Investigating virtual reality navigation in amnestic mild cognitive impairment using fMRI

This article was downloaded by: [Colin Dourish]On: 03 August 2015, At: 03:02Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954 Registeredoffice: 5 Howick Place, London, SW1P 1WG

Click for updates

Aging, Neuropsychology, and Cognition:A Journal on Normal and DysfunctionalDevelopmentPublication details, including instructions for authors andsubscription information:http://www.tandfonline.com/loi/nanc20

Investigating virtual reality navigationin amnestic mild cognitive impairmentusing fMRIE.M. Migoa, O. O’Dalya, M. Mitterschiffthalerab, E. Antonovaa, G.R.Dawsonc, C.T. Dourishc, K.J. Craigc, A. Simmonsade, G.K. Wilcockf,E. McCullochf, S.H.D. Jacksong, M.D. Kopelmana, S.C.R. Williamsa

& R.G. Morrisa

a King’s College London, Institute of Psychiatry, Psychology andNeuroscience, London, UKb Department for Psychotherapy and Psychosomatics, CampusInnenstadt, Ludwig-Maximilians-University, Munich, Germanyc P1vital, Wallingford, UKd NIHR Biomedical Research Centre for Mental Health at SouthLondon and Maudsley NHS Foundation Trust and Institute ofPsychiatry, Psychology and Neuroscience, King’s College London,London, UKe NIHR Biomedical Research Unit for Dementia at South Londonand Maudsley NHS Foundation Trust and Institute of Psychiatry,Psychology and Neuroscience, King’s College London, London, UKf Nuffield Department of Clinical Neurosciences, University ofOxford, John Radcliffe Hospital, Oxford, UKg Clinical Age Research Unit, King’s College Hospital, London, UKPublished online: 03 Aug 2015.

To cite this article: E.M. Migo, O. O’Daly, M. Mitterschiffthaler, E. Antonova, G.R. Dawson, C.T.Dourish, K.J. Craig, A. Simmons, G.K. Wilcock, E. McCulloch, S.H.D. Jackson, M.D. Kopelman,S.C.R. Williams & R.G. Morris (2015): Investigating virtual reality navigation in amnestic mildcognitive impairment using fMRI, Aging, Neuropsychology, and Cognition: A Journal on Normal andDysfunctional Development, DOI: 10.1080/13825585.2015.1073218

To link to this article: http://dx.doi.org/10.1080/13825585.2015.1073218

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the“Content”) contained in the publications on our platform. However, Taylor & Francis,our agents, and our licensors make no representations or warranties whatsoever as tothe accuracy, completeness, or suitability for any purpose of the Content. Any opinionsand views expressed in this publication are the opinions and views of the authors,and are not the views of or endorsed by Taylor & Francis. The accuracy of the Contentshould not be relied upon and should be independently verified with primary sourcesof information. Taylor and Francis shall not be liable for any losses, actions, claims,proceedings, demands, costs, expenses, damages, and other liabilities whatsoever orhowsoever caused arising directly or indirectly in connection with, in relation to or arisingout of the use of the Content.

This article may be used for research, teaching, and private study purposes. Anysubstantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions

Dow

nloa

ded

by [

Col

in D

ouri

sh]

at 0

3:02

03

Aug

ust 2

015

Investigating virtual reality navigation in amnestic mildcognitive impairment using fMRIE.M. Migoa, O. O’Dalya, M. Mitterschiffthalera,b, E. Antonovaa, G.R. Dawsonc,C.T. Dourishc, K.J. Craigc, A. Simmonsa,d,e, G.K. Wilcockf, E. McCullochf, S.H.D. Jacksong,M.D. Kopelmana, S.C.R. Williamsa and R.G. Morrisa

aKing’s College London, Institute of Psychiatry, Psychology and Neuroscience, London, UK; bDepartmentfor Psychotherapy and Psychosomatics, Campus Innenstadt, Ludwig-Maximilians-University, Munich,Germany; cP1vital, Wallingford, UK; dNIHR Biomedical Research Centre for Mental Health at South Londonand Maudsley NHS Foundation Trust and Institute of Psychiatry, Psychology and Neuroscience, King’sCollege London, London, UK; eNIHR Biomedical Research Unit for Dementia at South London andMaudsley NHS Foundation Trust and Institute of Psychiatry, Psychology and Neuroscience, King’s CollegeLondon, London, UK; fNuffield Department of Clinical Neurosciences, University of Oxford, John RadcliffeHospital, Oxford, UK; gClinical Age Research Unit, King’s College Hospital, London, UK

ABSTRACTSpatial navigation requires a well-established network of brainregions, including the hippocampus, caudate nucleus, and retro-splenial cortex. Amnestic Mild Cognitive Impairment (aMCI) is acondition with predominantly memory impairment, conferring ahigh predictive risk factor for dementia. aMCI is associated withhippocampal atrophy and subtle deficits in spatial navigation. Wepresent the first use of a functional Magnetic Resonance Imaging(fMRI) navigation task in aMCI, using a virtual reality analog of theRadial Arm Maze. Compared with controls, aMCI patients showedreduced activity in the hippocampus bilaterally, retrosplenial cor-tex, and left dorsolateral prefrontal cortex. Reduced activation inkey areas for successful navigation, as well as additional regions,was found alongside relatively normal task performance. Resultsalso revealed increased activity in the right dorsolateral prefrontalcortex in aMCI patients, which may reflect compensation forreduced activations elsewhere. These data support suggestionsthat fMRI spatial navigation tasks may be useful for staging ofprogression in MCI.

ARTICLE HISTORYReceived 14 April 2015Accepted 11 July 2015

KEYWORDSMild cognitive impairment;fMRI, navigation; spatialmemory; hippocampus

Introduction

Spatial navigation is the mechanism by which we move around an environment,determining a route between places and moving along it. It is complex, requiringperceptual identification and use of both self-centered (egocentric) and non-self-cen-tered (allocentric) frames of reference. Much of our understanding of the neurobiologyof immersive spatial navigation has come from rodent work in artificial mazes, such asthe Morris Swim Maze (Morris, Garrud, Rawlins, & O’Keefe, 1982), Olton Radial Arm

CONTACT: E.M. Migo [email protected]

AGING, NEUROPSYCHOLOGY, AND COGNITION, 2015http://dx.doi.org/10.1080/13825585.2015.1073218

© 2015 Taylor & Francis

Dow

nloa

ded

by [

Col

in D

ouri

sh]

at 0

3:02

03

Aug

ust 2

015

Maze (Olton & Samuelson, 1976), and cross mazes (O’Keefe & Conway, 1980). This hasshown the critical role of the hippocampus, especially for allocentric processing(Devan, Goad, & Petri, 1996), as well as non-medial temporal structures such as thecaudate nucleus (Devan et al., 1996) and retrosplenial cortex (Vann, Aggleton, &Maguire, 2009). More recently, the role of the prefrontal cortex in navigation hasbeen highlighted in terms of the problem-solving component associated with spatialnavigation (Spiers & Gilbert, 2015).

Examining spatial navigation in humans has required innovative real-world tests,including the use of virtual reality (see Bohil, Alicea, & Biocca, 2011 for a review). Thisallows navigational tasks to be performed in an imaging environment, where functionalactivations are found in similar key regions to those identified in animal work. Across avariety of tasks, hippocampal activation is reliably seen in young healthy participants(Antonova et al., 2009; Baumann, Chan, & Mattingley, 2012; Iaria, Chen, Guariglia, Ptito, &Petrides, 2007; Iaria, Petrides, Dagher, Pike, & Bohbot, 2003; Maguire et al., 1998; Marshet al., 2010; Parslow et al., 2004; Shipman & Astur, 2008). A number of studies haveshown a right lateralized effect across multiple imaging modalities, including functionalimaging tasks (Hartley, Maguire, Spiers, & Burgess, 2003; Iaria et al., 2003; Maguire et al.,1998; Wolbers, Wiener, Mallot, & Büchel, 2007) and when looking at associationsbetween right hippocampal structure and navigation performance (Bohbot, Lerch,Thorndycraft, Iaria, & Zijdenbos, 2007; Iaria, Lanyon, Fox, Giaschi, & Barton, 2008;Schinazi, Nardi, Newcombe, Shipley, & Epstein, 2013; Wegman et al., 2014).

Although the role of the hippocampus in spatial memory is clearly established,navigation tasks also activate a consistent range of other brain regions, including theparahippocampal place area and retrosplenial cortex (see Epstein, 2008 for a review),as well as the precuneus (Cavanna & Trimble, 2006). The retrosplenial cortex mayhave a specific role processing landmarks and permanent features (e.g., Auger,Mullally, & Maguire, 2012), whereas the parahippocampal place area appears to beinvolved in encoding the local scene for later memory (Epstein, 2008). The precuneusappears to be important for episodic retrieval as well as visuospatial imagery(Cavanna & Trimble, 2006).

In line with animal work, functional Magnetic Resonance Imaging (fMRI) activityduring spatial navigation tasks is found in the caudate nucleus when participantsuse egocentric or response strategies (Hartley et al., 2003; Iaria et al., 2003), asopposed to allocentric strategies. Caudate gray matter density is significantlypositively correlated with participants spontaneously choosing to use a responselearning strategy in an Olton Radial Arm Maze task (Bohbot et al., 2007) and,showing the opposite pattern to the hippocampus, negatively correlated withassessments of allocentric learning in a real-world navigation task (Schinazi et al.,2013). This interaction between hippocampal- and caudate-mediated spatial strate-gies has also been highlighted in a study using a virtual reality Morris Water Mazeprocedure, the Arena task, where a pharmacological challenge (scopolamine) wasused to disrupt normal hippocampal function (Antonova et al., 2011).Administration of scopolamine compared with placebo led to significant compen-satory recruitment of the caudate during the task. Most spatial tasks will require theuse of both egocentric and allocentric representations, although this will be influ-enced by participants' strategies and can change with practice on the task (Iaria

2 E.M. MIGO ET AL.

Dow

nloa

ded

by [

Col

in D

ouri

sh]

at 0

3:02

03

Aug

ust 2

015

et al., 2003). Another interpretation of caudate activity relates it to learning aboutlandmarks within navigation (Doeller, King, & Burgess, 2008).

Amnestic Mild Cognitive Impairment (aMCI) is a condition characterized by arelatively selective decline in memory over and above what is expected in normalaging (Petersen, 2004). Having the condition is widely considered to be a highlypredictive risk factor for the development of Alzheimer's Disease (AD; Morris et al.,2001; Petersen, 2004) and there is therefore diverse research into identifying aMCIpatients, understanding their cognitive impairments, and predicting later conversionto dementia (Blennow, Hampel, Weiner, & Zetterberg, 2010; Prvulovic, Bokde,Faltraco, & Hampel, 2011; Sperling, 2011; Wolk & Detre, 2012). There has been arecent emphasis on investigating spatial navigation in normal and pathological aging(Gazova et al., 2012; Iachini, Iavarone, Senese, Ruotolo, & Ruggiero, 2009; Lithfous,Dufour, & Després, 2013; Moffat, 2009). Spatial navigation impairments occur early inthe progress of AD, where patients can be disoriented even in familiar surroundings(Henderson, Mack, & Williams, 1989; Monacelli, Cushman, Kavcic, & Duffy, 2003).Hippocampal atrophy is well established in aMCI (Shi, Liu, Zhou, Yu, & Jiang, 2009;Yang et al., 2012), making tests that are sensitive to hippocampal damage/dysfunc-tion appropriate for investigation, and a number of behavioral studies have investi-gated navigation performance in these patients.

Work with a real-world navigation task based on the Morris Water Maze has shownthat aMCI patients can be impaired, especially when an allocentric strategy is required(Hort et al., 2007; Laczó et al., 2010, 2009). When using navigation tasks in a functionalimaging setting, the use of virtual reality tasks is critical. Such tasks have been validatedin studies demonstrating highly correlated performance between real-world and virtualnavigation within an MCI population (Cushman, Stein, & Duffy, 2008; Nedelska et al.,2012). Using virtual tasks, work has shown impairments in aMCI on route learning, butnot landmark recognition (DeIpolyi, Rankin, Mucke, Miller, & Gorno-Tempini, 2007), andin navigating through a virtual park and maze (Weniger, Ruhleder, Lange, Wolf, & Irle,2011). However, these behavioral impairments in navigation in aMCI are not universal(Cushman et al., 2008) and where multiple patient groups have been studied, single-domain and purely aMCI patients show much milder impairments than the multi-domain MCI groups (Hort et al., 2007). These performance differences are thereforemild and often do not reach statistical significance.

Some studies have used structural imaging to look for associations between naviga-tion performance and regional brain volume in aMCI patients. When combining MCI, AD,and healthy controls, correlations have been seen between performance and the righthippocampus (DeIpolyi et al., 2007; Nedelska et al., 2012). Other work has shown anassociation between precuneus volume and performance within aMCI (Weniger et al.,2011). Using a broader MCI group, associations between navigation performance onobjective tasks and subjective questionnaires have been seen in a range of brain regionsincluding the middle and medial frontal gyrus, superior temporal gyrus, and cuneus(Mitolo et al., 2013).

Here we investigate spatial navigation in aMCI using fMRI for the first time.Virtual reality navigation tasks are highly suitable for use in an MRI environment,and the widespread use of navigation tasks in healthy controls has identifiedregions that are reliably activated. We present data from a group of aMCI patients

AGING, NEUROPSYCHOLOGY, AND COGNITION 3

Dow

nloa

ded

by [

Col

in D

ouri

sh]

at 0

3:02

03

Aug

ust 2

015

and matched controls on a virtual reality Olton Radial Arm Maze analog. Given thevariation in the literature on existing navigation task performance in aMCI patients,we expected only relatively mild performance impairments, if any, in these patientscompared with the control group. Much work using fMRI in MCI has been focussedon episodic and working memory tasks (Li et al., 2015), with a few studies onvisuospatial processing and none using a navigation task on which to base anypredictions. We expected that the control group would show blood-oxygen-leveldependent (BOLD) task-related activations in areas known to be important fornavigation, such as the hippocampus, retrosplenial cortex, and caudate nucleus.We predicted decreases in BOLD activity in the aMCI patients in regions affected byearly dementia processing, most notably in the hippocampus.

Materials and methods

Participants

The patient group was composed of eight patients (5 male) with aMCI, recruitedfrom specialized memory clinics. All patients met the criteria for aMCI as defined inPetersen et al. (2001), including subjective memory complaints, objective memoryimpairments, normal general cognitive function, and intact activities of daily living.All scored 0.5 on the Clinical Dementia Rating Scale, as measured during this study(CDR; J. C. Morris, 1993), indicating that none had advanced to dementia. A controlgroup of 10 healthy volunteers (6 male), age and IQ matched to the patients, wererecruited from existing databases of research volunteers or in response to locallyplaced adverts.

Participants were aged between 61 and 80 years of age and were all native Englishspeakers. For both groups, a full medical history and medical examination was carriedout by a clinician employed for the study, with blood tests and urinalysis, to ensure thatall participants were healthy with no history of head injury, alcoholism, or any psychiatricor neurological condition (other than memory problems in the aMCI group). No parti-cipants were excluded based on their medical history or examination; no alternativeconditions to explain memory problems in the aMCI group were indicated. Informedwritten consent was obtained from all participants and the study was approved by theNational Hospital for Neurology and Neurosurgery and Institute of Neurology ResearchEthics Committee.

Apparatus

The fMRI paradigm software was programmed by Third Dimension (Dorset, UK)using Superscape Virtual Reality software (Superscape, Hampshire, UK). The experi-ment was presented on a 450 MHz microprocessor fitted with an 8 MB, three-dimensional graphics card. The image was displayed via a projector onto a Perspexscreen at the foot of the scanning table. To navigate in the virtual reality task,participants used a trackerball. Participants who had corrected vision wore contactlenses or used MR-compatible vision-correcting goggles, to ensure they couldaccurately see the screen.

4 E.M. MIGO ET AL.

Dow

nloa

ded

by [

Col

in D

ouri

sh]

at 0

3:02

03

Aug

ust 2

015

Procedure

Participants completed a screening visit and a scanning visit. On the screening visit,they completed a neuropsychological assessment and had a medical examination,with full medical history taken. The neuropsychological assessment covered tests ofmemory, executive function, and intelligence. The neuropsychological tests usedwere the Wechsler Abbreviated Scale of Intelligence (WASI) two subtest form(Wechsler, 1999), the California Verbal Learning Test (CVLT; Delis, Kramer, Kaplan,& Ober, 2000), the Logical Memory and Visual Reproduction Subtests of theWechsler Memory Scale-III (WMS-III; Wechsler, 1998), the Hayling and BrixtonTasks (Burgess & Shallice, 1997), and the National Adult Reading Test (NART;Nelson & Willison, 1991). Participants also underwent training on the platformtask (see below). On the scanning visit, the vital signs for participants were checkedand an alcohol breath test was administered. For the aMCI group, the CDR wasadministered and all participants completed the Trail Making Task (TMT; Delis,Kaplan, & Kramer, 2001). All participants completed a computerized N-back taskas an fMRI study; the behavioral data reported here indicate working memoryperformance, while fMRI data are reported elsewhere (Migo et al., 2015).

Task

Participants moved in a virtual circular arena containing a randomly arranged array ofcircular platforms positioned on the floor. Surrounding the arena were visual cues tohelp the participants navigate around the central space using a trackerball; see Figure 1for a plan view of the arena with four platforms and an example of the participants' eyeview during the task. The visual cues were buildings, people, vehicles, and trees.Participants could select platforms using a cursor on the screen using a trackerball tonavigate and select platforms.

The task was to visit each platform in turn, without going back to the sameplatform twice. Once a platform is selected, the participant is automatically movedto stand on it and turned to face the center of the arena. Other platforms are nowvisible from a different perspective, requiring an allocentric strategy linking plat-forms with landmarks. To avoid the use of simple egocentric strategies (e.g., always

Figure 1. Platform task. A: Plan view of arena space with landmarks: 1, wind turbine; 2, cottage withcar outside; 3, stone formation; 4, group of three houses; 5, statue; 6, castle; 7, fire engine; 8, buswith passengers; 9, tower block; 10, church. B: Participant’s view of task with two platforms, one ofwhich can be selected. The statue, castle, and fire engine landmarks can be seen, as well as the crosshair controlled by participants with the trackerball. C: Visual control image.

AGING, NEUROPSYCHOLOGY, AND COGNITION 5

Dow

nloa

ded

by [

Col

in D

ouri

sh]

at 0

3:02

03

Aug

ust 2

015

turning right), not every platform could be selected on any given choice. Platformscould only be selected when they were yellow; they could not be selected whenthey were red. This color change was independent of whether or not the platformshad been selected before (although there was always at least one correct availableplatform per choice). Participants received feedback in the form of a green tick orred cross after each choice. The task was self-paced and continued until all plat-forms had been selected. Before and after each trial there were 10 s periods ofblank screen (rest) and 10 s of a visual control image, which participants passivelyviewed (see Figure 1). Task difficulty could be manipulated by changing thenumber of platforms on each trial. Here, we used three trials of four platformsand three trials of six platforms in a fixed order for all participants.

Participants were trained on the task on a separate visit where the neuropsychologi-cal assessment was completed. After trials introducing the trackerball and the task, theycompleted a full practice run of the experiment, comprising three trials at four platformsand three trials at six platforms. On the day of the scan, each participant was reintro-duced to the task and completed another full dry run before entering the scanner.

Image acquisition

Images were acquired using a 3.0-Tesla, GE Signa HDx system running 14m4 software(General Electric, Milwaukee, WI, USA) at the Institute of Psychiatry, Psychology andNeuroscience Department of Neuroimaging. Image volumes (each consisting of 38 near-axial slices) were collected using a gradient-echo echo planar imaging sequence with arepetition time (TR) of 2000 ms, an echo time (TE) of 30 ms, and a 75° flip angle. Theslices were positioned parallel to the AC-PC line. The body coil was used for radio-frequency (RF) transmission and an 8-channel head coil for RF reception. Each imageslice was acquired using a 64 × 64 image matrix over a 24 cm field of view. The resultingin-plane pixel size of the images was 3.75 mm × 3.75 mm. The image slices had athickness of 3.0 mm with a 0.3 mm gap. Head movement was limited by foam paddingwithin the head coil and a restraining band across the forehead. At the same session a60-slice, high-resolution gradient-echo echo planar sequence was acquired in both thecoronal and axial planes with the same acquisition parameters apart from a 128 × 128matrix, giving 1.875 × 1.875 in-plane resolution. Data quality was assured using anautomated quality control procedure (Simmons, Moore, & Williams, 1999). A high-resolution 3D T1 weighted SPGR image was acquired in the coronal plane, with1.1 mm isotropic voxels using a 256 × 256 × 196 matrix with a TI of 450 ms, TR of7.1 ms, TE of 2.8 ms. and a flip angle of 20°.

Imaging data processing

Functional MRI data were preprocessed and analyzed using SPM8 (WellcomeDepartment of Cognitive Neurology, London, UK; www.fil.ion.ucl.ac.uk/spm). Imageswere preprocessed using the following steps: images were first spatially realigned tothe mean image from the series and then resliced. The mean functional image was co-registered to each subject's T1 structural scan. Spatial normalization into MontrealNeurological Institute (MNI) stereotactic space was carried out using Diffeomorphic

6 E.M. MIGO ET AL.

Dow

nloa

ded

by [

Col

in D

ouri

sh]

at 0

3:02

03

Aug

ust 2

015

Anatomical Registration Through Exponentiated Lie Algebra (DARTEL) (Ashburner, 2007),using a sample template generated from all participants' structural scans. The functionalimages were resampled into 1.5 mm3 voxels and spatially smoothed with an 8 mm full-width half-maximum Gaussian kernel.

At the single-subject level, the data were modeled by splitting responses intopairs of choices (i.e., a regressor encoding choices one and two at four platforms,regressors encoding choices three and four at four platforms, etc.). Error trials,visual control epochs, and a resting baseline condition were modeled separately.In each case the regressor encoded the relevant epochs (i.e., blocks), modeledusing a boxcar, convolved with the canonical hemodynamic response function inSPM. Importantly, for each trial the duration of the modeled epoch was dependentupon subject performance, and therefore varied among epochs. Six regressorsencoding movement parameters (i.e., translations and rotations around the x-, y-,z-axes) created during the realignment stage were included as covariates in all first-level models to control for residual motion-related signal intensity variation. In thesecond-level analysis, age and IQ (WASI) were included as covariates of nointerest to account for variance associated with these factors in both groups. Anumber of separate group-level random effects analyses were carried out to iden-tify task-related activation. Separate one-sample t-tests for patients and controlswere conducted using the first-level contrast of (i) all task-related activity (i.e.,across four- and six-platform trials) versus resting baseline; (ii) activity during thefour-platform trials alone, compared to the resting baseline; (iii) activity during thesix-platform trials alone, compared to the resting baseline; and (iv) to identify loadeffects, the first-level contrast of activity during the four-platform was compared tothe six-platform trials. Finally, a group-level three-way analysis of variance (ANOVA)was created with the following factors: group (patient versus control), tasklevel (four versus six platforms), and trial (first pair of choices versus last pair ofchoices).

The results were considered significant if they survived stringent family-wiseerror correction for multiple comparisons on the basis of cluster extent (p < .05corrected) using a cluster-forming height threshold of p < .001. For region ofinterest (ROI) analysis of the hippocampus/parahippocampal gyrus, we used asmall volume correction limited to an anatomical mask, defined a priori in theWFU Pickatlas (Wake Forest University, Winston-Salem, North Carolina). Statisticalsignificance was defined at the voxel level (p < .05) corrected for multiplecomparisons.

Structural data were processed for voxel-based morphometry (VBM) analysisusing the VBM8 toolbox within SPM8 (http://dbm.neuro.uni-jena.de/vbm/). Defaultsettings in VBM8 were used, which included normalization via DARTEL and seg-mentation into gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF).Total intracranial volume (ICV) was calculated for each participant by summingglobal tissue volume (i.e., GM, WM, and CSF) within VBM8. Images were smoothedwith an 8 mm full-width half-maximum Gaussian kernel. Statistical analysis usedunpaired t-tests in SPM8 to compare GM differences between healthy controls andaMCI patients. Age and ICV were included as variables of no interest. The threshold

AGING, NEUROPSYCHOLOGY, AND COGNITION 7

Dow

nloa

ded

by [

Col

in D

ouri

sh]

at 0

3:02

03

Aug

ust 2

015

of significance for this analysis was set to p < .001 (uncorrected) with an extentthreshold of 200 voxels.

Results

Neuropsychological assessment

The performance of both groups of participants is shown in Table 1. The control andaMCI groups were matched for age, years in education, IQ, and executive function. TheaMCI group were significantly impaired on all measures of memory performance, withno group differences on other measures.

Platform task performance

Both groups performed well at the task. The mean errors per trial (i.e., returning to apreviously visited platform) and total time taken per trial are shown in Table 2. Bothgroups made errors at both levels of task difficulty and although aMCI patientsmade numerically slightly more errors, a two-way repeated measures ANOVA withgroup as a between-subject factor and a number of platforms as a within-subjectfactor showed no significant effect of group (F(1,16) = 1.470, p = .243, f = .302). The

Table 1. Performance on neuropsychological test battery.Control aMCI p

N (male) 10 (6) 8 (5) >.999#

Age 70.3 (6.5) 69.6 (5.8) .822Years in education 16.0 (4.2) 17.0 (4.2) .622

IQ NART 121.3 (6.3) 119.8 (9.3) .679WASI 124.4 (15.9) 116.8 (16.7) .474

Executive function Hayling 6.0 (1.0) 6.0 (0.0) .740*Brixton 4.2 (2.5) 3.8 (2.1) .690Trails B Time 74.0 (34.8) 88.4 (29.2) .365

Memory CVLT T Score 56.0 (10.5) 35.9 (15.1) .004Logical Memory Immediate 11.9 (3.0) 7.6 (4.2) .022Logical Memory Delayed 13.1 (2.1) 9.5 (3.7) .018Visual Reproduction Immediate 12.7 (3.0) 8.9 (4.0) .033Visual Reproduction Delayed 14.9 (2.7) 8.8 (4.2) .002

Attention Trails A Time 39.4 (11.4) 35.8 (10.3) .492

Working memory N-Back 2-back d’ 3.8 (1.0) 3.8 (0.6) .721*N-Back 2-back RT 570.8 (107.0) 594.8 (94.7) .626

Values represent means for each group (SD) except for counts of participants, Hayling test, and N-Back 2-back d’ (Mdnwith IQR presented). p-Values represent independent t-tests except for #Chi squared with Fisher’s exact test and*Mann–Whitney U-test.

Table 2. Platform performance.Control aMCI

Errors per trial 4 Platforms 0.57 (.35) 0.83 (.44)6 Platforms 1.90 (.74) 2.13 (.62)

Time per trial (seconds) 4 Platforms 59.7 (18.4) 84.3 (32.6)6 Platforms 92.5 (19.6) 103.0 (31.8)

Mean errors or trial times are presented with standard deviation in parentheses.

8 E.M. MIGO ET AL.

Dow

nloa

ded

by [

Col

in D

ouri

sh]

at 0

3:02

03

Aug

ust 2

015

main effect of number of platforms was significant, with more errors being made at sixplatforms, (F(1,16) = 58.898, p < .001, f = 1.916), but the interaction of group and numberof platforms was not significant (F(1,16) = .015, p = .905, f = .032). The results from thetime taken per trial showed the same pattern, with a significant main effect of numberof platforms (F(1,16) = 15.399, p = .001, f = .980) and non-significant results for the maineffect of group (F(1,16) = 2.820, p = .107, f = .423) and interaction (F(1,16) = 1.147,p = .300, f = .268).

To investigate more subtle performance differences, errors and reaction time forcorrect choices were broken down within each trial (i.e., choices 1, 2, 3, and 4 at fourplatforms and choices 1–6 at six platforms). For the reaction time data for four platforms,a two-way repeated measures ANOVA was carried out, with group as a between-subjectsfactor and choice as a within-subjects factor; there was no significant main effect ofgroup (F(1,15) = 2.843, p = .112, f = .435), but the main effect of choice was significant (F(3,45) = 3.941, p = .014, f = .512). There was a trend for an interaction between platformchoice and group (F(3,45) = 2.500, p = .072, f = .408) due to the aMCI group being sloweras the number of choices increased. Planned post hoc t-tests showed no significantdifferences between reaction time for the first two choices (smallest p = .392), but trendsfor the last two choices (Choice 3 p = .115, d = .875; Choice 4 p = .058, d = 0.970).Although the results of the ANOVA tests were marginal, the effect sizes indicate strongeffects of the aMCI group being slower on the later trials. This pattern was not present atsix platforms (smallest p = .343 for ANOVA). Errors were compared at each choice withseparate Mann–Whitney U-tests, but no significant differences were seen (smallestuncorrected p = .148). Finally, we investigated any learning effects in reaction timeacross trials (trials 1–3 for both task levels) but none were seen for each group separatelyand there were no group by trial interactions (largest F = 2.371, p = .138 for main effectof trial number at four platforms).

fMRI results

Control groupIn the control group, a large network of brain regions showed increased BOLD responseduring the task compared with rest, including a large cluster with a peak in the posteriorparietal cortex that extended bilaterally into the precuneus and visual cortex. Theseregions are shown in Figure 2 and in Table 3. No regions were significantly deactivatedduring the task when comparing both task levels against rest or when looking at thetask levels separately. When activity at four versus six platforms was compared, the leftsupramarginal gyrus showed a trend for being more active at four than six platforms(Table 3), but no regions showed increased activity with increased task difficulty. In theROI analysis of the hippocampus/parahippocampus, a cluster within the right hippo-campus was significantly activated during the task compared with rest (x y z = 24, −25,−8; 144 voxels; Z = 3.96).

Patient groupIn the aMCI group, the brain regions that were significantly active covered a similarnetwork, but on visual inspection showed less activation (Figure 2, Table 4). Unlike in thecontrol group, an area in the midline cuneus showed task-related deactivations compared

AGING, NEUROPSYCHOLOGY, AND COGNITION 9

Dow

nloa

ded

by [

Col

in D

ouri

sh]

at 0

3:02

03

Aug

ust 2

015

with rest and no regions in the hippocampal/parahippocampal ROI were significantlyactivated or deactivated by the task. No regions were sensitive to a change in number ofplatforms, as there were no effects when comparing activation at four versus six platforms.

aMCI versus control groupWhen the control group was directly compared to the aMCI group across both tasklevels, a widespread network of regions was significantly more active in controls than inaMCI Patients (Table 5, Figure 3: red). This included areas important for navigation suchas the retrosplenial cortex, bilateral precuneus, left dorsolateral prefrontal cortex(DLPFC), and caudate nucleus, extending into the parahippocampal gyrus and hippo-campus. It also included areas such as the left middle frontal gyrus, bilateral insula, andoccipital cortex. In the reverse contrast, two regions were significantly less active in thecontrols compared with the aMCI patients (Table 5, Figure 3: blue). These regions are inthe right DLPFC, in a medial region of the superior frontal gyrus and in the primary

Figure 2. Activation patterns for task versus rest condition. Left side shows control data and rightside shows data from aMCI group.

Table 3. Task-related activations in control participants. BA = Brodmann area, L = left, R = right.x y z BA Voxels Z p(FWE-corr)

Task > RestPrecuneus R 21 −70 54 7 79569 6.04 <.001Supramarginal Gyrus R 56 −19 28 40 601 4.41 .002Insula L −51 17 −3 13 495 3.94 .005

4 Platforms > RestExtrastriate Cortex L −27 −72 34 19 100198 6.09 <.001

6 Platforms > RestPrecuneus R 20 −70 56 7 50807 5.84 <.001Precentral Gyrus R 56 11 10 6 2319 5.15 <.001Supramarginal Gyrus R 59 −18 28 40 335 4.39 .048Thalamus L −20 −22 0 − 552 3.54 .005

4 Platforms > 6 PlatformsSupramarginal Gyrus L −60 −51 36 40 291 4.25 .088

10 E.M. MIGO ET AL.

Dow

nloa

ded

by [

Col

in D

ouri

sh]

at 0

3:02

03

Aug

ust 2

015

Figure 3. Activation maps for controls greater than aMCI patients (red) and controls less than aMCIpatients (blue). To view this figure in color, please see the online version of this journal.

Table 4. Task-related activations and deactivations in the aMCI group.x y z BA Voxels Z p(FWE-corr)

Task > RestMiddle occipital gyrus L −30 −97 9 18 8692 5.34 <.001Superior frontal gyrus L 0 32 52 8 887 4.87 <.001Middle frontal gyrus L −33 24 22 46 354 4.00 .003Orbital frontal gyrus R 35 33 0 47 324 3.84 .005Supplementary motor area R 12 −19 57 6 413 4.00 .001Precentral gyrus L −28 −15 54 6 233 3.5 .026Postcentral gyrus L −54 −21 54 2 252 4.05 .018Superior parietal lobule R 20 −57 67 5 284 3.85 .010Middle temporal gyrus R 51 −64 −3 37 224 4.23 .031

4 Platforms > RestMiddle occipital gyrus L −28 −82 28 19 529 4.43 <.001Middle occipital gyrus L −32 −96 6 18 517 4.02 <.001Medial frontal gyrus L −12 −3 60 6 1224 4.33 <.001Medial frontal gyrus R 18 −6 61 6 347 3.79 .006Superior frontal gyrus R 6 35 51 8 1189 4.21 <.001Orbital frontal gyrus R 35 20 −14 47 529 4.73 <.001Superior parietal lobule L −28 −67 52 7 306 3.86 .001Lingual gyrus R 20 −87 −14 18 1264 4.59 <.001

6 Platforms > RestMiddle occipital gyrus L −30 −91 12 18 6385 5.64 <.001Middle frontal gyrus R 38 9 57 8 475 4.63 <.001Medial frontal gyrus R 4 11 48 32 404 4.47 <.001Medial frontal gyrus R 16 −6 60 6 226 3.79 .012Postcentral gyrus L −54 −21 52 2 3197 4.31 <.001Precentral gyrus L −32 −3 46 6 747 4.19 <.001Precuneus R 21 −63 48 7 1271 4.16 <.001Supramagrinal gyrus R 52 −43 43 40 305 4.76 .002

Task < RestCuneus R 6 −91 25 18 206 4.31 .045

6 Platforms < RestCuneus R 3 −87 30 18 225 3.94 .013

Table 5. Significant regions for aMCI patients versus controls.x y z BA Voxels Z p(FWE-corr)

Controls > aMCIInferior occipital gyrus L −39 −73 −3 19 43173 6.73 <.001Precuneus L −30 −67 33 7 1769 5.61 <.001Middle frontal gyrus L −24 38 24 46 6553 5.55 <.001Retrosplenial cortex R 12 −58 13 30 633 4.00 .040Controls > aMCI (Parahippocampal/Hippocampal ROI)Hippocampus L −16 −25 −11 – 253 4.29 .019Parahippocampus R 18 −42 −3 27 552 4.32 .003Controls < aMCIMedial frontal gyrus R 8 51 43 9 1008 5.63 .006Precentral gyrus L −16 −28 67 4 1099 5.11 .004

AGING, NEUROPSYCHOLOGY, AND COGNITION 11

Dow

nloa

ded

by [

Col

in D

ouri

sh]

at 0

3:02

03

Aug

ust 2

015

motor cortex, in the precentral gyrus. Using the parahippocampal/hippocampal ROI, onecluster in the left posterior hippocampus and one in the right parahippocampal gyrus,extending into the hippocampus, were significantly more active in the control groupthan in the patients (Table 5, Figure 4).

Across both groups, a region in the right supplementary motor area (BA 6) was signifi-cantly more active at four platforms versus six platforms (x y z = 14, −18, 55; 724 voxels;Z = 4.21). No regions were significant in the reverse contrast. No regions showed significanteffects of choice (start of trial versus end of trial). Finally, driven by the behavioral results, wecompared activity in the aMCI group versus controls at the end of the four-platform trials,where the only performance differences were seen. This showed significantly more activityin the right caudate in the control group compared with the patients (x y z = 21, −9, 31; 610voxels; Z = 4.73) and no regions for the reverse contrast.

VBM analysis

Multiple regions of reduced gray matter were seen in the aMCI group in comparison tothe controls (Table 6). This included clusters with peaks in the hippocampus bilaterally,as well as the left perirhinal cortex. Another cluster covered the right thalamus. Noregions showed significantly higher gray matter in the aMCI group.

Figure 4. Activation maps for controls greater than aMCI patients in hippocampal/parahippocampalROI. A: Left hemisphere peak and B: Right hemisphere peak. See also Table 5.

Table 6. Significant clusters for VBM results where reduced gray matter was found in aMCI patientscompared with controls.Region BA Area Peak MNI Co-ordinates Cluster Size Z p(uncorrected)Hippocampus L – −53 314 3.78 <.001Hippocampus R – 33–12 −11 246 3.62 <.001Perirhinal cortex L 36 −39 3–30 245 3.89 <.001Inferior temporal gyrus L 20 −106 1354 4.45 <.001Thalamus/vermis R – −41 1226 4.00 <.001Gyrus rectus R 11 10 27–17 399 4.12 <.001

12 E.M. MIGO ET AL.

Dow

nloa

ded

by [

Col

in D

ouri

sh]

at 0

3:02

03

Aug

ust 2

015

Discussion

This study used a virtual reality navigation task based on the Olton Radial ArmMaze in an fMRI study comparing aMCI patients and healthy control participants.Both groups performed well on the task and performance data were not sensitiveenough to differentiate patients from controls, except for a subtle deficit at the endof the easier task-level trials. Our fMRI results showed significantly different recruit-ment of areas known to be important for spatial navigation between the twogroups, covering widespread networks. Specifically, aMCI patients showed reducedrecruitment of the hippocampus alongside increased recruitment of the retrosple-nial cortex, precuneus, and caudate nucleus. Lateralized differences were seen inthe DLPFC.

In this study we have demonstrated that aMCI patients can successfully com-plete what seems on the surface to be a relatively complicated navigation task,albeit in the context of structured and standardized training beforehand at screen-ing and on the day of the scan. All patients (and control participants) successfullycompleted a dry run outside the scanner immediately prior to the scanning session.With this training, the patients performed as well as controls, except for a subtledifference toward the end of the four platforms trials. This subtle performancedeficit follows previous reports in the literature, where some studies show deficitsin navigation tasks while others are unimpaired, depending on the task.

Our healthy control group showed an expected right hippocampal activationduring the task in our ROI analysis. Right hippocampal involvement in spatialnavigation is well established from previous fMRI studies (Hartley et al., 2003;Iaria et al., 2003; Maguire et al., 1998; Wolbers et al., 2007), as well as MRI studieslinking the volume or structural integrity of the region to behavioral performance(Bohbot et al., 2007; Iaria et al., 2008). In the present group of healthy elderlycontrols, hippocampal activation was not significant in the whole-brain analysis, asexpected from results showing reduced hippocampal activity during navigationwith age (Antonova et al., 2009). In using a virtual reality Arena task, an analogof the Morris Swim Maze, multiple replications have shown robust hippocampalactivation in young, healthy control participants (Antonova et al., 2009, 2011;Parslow et al., 2004). However, when the same task was used with healthy olderparticipants, hippocampal activation was not seen in whole-brain analysis(Antonova et al., 2009). Significantly reduced activation in the hippocampus inolder versus younger participants has been replicated in different navigation tasks(Meulenbroek, Petersson, Voermans, Weber, & Fernández, 2004; Moffat, Elkins, &Resnick, 2006).

Within the hippocampal/parahippocampal ROI, our aMCI group showed significantlyreduced activation compared with controls, with the peak in the hippocampus proper inthe left hemisphere and the parahippocampal gyrus in the right hemisphere.Significantly reduced activity in this anterior region of the right parahippocampalgyrus is frequently reported in studies of episodic memory in MCI patients, and it isone of the few regions reliably implicated in MCI as revealed by meta-analysis(Browndyke et al., 2013). In a separate study involving the same participants as usedin this study, as well as some additional participants, reduced hippocampal activity was

AGING, NEUROPSYCHOLOGY, AND COGNITION 13

Dow

nloa

ded

by [

Col

in D

ouri

sh]

at 0

3:02

03

Aug

ust 2

015

found during a working memory N-back task (Migo et al., 2015), mirroring our resultshere. This suggests that reduced hippocampal activity is present both for tasks wherehippocampal involvement is expected, as well as in those, such as the N-back, where itnot expected to be critical.

The aMCI group showed a much restricted network of task-related activity,notably including the DLPFC. When the two groups were directly compared, awide network showed significantly more activation in the controls than in theaMCI group and some restricted areas were significantly more activated in patientsthan controls. These comparisons showed laterality-related dissociations within thefrontal cortex. Patients recruited the right DLPFC (BA 46) more than controls, whoinstead showed greater activation of the left DLPFC (BA 46 and 9) and leftmedial prefrontal cortex (BA 32). A very similar region in BA 9 has been found tobe reliability less activated in MCI than in healthy controls in a meta-analysis ofepisodic memory studies (Browndyke et al., 2013), showing some consistencyacross tasks. The role of the left DLPFC in verbal working memory (Owen,McMillan, Laird, & Bullmore, 2005; Wager & Smith, 2003) may explain its activationin controls, since participants often used a strategy of continually rehearsing whichplatforms had been visited in order to complete the task, allowing workingmemory to contribute to performance. In meta-analysis, the left DLPFC is morefrequently reported as being involved in verbal working memory than the right(Wager & Smith, 2003), consistent with verbal maintenance across the task incontrols. The right DLPFC may have been additionally recruited in aMCIpatients as a result of attentional demands, a role suggested, for example, bymeta-analysis of Stroop task performance in healthy controls (Vanderhasselt,Raedt, & Baeken, 2009).

The laterality effects in BA 46 may also reflect how demanding participantsfound the task. Meta-analysis shows that for spatial and verbal working memory,the left DLPFC is implicated for passive working memory storage whereas the rightDLPFC is implicated for executive working memory more than the left (Wager &Smith, 2003). Whilst the procedure used in the task was designed to prevent theemergence of organizational strategies, for the aMCI patients it may have beensufficiently demanding to recruit more executive working memory networks. Thetask is set up to limit the number of choices on specific trial to avoid strategiessuch as “clustering” or “ordering” spatial information through repetitive routes,done by using a pseudorandom method of certain locations not being accessiblefor checking on each trial (see Methods). However, greater demands on spatialworking memory may have occurred in the aMCI group, with the participantsinvolving more central executive resources. More recently, the importance of theprefrontal cortex interacting with the hippocampus for successful navigation hasbeen stressed by Spiers and Gilbert (2015). The right lateral prefrontal cortex maybe particularly important for detecting that a change in route is needed as a resultof environment alteration, with the frontopolar prefrontal cortex involved in re-planning the route. The posterior hippocampus processes the new path while thesuperior prefrontal cortex may be particularly important for dealing with conflictingroute options. This interpretation suggests that the aMCI and control participantswere differentially recruiting parts of a navigation network in order to succeed in

14 E.M. MIGO ET AL.

Dow

nloa

ded

by [

Col

in D

ouri

sh]

at 0

3:02

03

Aug

ust 2

015

the task. If spatial navigation involves much more than just hippocampal proces-sing in isolation, this could perhaps explain why our patients did well on the taskdespite hippocampal atrophy and demonstrable episodic memory impairmentsusing standard tests.

Other regions that were recruited significantly more in the healthy controls thanthe aMCI patients include those known to be important for spatial navigation, suchas the caudate, the precuneus, and the retrosplenial cortex. When looking at thespecific part of the task where performance differences were seen among thegroups, at the end of the four platform trials, the right caudate nucleus wassignificantly more active in controls versus patients. Caudate activation is alsoassociated with egocentric spatial processing (Postle & D’Esposito, 2003) and over-learning of spatial environments (Dahmani & Bohbot, 2015). In most spatial tasks,egocentric and allocentric processing interact and are coordinated to solve aparticular task, but here the task is designed to emphasize allocentric processing.It is notable that in a previous study in which hippocampal functioning wassuppressed using scopolamine in normal participants, the caudate nucleus becamemore active, which was interpreted as involving an egocentric compensatorymechanism (Antonova et al., 2011). Although it is possible that egocentric choicescould sometimes be successful on this task, an egocentric-based mnemonic wouldtend to make responses less accurate. The caudate nucleus is also associated withusing landmarks in spatial navigation (Doeller et al., 2008), which is consistent withthe design of the task to utilize major landmarks in the virtual environment tonavigate between platforms. Studies of brain dysfunction and/or atrophy in MCIhave tended to focus on the medial temporal lobes and the hippocampus, butcaudate atrophy has also been reported (Madsen et al., 2010). Rightcaudate atrophy has been particularly associated with MCI to AD conversion(Madsen et al., 2010).

A significant reduction in recruitment of the precuneus in MCI, as found in thisstudy, is one of the most widely reported results in fMRI studies of MCI usingtraditional memory tasks, such as associative memory (Pihlajamäki & Sperling,2008). Disruption here is thought to occur early in the progress of degenerationtoward dementia (Pihlajamäki & Sperling, 2008). The retrosplenial cortex was alsoactivated more strongly in controls than in our aMCI patients. The role of theregion in navigation is thought to be centered on switching between egocentricand allocentric frame of reference, and patients with retrosplenial damage struggleto navigate in familiar environments, even when landmarks are recognized (Vannet al., 2009). Hypoperfusion in the retrosplenial cortex is a key feature of MCI(Nestor, Fryer, Ikeda, & Hodges, 2003) and is noticeable before any changes inthe parahippocampal gyrus (Minoshima et al., 1997). fMRI investigations of cogni-tion beyond long-term episodic memory remain limited and therefore we arecautious that we should not over-interpret group differences in regions recruitedfor the task; as more studies are published it will be possible to establish whichregions show the most reliable group differences. A recent meta-analysis on allpublished fMRI tasks comparing MCI patients against age-matched controlsreported not having enough studies to be conclusive as to whether there are anysignificant regions implicated in visuospatial tasks (Li et al., 2015).

AGING, NEUROPSYCHOLOGY, AND COGNITION 15

Dow

nloa

ded

by [

Col

in D

ouri

sh]

at 0

3:02

03

Aug

ust 2

015

Widespread differences in BOLD activity were found despite the lack of majorperformance impairments in the aMCI group. Our lack of performance differenceson the platform task itself cannot be easily attributed to the sample size, since thepatient group showed significant deficits on all the standardized memory tasks. Theregions showing significantly reduced task-related activity in the patient groupcover the key regions known to be important for spatial navigation, and areconsistent with reports of early atrophy and changes in perfusion in MCI fromthe literature. The additional use of prefrontal regions in the aMCI group canpresumably compensate for the lack of recruitment of other regions in the standardnetwork used for navigation. Outside of the expected navigation network, ourresults are consistent with other studies using fMRI in MCI/aMCI patients. Someof the regions where we see reduced activation relative to controls are reliablyreported across a variety of cognitive tasks, as indicated by meta-analysis (Li et al.,2015), notably including the left middle frontal gyrus and right insula. It might beconcluded that reduced activation in aMCI patients reflects changes in those brainregions as a part of an early disease process, before atrophy can be seen, notnecessarily specific in relation to disruption of particular neurocognitive systems.

Our structural imaging VBM analysis found expected loss of gray matter in thehippocampus bilaterally in our aMCI patients compared with controls. We also foundvolume reductions in the left perirhinal cortex and in a cluster extending to cover theright thalamus. Gray matter reductions in medial temporal lobe regions are reliablyfound in meta-analysis, particularly on the left side (Ferreira, Diniz, Forlenza, Busatto, &Zanetti, 2011; Yang et al., 2012). Our results are therefore consistent with the extensiveprevious literature looking at atrophy in aMCI.

There was some evidence that patients were impaired toward the end of the four-platform trials, but not for six platforms. It might be expected that the task would beincreasingly sensitive to impairment with task difficulty. It is worth noting that thisremains a subtle deficit and that, despite showing problems with episodic memory,our aMCI patients performed well on all other cognitive tasks, including a measure ofworking memory. Since participants (both patients and controls) used verbal mainte-nance to keep in mind which platforms they had visited, this task will not haveextensively tapped the types of anterograde memory with which they are impaired onstandardized tests.

One other potential explanation for our aMCI patients performing well on thetask is the training they received before completing the scanning session. Thistraining ensured that all patients, and indeed, the controls, were comfortablewith both the cognitive and practical demands of the task. Training introducedthem to the virtual spatial environment and the use of the trackerball, which was anovel device to use for all participants. Standardized instructions, gradually intro-ducing the task complexity, ensured that all participants, even those with littleexperience of using a computer, were able to successfully complete the task. Thismay explain the lack of any learning effect across the individual task trials. In otherstudies, where participants are asked to navigate the same route, participants learntheir way around a virtual environment and aMCI patients show less improvementacross trials (Weniger et al., 2011). In our study, participants learned the layout ofthe virtual space before entering the scanner and the task instead measured their

16 E.M. MIGO ET AL.

Dow

nloa

ded

by [

Col

in D

ouri

sh]

at 0

3:02

03

Aug

ust 2

015

ability to navigate within a now familiar space. As more navigation tasks are usedwith MCI patients, and in aMCI in particular, these differences in design andtraining will be important for interpreting any performance differences betweengroups.

Much of the existing data using virtual or real-world tasks of spatial navigationshows only small impairments in MCI, particularly when the aMCI group is investi-gated (Cushman et al., 2008; Hort et al., 2007; Tippett et al., 2009). This is partly dueto the limited numbers of existing studies, which means that making replicationsand the development of novel tasks is important. Variation in results will also beinfluenced by the heterogeneity in MCI patients, both within any given group andamong studies (Stephan, Matthews, McKeith, Bond, & Brayne, 2007; Stephan et al.,2013). The outcome measures of our task, reaction time, and error rates may not besensitive enough to fully expose aMCI performance impairments in the same wayas tasks such as the Hidden Goal Task (Hort et al., 2007), based on the Morris WaterMaze (Morris et al., 1982), which measure distance errors. The procedure used wasdesigned to mimic the allocentric spatial memory demands of the Radial Arm Maze,but for humans. In real-life situations, there is often time pressure to make adecision and slower responding has negative consequences. Here, the experimentwas designed to be self-paced and therefore may not reflect the full nature ofeveryday life task demands. A challenge for creating ecologically valid virtual realitytasks for fMRI use is to create the real-world demands, whilst having sufficientexperimental control to ensure reliable measurement.

A limitation of our study is that our sample was relatively small, including onlyeight patients with aMCI. Again, this might have resulted in the task not detectingcognitive change if the task had low sensitivity in this respect. For activationmeasurement, to try to mitigate the risk that our results represent a type I error,our fMRI results are corrected at a whole-brain level with a height threshold of .001,indicating a strict statistical threshold. In order to fully compare and contrastresults, the use of and full reporting of broad neuropsychological backgroundtesting remains important. Here, our group of aMCI patients was carefully selected,screened, and trained on the task, helping us to have a relatively homogenousgroup which has allowed us to see statistically robust BOLD differences.

It should be noted that fMRI tasks like the Platform task are used experimentallyand do not currently replace standardized neuropsychological assessment used todefine aMCI. The lack of behavioral differences between the groups, althoughhelpful in avoiding a confound in interpreting fMRI differences (Price & Friston,1999), indicates that this task does not separate the groups based on performance.However, fMRI tasks such as these help us understand neural changes that precedeperformance deficits and already have potential in the evaluation of pharmaceuticaltreatment, using fMRI brain activations as outcome measures for drug screeningtrials, as suggested for schizophrenia research (Dourish & Dawson, 2014).

In conclusion, in this study we found statistically robust fMRI differences betweenhealthy controls and aMCI patients. Even in the absence of profound performance deficits,widespread differences in activation patterns were found when considering the groupsseparately, but also when directly compared. We believe this to be the first use of an fMRIspatial navigation task in aMCI, showing reduced recruitment of many regions important

AGING, NEUROPSYCHOLOGY, AND COGNITION 17

Dow

nloa

ded

by [

Col

in D

ouri

sh]

at 0

3:02

03

Aug

ust 2

015

for spatial navigation in aMCI, including the hippocampus, caudate, retrosplenial cortex,and precuneus. We also saw reduced activation in a wide network of other regions,matching results from a variety of fMRI tasks used with this population. In contrast, weobserved significantly increased recruitment of the right prefrontal regions in the patientgroup. These effects add weight to suggestions that spatial navigation tasks may bepotentially useful in staging of the disease (Gazova et al., 2012; Lithfous et al., 2013)based on the differential activation patterns rather than performance.

Disclosure statement

No potential conflict of interest was reported by the authors.

Funding

This study was supported by the P1vital CNS Experimental Medicine Consortium (membersAstraZeneca, GlaxoSmithKline, Lundbeck, Organon (a subsidiary of Merck) and Pfizer). GKW waspartly supported by the NIHR Comprehensive Biomedical Research Centre Programme, Oxford. ASand SCRW were supported by the NIHR Biomedical Research Centre for Mental Health at SouthLondon and Maudsley NHS Foundation Trust and King's College London, IoPPN. AS and MDK weresupported by the NIHR Biomedical Research Unit for Dementia at South London and MaudsleyNHS Foundation Trust and King's College London, IoPPN.

References

Antonova, E., Parslow, D., Brammer, M., Dawson, G. R., Jackson, S. H. D., & Morris, R. G. (2009). Age-related neural activity during allocentric spatial memory. Memory, 17, 125–143. doi:10.1080/09658210802077348

Antonova, E., Parslow, D., Brammer, M., Simmons, A., Williams, S., Dawson, G. R., & Morris, R. (2011).Scopolamine disrupts hippocampal activity during allocentric spatial memory in humans: AnfMRI study using a virtual reality analogue of the Morris Water Maze. Journal ofPsychopharmacology, 25, 1256–1265. doi:10.1177/0269881110379285

Ashburner, J. (2007). A fast diffeomorphic image registration algorithm. NeuroImage, 38, 95–113.doi:10.1016/j.neuroimage.2007.07.007

Auger, S. D., Mullally, S. L., & Maguire, E. A. (2012). Retrosplenial cortex codes for permanentlandmarks. PLoS ONE, 7, e43620. doi:10.1371/journal.pone.0043620

Baumann, O., Chan, E., & Mattingley, J. B. (2012). Distinct neural networks underlie encoding ofcategorical versus coordinate spatial relations during active navigation. NeuroImage, 60, 1630–1637. doi:10.1016/j.neuroimage.2012.01.089

Blennow, K., Hampel, H., Weiner, M., & Zetterberg, H. (2010). Cerebrospinal fluid and plasmabiomarkers in Alzheimer disease. Nature Reviews Neurology, 6, 131–144. doi:10.1038/nrneurol.2010.4

Bohbot, V. D., Lerch, J., Thorndycraft, B., Iaria, G., & Zijdenbos, A. P. (2007). Gray matter differencescorrelate with spontaneous strategies in a human virtual navigation task. The Journal ofNeuroscience, 27, 10078–10083. doi:10.1523/jneurosci.1763-07.2007

Bohil, C. J., Alicea, B., & Biocca, F. A. (2011). Virtual reality in neuroscience research and therapy.Nature Reviews Neuroscience, 12, 752–762. doi:10.1038/nrn3122

Browndyke, J. N., Giovanello, K., Petrella, J., Hayden, K., Chiba-Falek, O., Tucker, K. A., . . . Welsh-Bohmer, K. A. (2013). Phenotypic regional functional imaging patterns during memory encodingin mild cognitive impairment and Alzheimer’s disease. Alzheimer’s & Dementia, 9, 284–294.doi:10.1016/j.jalz.2011.12.006

18 E.M. MIGO ET AL.

Dow

nloa

ded

by [

Col

in D

ouri

sh]

at 0

3:02

03

Aug

ust 2

015

Burgess, P., & Shallice, T. (1997). The hayling and brixton tests. Bury St Edmunds: Thames Valley TestCompany.

Cavanna, A. E., & Trimble, M. R. (2006). The precuneus: A review of its functional anatomy andbehavioural correlates. Brain, 129, 564–583. doi:10.1093/brain/awl004

Cushman, L. A., Stein, K., & Duffy, C. J. M. D. (2008). Detecting navigational deficits in cognitiveaging and Alzheimer disease using virtual reality. Neurology, 71, 888–895. doi:10.1212/01.wnl.0000326262.67613.fe

Dahmani, L., & Bohbot, V. D. (2015). Dissociable contributions of the prefrontal cortex to hippo-campus- and caudate nucleus-dependent virtual navigation strategies. Neurobiology of Learningand Memory, 117, 42–50. doi:10.1016/j.nlm.2014.07.002

DeIpolyi, A. R., Rankin, K. P., Mucke, L., Miller, B. L., & Gorno-Tempini, M. L. (2007). Spatial cognitionand the human navigation network in AD and MCI. Neurology, 69, 986–997. doi:10.1212/01.wnl.0000271376.19515.c6

Delis, D. C., Kaplan, E., & Kramer, J. H. (2001). Delis-Kaplan executive function system. San Antonio,TX: Psychological Corporation.

Delis, D. C., Kramer, J., Kaplan, E., & Ober, B. (2000). The California verbal learning test. San Antonio,TX: Psychological Corporation.

Devan, B. D., Goad, E. H., & Petri, H. L. (1996). Dissociation of hippocampal and striatal contribu-tions to spatial navigation in the water maze. Neurobiology of Learning and Memory, 66, 305–323. doi:10.1006/nlme.1996.0072

Doeller, C. F., King, J. A., & Burgess, N. (2008). Parallel striatal and hippocampal systems forlandmarks and boundaries in spatial memory. Proceedings of the National Academy ofSciences, 105, 5915–5920. doi:10.1073/pnas.0801489105

Dourish, C. T., & Dawson, G. R. (2014). Precompetitive consortium approach to validation of the nextgeneration of biomarkers in schizophrenia. Biomarkers in Medicine, 8, 5–8. doi:10.2217/bmm.13.89

Epstein, R. A. (2008). Parahippocampal and retrosplenial contributions to human spatial naviga-tion. Trends in Cognitive Sciences, 12, 388–396. doi:10.1016/j.tics.2008.07.004

Ferreira, L. K., Diniz, B. S., Forlenza, O. V., Busatto, G. F., & Zanetti, M. V. (2011). Neurostructuralpredictors of Alzheimer’s disease: A meta-analysis of VBM studies. Neurobiology of Aging, 32,1733–1741. doi:10.1016/j.neurobiolaging.2009.11.008

Gazova, I., Vlcek, K., Laczó, J., Nedelska, Z., Hyncicova, E., Mokrisova, I., . . . Hort, J. (2012). Spatialnavigation-A unique window into physiological and pathological aging. Frontiers in AgingNeuroscience, 4. doi:10.3389/fnagi.2012.00016

Hartley, T., Maguire, E. A., Spiers, H. J., & Burgess, N. (2003). The well-worn route and the path lesstraveled: Distinct neural bases of route following and wayfinding in humans. Neuron, 37, 877–888. doi:10.1016/s0896-6273(03)00095-3

Henderson, V. W., Mack, W., & Williams, B. W. (1989). Spatial disorientation in Alzheimer’sdisease. Archives of Neurology, 46, 391–394. doi:10.1001/archneur.1989.00520400045018

Hort, J., Laczó, J., Vyhnálek, M., Bojar, M., Bureš, J., & Vlček, K. (2007). Spatial navigation deficit inamnestic mild cognitive impairment. Proceedings of the National Academy of Sciences of theUnited States of America, 104, 4042–4047. doi:10.1073/pnas.0611314104

Iachini, T., Iavarone, A., Senese, V. P., Ruotolo, F., & Ruggiero, G. (2009). Visuospatial memory inhealthy elderly, AD and MCI: A review. Current Aging Science, 2, 43–59. doi:10.2174/1874609810902010043

Iaria, G., Chen, J.-K., Guariglia, C., Ptito, A., & Petrides, M. (2007). Retrosplenial and hippocampalbrain regions in human navigation: Complementary functional contributions to the formationand use of cognitive maps. European Journal of Neuroscience, 25, 890–899. doi:10.1111/j.1460-9568.2007.05371.x

Iaria, G., Lanyon, L. J., Fox, C. J., Giaschi, D., & Barton, J. J. S. (2008). Navigational skills correlate withhippocampal fractional anisotropy in humans. Hippocampus, 18, 335–339. doi:10.1002/hipo.20400

Iaria, G., Petrides, M., Dagher, A., Pike, B., & Bohbot, V. D. (2003). Cognitive strategies dependent onthe hippocampus and caudate nucleus in human navigation: Variability and change withpractice. Journal of Neuroscience, 23, 5945–5952.

AGING, NEUROPSYCHOLOGY, AND COGNITION 19

Dow

nloa

ded

by [

Col

in D

ouri

sh]

at 0

3:02

03

Aug

ust 2

015

Laczó, J., Andel, R., Vyhnalek, M., Vlcek, K., Magerova, H., Varjassyova, A., & Hort, J. (2010). Humananalogue of the Morris Water Maze for testing subjects at risk of Alzheimer’s disease.Neurodegenerative Diseases, 7, 148–152.

Laczó, J., Vlček, K., Vyhnálek, M., Vajnerová, O., Ort, M., Holmerová, I., . . . Hort, J. (2009). Spatialnavigation testing discriminates two types of amnestic mild cognitive impairment. BehaviouralBrain Research, 202, 252–259. doi:10.1016/j.bbr.2009.03.041

Li, H.-J., Hou, X.-H., Liu, -H.-H., Yue, C.-L., He, Y., & Zuo, X.-N. (2015). Toward systems neuroscience inmild cognitive impairment and Alzheimer’s disease: A meta-analysis of 75 fMRI studies. HumanBrain Mapping, 36, 1217–1232. doi:10.1002/hbm.22689

Lithfous, S., Dufour, A., & Després, O. (2013). Spatial navigation in normal aging and the prodromalstage of Alzheimer’s disease: Insights from imaging and behavioral studies. Ageing ResearchReviews, 12, 201–213. doi:10.1016/j.arr.2012.04.007

Madsen, S. K., Ho, A. J., Hua, X., Saharan, P. S., Toga, A. W., Jack, C. R., . . . Thompson, P. M. (2010). 3Dmaps localize caudate nucleus atrophy in 400 Alzheimer’s disease, mild cognitive impairment,and healthy elderly subjects. Neurobiology of Aging, 31, 1312–1325. doi:10.1016/j.neurobiolaging.2010.05.002

Maguire, E. A., Burgess, N., Donnett, J. G., Frackowiak, R. S. J., Frith, C. D., & O’Keefe, J. (1998).Knowing where and getting there: A human navigation network. Science, 280, 921–924.doi:10.1126/science.280.5365.921

Marsh, R., Hao, X., Xu, D., Wang, Z., Duan, Y., Liu, J., . . . Peterson, B. S. (2010). A virtual reality-basedFMRI study of reward-based spatial learning. Neuropsychologia, 48, 2912–2921. doi:10.1016/j.neuropsychologia.2010.05.033

Meulenbroek, O., Petersson, K. M., Voermans, N., Weber, B., & Fernández, G. (2004). Age differencesin neural correlates of route encoding and route recognition. NeuroImage, 22, 1503–1514.doi:10.1016/j.neuroimage.2004.04.007

Migo, E. M., Mitterschiffthaler, M., O’Daly, O., Dawson, G. R., Dourish, C. T., Craig, K. J., . . . Morris, R.G. (2015). Alterations in working memory networks in amnestic mild cognitive impairment.Aging, Neuropsychology, and Cognition, 22, 106–127. doi:10.1080/13825585.2014.894958

Minoshima, S., Giordani, B., Berent, S., Frey, K. A., Foster, N. L., & Kuhl, D. E. (1997). Metabolicreduction in the posterior cingulate cortex in very early Alzheimer’s disease. Annals ofNeurology, 42, 85–94. doi:10.1002/ana.410420114

Mitolo, M., Gardini, S., Fasano, F., Crisi, G., Pelosi, A., Pazzaglia, F., & Caffarra, P. (2013). Visuospatialmemory and neuroimaging correlates in mild cognitive impairment. Journal of Alzheimer’sDisease, 35, 75–90. doi:10.3233/JAD-121288

Moffat, S. D. (2009). Aging and spatial navigation: What do we know and where do we go?Neuropsychology Review, 19, 478–489. doi:10.1007/s11065-009-9120-3

Moffat, S. D., Elkins, W., & Resnick, S. M. (2006). Age differences in the neural systems supportinghuman allocentric spatial navigation. Neurobiology of Aging, 27, 965–972. doi:10.1016/j.neurobiolaging.2005.05.011

Monacelli, A. M., Cushman, L. A., Kavcic, V., & Duffy, C. J. (2003). Spatial disorientation inAlzheimer’s disease: The remembrance of things passed. Neurology, 61, 1491–1497.doi:10.1212/WNL.61.11.1491

Morris, J. C. (1993). The clinical dementia rating (CDR): Current version and scoring rules.Neurology, 43, 2412–2412. doi:10.1212/WNL.43.11.2412-a

Morris, J. C., Storandt, M., Miller, J. P., McKeel, D. W., Price, J. L., Rubin, E. H., & Berg, L. (2001). Mildcognitive impairment represents early-stage Alzheimer disease. Archives of Neurology, 58, 397–405. doi:10.1001/archneur.58.3.397

Morris, R. G. M., Garrud, P., Rawlins, J. N. P., & O’Keefe, J. (1982). Place navigation impaired in ratswith hippocampal lesions. Nature, 297, 681–683. doi:10.1038/297681a0

Nedelska, Z., Andel, R., Laczó, J., Vlcek, K., Horinek, D., Lisy, J., . . . Hort, J. (2012). Spatial navigationimpairment is proportional to right hippocampal volume. Proceedings of the National Academyof Sciences, 109, 2590–2594. doi:10.1073/pnas.1121588109

Nelson, H. E., & Willison, J. R. (1991). National adult reading test (Part II): Test manual. Windsor:NFER-NELSON.

20 E.M. MIGO ET AL.

Dow

nloa

ded

by [

Col

in D

ouri

sh]

at 0

3:02

03

Aug

ust 2

015

Nestor, P. J., Fryer, T. D., Ikeda, M., & Hodges, J. R. (2003). Retrosplenial cortex (BA 29/30)hypometabolism in mild cognitive impairment (prodromal Alzheimer’s disease). EuropeanJournal of Neuroscience, 18, 2663–2667. doi:10.1046/j.1460-9568.2003.02999.x

O’Keefe, J., & Conway, D. H. (1980). On the trail of the hippocampal engram. PhysiologicalPsychology, 8, 229–238. doi:10.3758/BF03332854

Olton, D. S., & Samuelson, R. J. (1976). Remembrance of places passed: Spatial memory in rats.Journal of Experimental Psychology: Animal Behavior Processes, 2, 97–116. doi:10.1037/0097-7403.2.2.97

Owen, A. M., McMillan, K. M., Laird, A. R., & Bullmore, E. (2005). N-back working memory paradigm:A meta-analysis of normative functional neuroimaging studies. Human Brain Mapping, 25, 46–59. doi:10.1002/(ISSN)1097-0193

Parslow, D. M., Rose, D., Brooks, B., Fleminger, S., Gray, J. A., Giampietro, V., . . . Morris, R. G. (2004).Allocentric spatial memory activation of the hippocampal formation measured with fMRI.Neuropsychology, 18, 450–461. doi:10.1037/0894-4105.18.3.450

Petersen, R. C. (2004). Mild cognitive impairment as a diagnostic entity. Journal of Internal Medicine,256, 183–194. doi:10.1111/j.1365-2796.2004.01388.x

Petersen, R. C., Doody, R., Kurz, A., Mohs, R. C., Morris, J. C., Rabins, P. V., . . . Winblad, B. (2001).Current concepts in mild cognitive impairment. Archives of Neurology, 58, 1985–1992.doi:10.1001/archneur.58.12.1985

Pihlajamäki, M., & Sperling, R. A. (2008). fMRI: Use in early Alzheimer’s disease and in clinical trials.Future Neurology, 3, 409–421. doi:10.2217/14796708.3.4.409

Postle, B. R., & D’Esposito, M. (2003). Spatial working memory activity of the caudate nucleus issensitive to frame of reference. Cognitive, Affective, & Behavioral Neuroscience, 3, 133–144.doi:10.3758/CABN.3.2.133

Price, C. J., & Friston, K. J. (1999). Scanning patients with tasks they can perform. Human BrainMapping, 8, 102–108. doi:10.1002/(SICI)1097-0193(1999)8:2/3<102::AID-HBM6>3.0.CO;2-J

Prvulovic, D., Bokde, A. L. W., Faltraco, F., & Hampel, H. (2011). Functional magnetic resonanceimaging as a dynamic candidate biomarker for Alzheimer’s disease. Progress in Neurobiology, 95,557–569. doi:10.1016/j.pneurobio.2011.05.008

Schinazi, V. R., Nardi, D., Newcombe, N. S., Shipley, T. F., & Epstein, R. A. (2013). Hippocampal sizepredicts rapid learning of a cognitive map in humans. Hippocampus, 23, 515–528. doi:10.1002/hipo.v23.6

Shi, F., Liu, B., Zhou, Y., Yu, C., & Jiang, T. (2009). Hippocampal volume and asymmetry in mildcognitive impairment and Alzheimer’s disease: Meta-analyses of MRI studies. Hippocampus, 19,1055–1064. doi:10.1002/hipo.20573

Shipman, S. L., & Astur, R. S. (2008). Factors affecting the hippocampal BOLD response duringspatial memory. Behavioural Brain Research, 187, 433–441. doi:10.1016/j.bbr.2007.10.014

Simmons, A., Moore, E., & Williams, S. C. R. (1999). Quality control for functional magneticresonance imaging using automated data analysis and Shewhart charting. MagneticResonance in Medicine, 41, 1274–1278. doi:10.1002/(sici)1522-2594(199906)41:6<1274::aid-mrm27>3.0.co;2-1

Sperling, R. (2011). The potential of functional MRI as a biomarker in early Alzheimer’s disease.Neurobiology of Aging, 32, S37–S43. doi:10.1016/j.neurobiolaging.2011.09.009

Spiers, H. J., & Gilbert, S. J. (2015). Solving the detour problem in navigation: A model of prefrontaland hippocampal interactions. Frontiers in Human Neuroscience, 9. doi:10.3389/fnhum.2015.00125

Stephan, B. C. M., Matthews, F. E., McKeith, I. G., Bond, J., & Brayne, C. (2007). Early cognitivechange in the general population: How do different definitions work? Journal of the AmericanGeriatrics Society, 55, 1534–1540. doi:10.1111/j.1532-5415.2007.01386.x

Stephan, B. C. M., Minett, T., Pagett, E., Siervo, M., Brayne, C., & McKeith, I. G. (2013). Diagnosingmild cognitive impairment (MCI) in clinical trials: A systematic review. BMJ Open, 3. doi:10.1136/bmjopen-2012-001909

Tippett, W. J., Lee, J.-H., Zakzanis, K. K., Black, S. E., Mraz, R., & Graham, S. J. (2009). Visuallynavigating a virtual world with real-world impairments: A study of visually and spatially guided

AGING, NEUROPSYCHOLOGY, AND COGNITION 21

Dow

nloa

ded

by [

Col

in D

ouri

sh]

at 0

3:02

03

Aug

ust 2

015

performance in individuals with mild cognitive impairments. Journal of Clinical and ExperimentalNeuropsychology, 31, 447–454. doi:10.1080/13803390802251360

Vanderhasselt, M.-A., Raedt, R., & Baeken, C. (2009). Dorsolateral prefrontal cortex and stroopperformance: Tackling the lateralization. Psychonomic Bulletin & Review, 16, 609–612.doi:10.3758/pbr.16.3.609

Vann, S. D., Aggleton, J. P., & Maguire, E. A. (2009). What does the retrosplenial cortex do? NatureReviews Neuroscience, 10, 792–802. doi:10.1038/nrn2733

Wager, T. D., & Smith, E. E. (2003). Neuroimaging studies of working memory: A meta-analysis.Cognitive, Affective, & Behavioral Neuroscience, 3, 255–274. doi:10.3758/CABN.3.4.255

Wechsler, D. (1998). Wechsler memory scale ® - 3rd UK edition. London: Psychological Corporation.Wechsler, D. (1999). WASI manual. San Antonio, TX: Psychological Corporation.Wegman, J., Fonteijn, H. M., Van Ekert, J., Tyborowska, A., Jansen, C., & Janzen, G. (2014). Gray and

white matter correlates of navigational ability in humans. Human Brain Mapping, 35, 2561–2572.doi:10.1002/hbm.22349

Weniger, G., Ruhleder, M., Lange, C., Wolf, S., & Irle, E. (2011). Egocentric and allocentric memory asassessed by virtual reality in individuals with amnestic mild cognitive impairment.Neuropsychologia, 49, 518–527. doi:10.1016/j.neuropsychologia.2010.12.031

Wolbers, T., Wiener, J. M., Mallot, H. A., & Büchel, C. (2007). Differential recruitment of thehippocampus, medial prefrontal cortex, and the human motion complex during path inte-gration in humans. Journal of Neuroscience, 27, 9408–9416. doi:10.1523/JNEUROSCI.2146-07.2007

Wolk, D. A., & Detre, J. A. (2012). Arterial spin labeling MRI: An emerging biomarker for Alzheimer’sdisease and other neurodegenerative conditions. Current Opinion in Neurology, 25, 421–428.doi:10.1097/WCO.0b013e328354ff0a

Yang, J., Pan, P., Song, W., Huang, R., Li, J., Chen, K., . . . Shang, H. (2012). Voxelwise meta-analysis ofgray matter anomalies in Alzheimer’s disease and mild cognitive impairment using anatomiclikelihood estimation. Journal of the Neurological Sciences, 316, 21–29. doi:10.1016/j.jns.2012.02.010

22 E.M. MIGO ET AL.

Dow

nloa

ded

by [

Col

in D

ouri

sh]

at 0

3:02

03

Aug

ust 2

015