., & Howard, R. J. (2006). Task-induced deactivations during successful paired associates learning: An effect ofage but not Alzheimer’s disease. Neuroimage, 31, 818–831.

Gould, R. L., Brown, R. G., Owen, A. M., Bullmore, E. T., Williams, S. C., & Howard,R. J. (2005). Functional neuroanatomy of successful paired associate learning inAlzheimer’s disease. American Journal of Psychiatry, 162, 2049–2060.

Gould, R. L., Brown, R. G., Owen, A. M., Ffytche, D. H., & Howard, R. J. (2003). fMRIBOLD response to increasing task difficulty during successful paired associateslearning. Neuroimage, 20, 1006–1019.

Greicius, M. D., Krasnow, B., Reiss, A. L., & Menon, V. (2003). Functional connectivityin the resting brain: A network analysis of the default mode hypothesis. Pro-ceedings of the National Academy of Sciences of the United States of America, 100,253–258.

Greig, N. H., Sambamurti, K., Yu, Q.-S., Brossi, A., Bruinsma, G. B., & Lahiri, D. K. (2005).An overview of phenserine tartrate, a novel acetylcholinesterase inhibitor for thetreatment of Alzheimer’s disease. Current Alzheimer Research, 2, 281–290.

Gusnard, D. A., Raichle, M. E., & Raichle, M. E. (2001). Searching for a baseline: Func-tional imaging and the resting human brain. Nature Reviews Neuroscience, 2,685–694.

Hamalainen, A., Pihlajamaki, M., Tanila, H., Hanninen, T., Niskanen, E., Tervo, S., et al.(2006). Increased fMRI responses during encoding in mild cognitive impairment.Neurobiology of Aging, 28(12), 1889–1903.

Hayasaka, S., & Nichols, T. E. (2003). Validating cluster size inference: Random fieldand permutation methods. Neuroimage, 20, 2343–2356.

Jack, C. R., Jr., Lowe, V. J., Senjem, M. L., Weigand, S. D., Kemp, B. J., Shiung, M. M.,et al. (2008). 11C PiB and structural MRI provide complementary information inimaging of Alzheimer’s disease and amnestic mild cognitive impairment. Brain,131, 665–680.

Jack, C. R., Jr., Lowe, V. J., Weigand, S. D., Wiste, H. J., Senjem, M. L., Knopman, D. S., et al.(2009). Serial PIB and MRI in normal, mild cognitive impairment and Alzheimer’sdisease: Implications for sequence of pathological events in Alzheimer’s disease.Brain, 132, 1355–1365.

Johnson, S. C., Baxter, L. C., Susskind-Wilder, L., Connor, D. J., Sabbagh, M. N., & Caselli,R. J. (2004). Hippocampal adaptation to face repetition in healthy elderly andmild cognitive impairment. Neuropsychologia, 42, 980–989.

Jones, E. G., & Powell, T. P. (1970). An anatomical study of converging sensory path-ways within the cerebral cortex of the monkey. Brain, 93, 793–820.

Lee, A. C., Robbins, T. W., & Owen, A. M. (2000). Episodic memory meets workingmemory in the frontal lobe: Functional neuroimaging studies of encoding andretrieval. Critical Reviews in Neurobiology, 14, 165–197.

Machulda, M. M., Ward, H. A., Borowski, B., Gunter, J. L., Cha, R. H., O’Brien, P. C.,et al. (2003). Comparison of memory fMRI response among normal, MCI, andAlzheimer’s patients. Neurology, 61, 500–506.

Maguire, E. A., Frith, C. D., Burgess, N., Donnett, J. G., & O’Keefe, J. (1998). Knowingwhere things are parahippocampal involvement in encoding object locations invirtual large-scale space. Journal of Cognitive Neuroscience, 10, 61–76.

Maldjian, J. A., Laurienti, P. J., Kraft, R. A., & Burdette, J. H. (2003). An automated

method for neuroanatomic and cytoarchitectonic atlas-based interrogation offMRI data sets. Neuroimage, 19, 1233–1239.

Masters, C. L., & Beyreuther, K. (2006). Alzheimer’s centennial legacy: Prospects forrational therapeutic intervention targeting the Abeta amyloid pathway. Brain,129, 2823–2839.

2 sychol

M

M

M

M

M

N

N

O

O

O

O

P

P

P

P

P

P

P

P

R

R

070 M. de Rover et al. / Neurop

azoyer, B., Zago, L., Mellet, E., Bricogne, S., Etard, O., Houde, O., et al. (2001). Corticalnetworks for working memory and executive functions sustain the consciousresting state in man. Brain Research Bulletin, 54, 287–298.

cKiernan, K. A., Kaufman, J. N., Kucera-Thompson, J., & Binder, J. R. (2003). A para-metric manipulation of factors affecting task-induced deactivation in functionalneuroimaging. Journal of Cognitive Neuroscience, 15, 394–408.

inoshima, S., Giordani, B., Berent, S., Frey, K. A., Foster, N. L., & Kuhl, D. E. (1997).Metabolic reduction in the posterior cingulate cortex in very early Alzheimer’sdisease. Annals of Neurology, 42, 85–94.

ioshi, E., Dawson, K., Mitchell, J., Arnold, R., & Hodges, J. R. (2006). TheAddenbrooke’s Cognitive Examination Revised (ACE-R): A brief cognitive testbattery for dementia screening. International Journal of Geriatric Psychiatry, 21,1078–1085.

orris, J. C. (1993). The Clinical Dementia Rating (CDR): Current version and scoringrules. Neurology, 43, 2412–2414.

agahama, Y., Okina, T., Suzuki, N., & Matsuda, M. (2006). The Cambridge BehavioralInventory: Validation and application in a memory clinic. Journal of GeriatricPsychiatry and Neurology, 19, 220–225.

estor, P. J., Fryer, T. D., Smielewski, P., & Hodges, J. R. (2003). Limbichypometabolism in Alzheimer’s disease and mild cognitive impairment. Annalsof Neurology, 54, 343–351.

’Brien, J. L., O’Keefe, K. M., LaViolette, P. S., DeLuca, A. N., Blacker, D., Dickerson, B. C.,et al. (2010). Longitudinal fMRI in elderly reveals loss of hippocampal activationwith clinical decline. Neurology, 74, 1969–1976.

’Keefe, J., & Nadel, L. (1978). The hippocampus as a cognitive map. Oxford: TheClaredon Press. (chapter 2).

wen, A. M., Miller, B., Petrides, M., & Evans, A. C. (1996). A specific role for the rightparahippocampal gyrus in the retrieval of object-location: A positron emissiontomography study. Journal of Cognitive Neuroscience, 12, 163–173.

wen, A. M., Sahakian, B. J., Semple, J., Polkey, C. E., & Robbins, T. W. (1995).Visuo-spatial short-term recognition memory and learning after temporal lobeexcisions, frontal lobe excisions or amygdalo-hippocampectomy in man. Neu-ropsychologia, 33, 1–24.

arkinson, J. K., Murray, E. A., & Mishkin, M. (1988). A selective mnemonic role forthe hippocampus in monkeys: Memory for the location of objects. Journal ofNeuroscience, 8, 4159–4167.

engas, G., Hodges, J. R., Watson, P., & Nestor, P. J. (2010). Focal posterior cingulateatrophy in incipient Alzheimer’s disease. Neurobiology of Aging, 31, 25–33.

engas, G., Pereira, J. M., Williams, G. B., & Nestor, P. J. (2009). Comparative reli-ability of total intracranial volume estimation methods and the influence ofatrophy in a longitudinal semantic dementia cohort. Journal of Neuroimaging, 19,37–46.

ennanen, C., Testa, C., Laakso, M. P., Hallikainen, M., Helkala, E. L., Hanninen, T.,et al. (2005). A voxel based morphometry study on mild cognitive impairment.Journal of Neurology, Neurosurgery and Psychiatry, 76, 11–14.

ereira, J. M., Xiong, L., Acosta-Cabronero, J., Pengas, G., Williams, G. B., & Nestor, P.J. (2010). Registration accuracy for VBM studies varies according to region anddegenerative disease grouping. Neuroimage, 49, 2205–2215.

etersen, R. C. (2004). Mild cognitive impairment as a diagnostic entity. Journal ofInternal Medicine, 256, 183–194.

etersen, R. C., Smith, G. E., Waring, S. C., Ivnik, R. J., Tangalos, E. G., & Kokmen,E. (1999). Mild cognitive impairment: Clinical characterization and outcome.Archives of Neurology, 56, 303–308.

etrella, J. R., Krishnan, S., Slavin, M. J., Tran, T. T., Murty, L., & Doraiswamy, P. M.(2006). Mild cognitive impairment: Evaluation with 4-T functional MR imaging.Radiology, 240, 177–186.

aichle, M. E., MacLeod, A. M., Snyder, A. Z., Powers, W. J., Gusnard, D. A., & Shulman,G. L. (2001). A default mode of brain function. Proceedings of the National Academyof Sciences of the United States of America, 98, 676–682.

owe, C. C., Ng, S., Ackermann, U., Gong, S. J., Pike, K., Savage, G., et al. (2007). Imagingbeta-amyloid burden in aging and dementia. Neurology, 68, 1718–1725.

ogia 49 (2011) 2060–2070

Sahakian, B. J., Morris, R. G., Evenden, J. L., Heald, A., Levy, R., Philpot, M., et al. (1988).A comparative study of visuospatial memory and learning in Alzheimer-typedementia and Parkinson’s disease. Brain, 111(Pt 3), 695–718.

Schmand, B., Huizenga, H. M., & van Gool, W. A. (2010). Meta-analysis of CSF and MRIbiomarkers for detecting preclinical Alzheimer’s disease. Psychological Medicine,40, 135–145.

Segonne, F., Dale, A. M., Busa, E., Glessner, M., Salat, D., Hahn, H. K., et al. (2004).A hybrid approach to the skull stripping problem in MRI. Neuroimage, 22,1060–1075.

Sled, J. G., Zijdenbos, A. P., & Evans, A. C. (1998). A nonparametric method for auto-matic correction of intensity nonuniformity in MRI data. IEEE Transactions ofMedical Imaging, 17, 87–97.

Small, S. A., Perera, G. M., DeLaPaz, R., Mayeux, R., & Stern, Y. (1999). Differentialregional dysfunction of the hippocampal formation among elderly with memorydecline and Alzheimer’s disease. Annals of Neurology, 45, 466–472.

Smith, M. L., & Milner, B. (1981). The role of the right hippocampus in the recall ofspatial location. Neuropsychologia, 19, 781–793.

Smith, S. M. (2002). Fast robust automated brain extraction. Human Brain Mapping,17, 143–155.

Stark, C. (2007). Functinal role of the human hippocampus. In P. Andersen, R. Morris,D. Amaral, T. Bliss, & J. O’Keefe (Eds.), The hippocampus book (pp. 549–579). NewYork: Oxford Univerity Press.

Svoboda, E., McKinnon, M. C., & Levine, B. (2006). The functional neuroanatomy ofautobiographical memory: A meta-analysis. Neuropsychologia, 44, 2189–2208.

Swainson, R., Hodges, J. R., Galton, C. J., Semple, J., Michael, A., Dunn, B. D., et al. (2001).Early detection and differential diagnosis of Alzheimer’s disease and depressionwith neuropsychological tasks. Dementia and Geriatric Cognitive Disorders, 12,265–280.

Taffe, M. A., Weed, M. R., Gutierrez, T., Davis, S. A., & Gold, L. H. (2002). Differentialmuscarinic and NMDA contributions to visuo-spatial paired-associate learningin rhesus monkeys. Psychopharmacology (Berl), 160, 253–262.

Talpos, J. C., Winters, B. D., Dias, R., Saksida, L. M., & Bussey, T. J. (2009). A noveltouchscreen-automated paired-associate learning (PAL) task sensitive to phar-macological manipulation of the hippocampus: A translational rodent modelof cognitive impairments in neurodegenerative disease. Psychopharmacology(Berl), 205, 157–168.

Trivedi, M. A., Murphy, C. M., Goetz, C., Shah, R. C., Gabrieli, J. D., Whitfield-Gabrieli,S., et al. (2008). fMRI activation changes during successful episodic memoryencoding and recognition in amnestic mild cognitive impairment relative tocognitively healthy older adults. Dementia and Geriatric Cognitive Disorders, 26,123–137.

Tulving, E., Kapur, S., Markowitsch, H. J., Craik, F. I., Habib, R., & Houle, S. (1994).Neuroanatomical correlates of retrieval in episodic memory: Auditory sentencerecognition. Proceedings of the National Academy of Sciences of the United Statesof America, 91, 2012–2015.

Vellas, B., Andrieu, S., Sampaio, C., & Wilcock, G. (2007). Disease-modifying trialsin Alzheimer’s disease: A European task force consensus. Lancet Neurology, 6,56–62.

Weed, M. R., Taffe, M. A., Polis, I., Roberts, A. C., Robbins, T. W., Koob, G. F., et al. (1999).Performance norms for a rhesus monkey neuropsychological testing battery:Acquisition and long-term performance. Cognitive Brain Research, 8, 185–201.

Whitwell, J. L., Przybelski, S. A., Weigand, S. D., Knopman, D. S., Boeve, B. F., Petersen,R. C., et al. (2007). 3D maps from multiple MRI illustrate changing atrophypatterns as subjects progress from mild cognitive impairment to Alzheimer’sdisease. Brain, 130, 1777–1786.

Worsley, K. J., Marrett, S., Neelin, P., Vandal, A. C., Friston, K. J., & Evans, A. C. (1996).

A unified statistical approach for determining significant signals in images ofcerebral activation. Human Brain Mapping, 4, 58–73.

Ziolko, S. K., Weissfeld, L. A., Klunk, W. E., Mathis, C. A., Hoge, J. A., Lopresti, B. J., et al.(2006). Evaluation of voxel-based methods for the statistical analysis of PIB PETamyloid imaging studies in Alzheimer’s disease. Neuroimage, 33, 94–102.