B

H

PAa

b

c

d

e

a

ARR2A

KB3AMVM

1

adp

fliEmn

q

LYc

0d

Journal of Neuroscience Methods 205 (2012) 86– 95

Contents lists available at SciVerse ScienceDirect

Journal of Neuroscience Methods

journa l h omepa g e: www.elsev ier .com/ locate / jneumeth

asic Neuroscience

igh resolution 3 T fMRI in anesthetized monkeys

alma Pró-Sistiagaa,b,c,∗, Franck Lambertona, Thomas Boraudb,c, Romaric Saulniera,lan R. Younga,d, Bernard Bioulacb,c, Christian Grossb,c,e, Bernard Mazoyera

GIP Cyceron, Campus Jules Horowitz, Bd. Henri Becquerel, 14074 Caen, FranceUniversité de Bordeaux, Institut des Maladies Neurodégénératives, UMR 5293, F-33000 Bordeaux, FranceCNRS, Institut des Maladies Neurodégénératives, UMR 5293, F-33000 Bordeaux, FranceINSERM U.919, Serine Proteases and Pathophysiology of the Neurovascular Unit – SP2U, Caen, FranceCHU de Bordeaux, F-33000 Bordeaux, France

r t i c l e i n f o

rticle history:eceived 28 October 2011eceived in revised form3 December 2011ccepted 23 December 2011

eywords:OLD

T magnetnesthesiaacaca mulattaisual stimulationethodology

a b s t r a c t

Although there are numerous 3 T MRI research devices all over the world, only a few functional studiesat 3 T have been done in anesthetized monkeys. In the past, anesthetized preparations were reported tobe misleading when exploring cortical brain regions outside the primary sensory areas. Nonetheless, agreat improvement has been achieved in the limited effect of anesthetic agents on the reactivity of thebrain.

Here, we re-address the feasibility and potential applications of the brain oxygen level dependent(BOLD) fMRI signal in Macaca mulatta monkeys that have been lightly anesthetized with sevoflurane andcurarized. The monkeys were studied with commercially available coils and sequences using a 3 T clinicalmagnet. We obtained sagittal T1 scout images, gray matter double inversion recovery, standard gradientecho sequences and gradient echo functional imaging sequences. Given that fMRI signals are most readilyidentified in the cerebral cortices, we optimized Echo Planar Imaging sequences to reproduce significantchanges in the BOLD signal subsequent to a visual stimulation paradigm.

Our results provide a satisfactory signal to noise ratio with a limited standard deviation range, when

compared with studies on alert macaques.

We suggest that the 3 T magnet remains a valuable tool to analyze neural pathways in the macaquebrain under light anesthesia and report the use of spatially resolved fMRI in higher visual areas of anes-thetized monkeys. This methodology avoids the need for time-consuming training of awake monkeys, isstable over many hours, provides reproducible data and could be applied successfully to future functional

studies.

. Introduction

Magnetic resonance imaging (MRI) techniques are widely used

nd contribute to the understanding of the brain in health andisease. This methodology allows the examination of the globalroperties of the brain in a non-invasive way.

Abbreviations: BOLD, brain oxygen level dependent signal; CBF, cerebral bloodow; CMRO2, cerebral metabolism; MRI, magnetic resonance imaging; DIR, double

nversion recovery sequence; fMRI, functional magnetic resonance imaging; EPI,cho Planar Imaging; FWE, corrected family-wise error rate; GLM, general linearodel; GM, gray matter only images; GRE, standard gradient echo sequence; NHP,

on-human primates; TE, echo time; TI, inversion time; TR, repetition time.∗ Corresponding author at: GIP Cyceron, Campus Jules Horowitz, Bd. Henri Bec-uerel, 14074 Caen, France. Tel.: +33 0231470131; fax: +33 0231470222.

E-mail addresses: Pro@cyceron.fr (P. Pró-Sistiaga), Lamberton@cyceron.fr (F.amberton), tboraud@u-bordeaux2.fr (T. Boraud), Saulnier@cyceron.fr (R. Saulnier),oung@cyceron.fr (A.R. Young), bernard.bioulac@u-bordeaux2.fr (B. Bioulac),hristian.gross@u-bordeaux2.fr (C. Gross), Mazoyer@cyceron.fr (B. Mazoyer).

165-0270/$ – see front matter © 2011 Elsevier B.V. All rights reserved.oi:10.1016/j.jneumeth.2011.12.023

© 2011 Elsevier B.V. All rights reserved.

Neuronal activation is accompanied by an increase in bothenergy metabolism and local cerebral blood flow (Kida and Hyder,2006). These physiological functions permit the identification ofbrain activation through the use of blood oxygenation level-dependent (BOLD) contrast in functional MRI (fMRI). BOLD-fMRI issensitive to the decrease in deoxyhemoglobin concentration dur-ing neuronal activation (Ogawa et al., 1990). On the basis of thisunique property, studies with fMRI have multiplied in human andnon-human primates (NHP).

fMRI studies can be conducted in animal experimental proto-cols, with behaving or with anesthetized preparations. Work hasbeen performed with behaving macaque monkeys in functionalstudies (Goense et al., 2010) but, a major constraint on such pro-cedures is the fact that only a very limited number of macaquesper protocol can be trained and are finally used for fMRI studies

(Goense et al., 2010; Joseph et al., 2006). In fact, this is the principaladvantage of the use of anesthetic preparations in functional stud-ies, but it is thought that anesthesia would greatly influence, or eveninvalidate, the results obtained in both basic electrophysiological

eurosc

amoc

mbcHieeacabccE

atiupbd

fvhnw

o(umucapori

2

(Asaaa

wft(22tRN

P. Pró-Sistiaga et al. / Journal of N

nd MRI investigations (Sloan and Erian, 1993). While this conceptay have been relevant in the past (see below), today’s generation

f anesthetic agents appear more suited to studies of the cerebralirculation if used at an appropriate concentration (Ishizawa, 2007).

The effects of anesthesia on the cerebral circulation andetabolism have always been of concern to physiologists. It has

een well documented in humans that barbiturates depress botherebral blood flow (CBF) and metabolism (CMRO2) (Kida andyder, 2006), and that halogenic agents such as halothane and

soflurane can alter the reactivity of the cerebral circulation to anxternal challenge by, for example, hypo- or hypercapnia (Goodet al., 2009). However, the more recent generation of halogenicgents, which includes desflurane and sevoflurane, when used atoncentrations that do not induce surgical anesthesia, may have

positive role to play in study of the functional reactivity of therain to a stimulus that is repeated over time. In addition, the use ofurare allows muscle relaxation without affecting the neuromus-ular junction, so the neuronal circuits are preserved (Sloan andrian, 1993).

Indeed, if it could be demonstrated that a light maintenancenesthetic regime could be employed during functional activa-ion of the brain, and that fMRI could be used to map changesn the resulting BOLD signal, then such a methodology could besed to circumvent the problems associated with the use of awakerimates (environment habituation, control of physiological andiochemical parameters, movement, experimental studies of longuration) through the use of a readily available clinical 3 T camera.

Therefore, we have re-addressed the use of anesthetized NHP inMRI studies following a visual stimulation paradigm. To circum-ent the problems associated with the use of awake primates, weave addressed the feasibility of fMRI monitoring of the BOLD sig-al in lightly anesthetized and curarized monkeys (Macaca mulatta)ith a clinical 3 T camera.

Given that activity in primary cortical structures is readily rec-gnizable from the fMRI signal, we optimized Echo Planar ImagingEPI) sequences to identify the BOLD signal following a visual stim-lation paradigm, as already published by other authors in awakeonkeys. We show that the BOLD response to the visual stim-

lation is strongly repeatable in anesthetized monkeys and thathanges in intensity are comparable with those of awake primates,s reported by other authors. Beyond reporting the feasibility ofreviously described spatially resolved fMRI in higher visual areasf the NHP in anesthetized M. mulatta, the present results may be ofelevance to future functional studies of the central nervous systemn anesthetized primates with available clinical 3 T magnets.

. Methods

The study was conducted on two adult male rhesus monkeysM. mulatta, 0303 and 0390, weighing 6 and 7 kg, respectively).ll experiments were performed during daytime. A veterinariankilled in the healthcare and maintenance of NHP supervised allspects of animal care. Animals were checked at least daily by

competent person. These checks ensure that all sick or injurednimals are identified and appropriate action is taken.

The experimental procedures were performed in accordanceith the European Directive on the Protection of Animals Used

or Scientific Purposes (2010/63/UE) and the recommendations ofhe Weatherall report “The use of non-human primates in research”http://www.bprc.nl/BPRCE/L4/newsdownloads/The%20use%20of%0non-human%20primates%20in%20research%20-%20The%

0Weatherall%20Report.pdf). The protocol was approved byhe local committee on the ethics of animal experiments (Comiteegional d’Ethique en Experimentation Animale Normandie/03-09-10/15/09-13). Non-invasive imaging of animals (e.g.

ience Methods 205 (2012) 86– 95 87

MRI) with appropriate sedation or anesthesia is considered milddistress experience by the European Directive on the Protection ofAnimals Used for Scientific Purposes (2010/63/UE). Therefore, allefforts were then made to minimize suffering, as initial sedationwas performed before any manipulation and general anesthesiawas maintained throughout the experiment. Animals were alsochecked after recovery from general anesthesia.

Data were obtained from over 11 individual imaging sessions forfMRI (Table 1: 7 sessions from case 0390 and 4 sessions from case0303). The animals were maintained in individual primate cages(at least 1 m3 volume free per animal) on a 12 h light/12 h darkcycle. Fresh water and primate biscuits were available ad libitumand fresh fruits and vegetables were provided daily. Access to foodwas withdrawn at least 12 h before an experimental session. Allexperiments were conducted in the GIP Cyceron facilities whichhave accreditation for non-human primate research (Licence No. D14118001). Regular and efficient cleaning schedule for the roomsmaintained satisfactory hygienic standards.

2.1. Anesthesia and general preparation

In each experiment, the monkey was sedated initially by anintramuscular injection of ketamine (0.2 mg/kg Virbac; Carros,France). The hair on the head and hind legs was shaved anda perfusion line was placed in the saphenous vein. Followingintravenous administration of atropine 0.25 mg (to prevent buc-cal secretions) and curare (Atracurium, 5 mg, to achieve adequatemuscular relaxation) (Hospira; Meudon La Forêt, France), the mon-key was intubated under 2.5% sevofluarne. Atracurium was infusedcontinuously at a rate of 0.75 mg/kg/h throughout the experimentto minimize residual eye movements (Tootell et al., 1988), andanesthesia was maintained with sevoflurane (1.0–1.5%) in a mix-ture of N2O:O2 of 2:1 at a respiratory frequency of 18 cycles perminute with an MRI compatible ventilator (AestivaTM 5MRI; Datex-Ohmeda Inc., Madison, USA). Heart rate, oximetry (spO2), end-tidalCO2 pressure (etCO2) and arterial pressure were monitored contin-uously (Millennia 3155 MVS monitor, In Vivo Research, Orlando,Florida) and the body temperature was maintained at a constantvalue (38.2 ± 0.5 ◦C) by the use of hot water bottles.

The monkeys were placed in the sphinx position in an MRI com-patible stereotaxic device (1430 M MRI, David Kopf Instruments,Tujunga, CA; USA) (Fig. 1). Care was taken to limit the pressurecaused by the introduction of the ear bars into the external auditorymeatus because this procedure may cause undue pain to the animal.During visual stimulation, both eyelids were kept open throughoutthe experiment with surgical adhesive tape, while optimal hydra-tion of the eyes was maintained by the application of physiologicalsaline every 15 min. The room was darkened and a visual stimuluswas presented which consisted of a high-contrast black and whitecheckerboard pattern rotating at 16 Hz and placed 150 cm from theeyes with an angle of 10–11◦ (Fig. 1). The stimuli were presented ina synchronized manner towards the magnet through the use of acomputer connected to a video projector (Panasonic PT-LB30NTE)which projected the image onto a screen placed at the end of thescanner bore (Fig. 1). This stimulus was presented alternately withperiods of blackboard uniform intensity (background) illumination.

2.2. MRI data acquisition

Experiments were performed in a 3 T scanner (Achieva quasardual, Philips Healthcare; Amsterdam; the Netherlands) using smalland medium two element flexible surface coils (small flexible coil

(FlexS) and medium flexible coil (FlexM), Philips Healthcare; Ams-terdam, the Netherlands) positioned either with only one elementfitting the occipital pole (FlexS) (Fig. 1) or with both elements lat-eral to the brain at the level of the ears when necessary (FlexM). The

88 P. Pró-Sistiaga et al. / Journal of Neuroscience Methods 205 (2012) 86– 95

Table 1Summary of the conditions for the visual fMRI sessions used for the functional visual paradigm.a

Session Case FOV (mm) Voxel size (mm) EPI factor Bandwidth (Hz) Coil TR (ms) Nvol Nrun Paradigm (n of TR) RSD (%)

1 0390 122 × 140 × 56 2.2 × 2.2 × 2.0 55 28.98 FlexM 2000 135 20 15OFF + 15ON + 15OFF 2.342 122 × 140 × 56 2.2 × 2.2 × 2.0 55 28.98 FlexM 2000 135 18 15OFF + 15ON + 15OFF 2.533 122 × 140 × 48 2.2 × 2.2 × 2.0 55 28.98 FlexS 2000 120 20 30OFF +15 ON + 30OFF 1.204 122 × 140 × 34 2.2 × 2.2 × 2.0 55 28.98 FlexS 1000 90 50 30OFF + 30ON + 30OFF 1.235 122 × 140 × 28 2.2 × 2.2 × 2.0 55 28.98 FlexS 1000 60 40 20OFF + 20ON + 20OFF 1.776 113 × 130 × 28 2.2 × 2.2 × 2.0 55 28.98 FlexS 1000 60 40 20OFF + 20ON + 20OFF 2.007 113 × 130 × 28 2.2 × 2.2 × 2.0 55 28.98 FlexS 1000 60 30 20OFF + 20ON + 20OFF 1.888 0303 122 × 140 × 34 2.2 × 2.2 × 2.0 55 28.98 FlexS 1000 90 50 30OFF + 30ON + 30OFF 1.159 122 × 140 × 34 2.2 × 2.2 × 2.0 55 28.98 FlexM 1000 90 50 30OFF + 30ON + 30OFF 3.1310 119 × 140 × 28 1.7 × 1.7 × 1.8 65 23.88 FlexM 1000 90 50 30OFF + 30ON + 30OFF 3.8811 104 × 130 × 30 1.6 × 1.6 × 1.5 63 23.84 FlexM 1250 50 50 20OFF + 10ON + 20OFF 2.19

F ber oa al ove

e.

s1h

wgcfAflaa

twsrrT

tfitfha

FpsbaibMs

OV: field of view; TR: repetition time; Nvol: number of acquired volumes (or numcquisitions); RSD: standard deviation over time divided by the mean intensity sign

a The smaller coil used with the larger voxel size gives a smaller noise percentag

mall flexible coil has a circular shape with an external diameter of1 cm and internal diameter of 8 cm, and the medium flexible coilas an elliptic shape of 14 cm × 17 cm.

In all fMRI sessions, a standard sagittal 3D T1 weighted sequenceas performed for scout images in order to obtain the desired

eometry of the fMRI slices. The anterior commissure-posteriorommissure line was visualized and used as a horizontal referenceor the correct delineation of perpendicular coronal slices (Fig. 1).n echo time (TE) of 2.65 ms, a repetition time (TR) of 25 ms and aip angle of 30◦ were used, with a field of view of 180 × 180 × 147nd a matrix of 256 × 256 × 210 during a scanning time of 6 minnd 11 s. Phase encoding was applied along the dorso-ventral axis.

Once the geometry for the slice choice had been obtained, func-ional gradient Echo Planar Imaging (EPI) sequences were acquiredith a volumetric shimming localized to the visual cortex. The

equence consists of a volume acquired in a repetition time TR andepeated Nvol times in order to obtain the dynamic acquisition oruns needed for fMRI. These runs were also repeated Nrun times (seeable 1 for details).

Several scans were run for each animal. The rationale for varyinghe scan parameters was to optimize a gradient echo EPI sequenceor fMRI application on macaque monkeys. Of course, this goalmplies to find the best compromise between spatial resolution and

emporal resolution leading the lowest noise level compared to theew percent of signal modulation due to the BOLD. Practically, weave three constraints. First, we decided to acquire images withn isotropic voxel size limiting partial volume effects and lost of

ig. 1. Schema showing the positioning of the macaque inside the magnet bore, the paradlaced in the sphinx position in a MRI compatible stereotaxic device. For the visual stimerum to maintain the optimal hydration of the eyes. In this schema one element of the Foth elements are placed lateral to the brain at the ears level. (B) Schema illustrating thend white radial checkerboard flickering at 16 Hz stimulus were presented binocularly

mage was projected onto a screen placed at the end of the scanner bore, 150 cm in frooard. The ON and OFF condition duration was chosen as multiple of the repetition timeRI image of the M. mulatta 0390 case presented as an example for fMRI slices geometry

lices selection. An arrow delimitates the approximate location of the coronal image retri

f TR per run); Nrun: number of runs done in each session (or number of dynamicr time, multiplied by 100.

signal. Second, in our scanner, the acquisition matrix size is fixedto either 64 × 64 pixels or 80 × 80, other sizes being not compat-ible with imaging of brain of macaque monkeys. Third, we fixeda unique bandwidth in the phase coding direction, the value wasdefined in order to limit geometrical distortion and the fixed valueinsure the same point spread function between acquisitions and soallowing direct comparison of noise measures.

Having these constraints, we tested two standard surface coilsdedicated for clinical applications. We also tested different dynamicfMRI acquisitions, modifying the TR and the number of volumes ineach run. This gradient echo EPI sequence was used with a TE of30 ms and a TR of 2000 ms, 1250 or 1000 ms and a flip angle of 80◦

or 70◦. The flip angle was adapted according to the TR in order tooptimize the signal. Details of voxel size, FOV, bandwidth, EPI factorand duration are given in Table 1.

Two sequences were used to localize the activation mapsspatially with the same geometry (center of volume prescribed,angulation and thickness of the slices) and local volumetric shim-ming, restrained to the visual cortex, as for the EPI sequence. First,in order to preserve only gray matter signal suppressing the cere-brospinal fluid and white matter signals simultaneously, a doubleinversion recovery sequence was used as already fully described inthe literature as DIR sequence (Bedell and Narayana, 1998). A TE

of 25 ms, a TR of 11,000 ms with a turbo factor of 13, a flip angleof 90◦ and with no interslice gap, were used, in 14 or 16 slices ina total scanning time of 6 min and 14 s. Two 180◦ pulses with dif-ferent times of inversion (TI) were used (TI 325 ms and 3400 ms,

igm of stimulation and example of fMRI slices geometry choice. (A) Monkeys wereulation the eyelids were kept open through the experiment, adding physiologicallexS coil is placed fitting the occipital pole as example but when FlexM coil is used

paradigm of visual stimulation. The room was darkened and a high-contrast blackwith a computer synchronized to the magnet connected to a video projector. Thent of the animal eyes with an eccentricity of 10–11◦ . The OFF condition is a black

of the sequence in each scan session. (C) Sagittal 3D T1 weighted Fast Field Echochoice. The horizontal CA-CP line is shown as a reference for perpendicular coronaleved for posterior checking of the alignment between sequences (cf. Fig. 2).

eurosc

rsootvagaihdc

c9tatdcob

2

r(utaaawoaetvwafi

atdwsTtuaoTvifitE

dt(b

P. Pró-Sistiaga et al. / Journal of N

espectively). This anatomical sequence will be abbreviated as DIRequence for GM images in the following text. It allows recognitionf the visual cortex in a 3D manner and also allows the delineationf the brain contour and retrieval of the correct mask adapted tohe cerebral volume of each slide considered for the functional acti-ation maps. Second, for correct imaging registration verification,

standard gradient echo sequence (GRE) was performed with thiseometry with a first minimal echo of 7 ms, a second echo of 30 msnd a TR and flip angle equal to the TR and angle in the EPI sequence,n 22 slices in a scanning time of 2 min and 54 s. This GRE sequencead the same contrast as our EPI functional sequence with oneifference: no image deformation was present and this allows fororrect visual imaging registration checking.

For the functional visual stimulation paradigm, a checkerboardycle of alternating OFF, ON and OFF stimuli was used in 135, 120,0 or 60 volumes (Table 1). The cycle was set up in a unique eventrial with a simple OFF ON OFF, in order to keep runs as shorts possible in order to interrupt the scan for physiological con-rol (temperature, arterial pressure or anesthetic levels) withoutisturbing the acquisition. In addition, independent runs allow dis-arding runs if artifactual images are seen. Besides, the repetitionf short runs desynchronizes the eventual periodic artifacts causedy the respiration.

.3. MRI data analyses

All data were analyzed off-line on a Pentium computerunning the Linux operating system as implemented in SPM5www.fil.ion.ucl.ac.uk/spm). The rough original images were eval-ated by visual inspection with FSL tools (www.fmrib.ox.ac.uk/fsl/)o check the different volumes and see the potential susceptibilityrtifacts, i.e. image distortions or signal dropouts. Given that visu-lization of the dynamics of all the repetitions showed no motionrtifacts (FSLView version 3.1; www.fmrib.ox.ac.uk), all volumesere taken into account. We chose to make several repetitions

f the visual paradigm, in order to be able to average all volumesnd reinforce statistical power. This precaution makes it possibleventually to remove artifactual volumes. After this FSL analysis,he multi-slice data collected during the experiments were con-erted into a time series of voxels and two pre-processing stepsere applied: realignment with the first volume of the first run

s volume reference and spatial smoothing with a Gaussian kernellter (FWHM) of 3 mm.

As stated previously, the DIR sequence has the same geometrys the EPI sequence (center of volume prescribed, angulation andhickness of the slices). It is then not necessary to corregister the EPIata to the anatomical DIR sequence. In addition, the GRE sequenceith two different echo times had a minimal first echo time and a

econd echo time equal to the echo time used for the EPI images. TheR used, the volumetric shimming and the chosen slices were alsohe same as in the functional images. Those precautions alloweds to obtain images that were intermediate between the functionalnd the anatomical DIR sequence, with exactly the same contrastf the functional images but with minimal geometric deformation.he SPM check registration tool was used subsequently to verifyisually the correct imaging registration of the resulting functionalmages to the GRE (with both echo times) and the GM images (Fig. 2)or data acquisition from each functional session. In addition, andn order to visually check the correct overlay of GM and EPI images,he gray level isocontour of all the GM images was traced over thePI images (Fig. 3).

Furthermore, because we performed intra-subject analyses, we

id not need to apply a normalization step. Statistical analysis athe subject level was performed using the General Linear ModelGLM) as implemented in SPM5. Functional maps were generatedy cross-correlating the post-convoluted time course at each voxel

ience Methods 205 (2012) 86– 95 89

with a boxcar model of the stimulus presentation protocol in eachexperiment. Statistical comparisons were performed using t-testson the resulting beta weights. This GLM analysis resulted in statisti-cal contrast activation maps that were thresholded at a voxel-wiselevel at p < 0.05 with the corrected family-wise error rate (FWE).Data are presented in SPM5 results with a cluster-extent thresholdof 10 voxels.

In order to have comparable data about the different sessionconditions, we calculate in each run the mean value and the stan-dard deviation of the signal modulations over time of the baseline.Baseline is the signal during the time preceding the onset of thestimulation, with no induced BOLD activity. We had computedimages of the baseline signal of the mean signal modulations overtime of all individual voxels (MEAN), standard deviation of the sig-nal modulations over time for all images (STD), and the percentageof noise in this region of interest as the coefficient of variation ofthe brain over time, or the standard deviation range (RSD: meanstandard deviation of the signal modulations of the ROI over timedivided by the mean intensity signal of the ROI over time, multi-plied by 100). With these new data images for each session, wechose a region of interest in the visual cortex arbitrarily and cal-culated the spatial mean data of the MEAN, STD, and RSD with itscoefficient of variation (Table 2). We keep the ROI in visual cortexto have the data conducing to a better signal and lower STD directlyin our zone of interest. As the noise was calculated in a temporalmoment with no stimulation and then no induced BOLD response,we could have the general noise of our acquisition paradigm in thearea of interest.

In order to convert the data to s.d. units to compare session con-ditions and to present data comparable to others authors data, theBOLD signal was normalized for each session by subtracting themean signal of the baseline preceding the onset of the stimulationand dividing it by the standard deviation of the baseline.

Finally, we plotted the BOLD effect temporal profile of the moststatistically significant voxel of the ROI expressed in standard units.

3. Results

We obtained a BOLD activity signal in the monkeys anesthetizedwith 1.0–1.5% sevoflurane after binocular visual stimuli under thedifferent conditions that were tested. The signal modulations overtime of the baseline period preceding the onset of the stimulusare presented in Table 2 for the visual cortex region of interest:mean of signal modulations over time, the mean standard deviationover time, the signal to noise ratio (mean intensity signal over timedivided by the standard deviation over time) and the percentage ofnoise as the variation coefficient of the brain over time with its coef-ficient of variation. Functional maps for both monkeys are shown inFig. 4 and the mean time-course of the BOLD response observed inthe local maximum voxel of the V1 region of interest is illustratedin Fig. 5 for both monkeys. We obtained a mean time brain averagesignal resulting in 0.5–1 standard deviation units with a maximumsignal to noise ratio of 80–87 for a relative background noise thatvaried from 1.15 to 3.88%.

3.1. Signal level analysis

We chose to have the most isotropic voxel size possible, limit-ing partial volume effects and loss of signal. Having a fixed phasecoding bandwidth (identical geometric deformation and impulsionresponse) and EPI factor we tried to modify the TR, with different

number of volumes and with two different surface coils to be able toretrieve higher BOLD signals (Tables 1 and 2, and Fig. 6). The signallevel of the baseline is summarized in Fig. 6, as a function of acqui-sition parameters. Briefly, we performed several sessions with the

90 P. Pró-Sistiaga et al. / Journal of Neuroscience Methods 205 (2012) 86– 95

Fig. 2. Check registration example images as implemented in SPM5 (www.fil.ion.ucl.ac.uk/spm). Functional images (A), GRE images (with both echo times; (B) TE1: 6.9 ms,(D) TE2: 30 ms) and the GM images (C) are correctly aligned. One coronal level is shown as example to visually verify the correct alignment of the resulting functional images(cf. Fig. 1).

Table 2Summary of the signal analysis for all the individual sessions of cases 0390 and 0303.a

Session Case MEAN STD MEAN/STD RSD (%) CV RSD (%)

1 0390 1010.41 23.48 43.03 2.34 ±0.132 1122.60 28.09 39.96 2.53 ±0.133 1071.53 12.71 84.31 1.20 ±0.074 1245.05 15.26 81.59 1.23 ±0.065 1250.99 21.95 56.99 1.77 ±0.126 1194.65 23.69 50.43 2.00 ±0.167 1278.28 24.04 53.17 1.88 ±0.108 0303 1505.79 17.17 87.70 1.15 ±0.069 1400.20 43.38 32.28 3.13 ±0.1610 1393.01 53.95 25.82 3.88 ±0.1411 1343.74 29.14 46.11 2.19 ±0.08

STD, mean standard deviation of the signal modulations over time for all images; MEAN, mean signal modulations over time of all individual voxels; RSD, coefficient ofvariation over time (STD/MEAN multiplied by 100); CV RSD, coefficient of variation of the RSD.

a A higher signal to noise ratio with lower standard deviation range was found for the 3 and 4 scans for case 0390 and for the 8 scan for case 0303. Data are calculated foreach session, with an arbitrarily chosen region of interest in the primary visual cortex.

P. Pró-Sistiaga et al. / Journal of Neuroscience Methods 205 (2012) 86– 95 91

F es. Dif( rders

a

Fw1(

bttnlfr3iabw2e96oa

wsIun

wTat

average BOLD signal resulted in 0.5–1.0 standard deviation units.The BOLD signal is presented in Fig. 5 as an example for each

ig. 3. Gray level isocontour of the GM anatomical images overlaid to the EPI imag−36). The contour of the anatomical images showed a signal loss in the external boll along the continuity of the calcarine fissure.

lexM oval coil with a diameter of 12 cm and a FlexS circular coilith a diameter of 8 cm, we tested different repetition times (from

000 to 2000 ms) and different voxel sizes. With a fixed voxel size9.68 mm3) we have got different baseline durations.

In our visual stimulation experiments, the best compromiseetween signal and noise was obtained by using one element ofhe FlexS coil fitted to the occipital pole closer to the activated cor-ical area. We nevertheless also achieved a relatively good signal tooise ratio when both elements of the FlexM coil were positioned

ateral to the brain at the level of the ears, inside the stereotaxicrame. With FlexM coil, the higher mean signal to noise ratio wasetrieved when a TR of 1250 ms was used with a voxel size of.84 mm3 (MEAN/STD of 46.11 with RSD of 2.19%, see session 11

n Table 2 and Fig. 6). Nevertheless, with the same fixed conditionss used in other sessions of voxel size, EPI factor and bandwidth, theest mean signal to noise ratio was retrieved when a TR of 2000 msas used (voxel size 9.68 mm3 MEAN/STD of 43.03 with RSD of

.34%, see session 1 in Table 2 and Fig. 6). With FlexS coil, the high-st mean signal to noise ratio was retrieved when a voxel size of.68 mm3 was used with a TR of 2000 ms or a TR of 1000 ms, with a0 s baseline time or a 30 s baseline (Mean/STD of 84.31 with RSDf 1.20%, and 81.59 with RSD of 1.23%, respectively) (see sessions 3nd 4 in Table 2 and Fig. 6).

It is important to note that exactly the same session conditionsere repeated with both monkeys, producing the same range of

ignal to noise and RSD (see sessions 4 and 8 in Table 2 and Fig. 6).n addition, in one monkey the same session conditions had beensed with the two different coils with a dramatic loss of signal tooise and RSD increasing (see sessions 8 and 9 in Table 2 and Fig. 6).

It is also noteworthy, that with the same resolution parameters,ith the FlexS coil, changing the baseline duration (having 20 or 30

R, see Table 1) did not change the global RSD data, influencing only

nd slightly the confidence interval (see 3, 4 or 8 sessions comparedo 5, 6 and 7 sessions in Fig. 6).

ferent antero-posterior coronal levels from more anterior (−58) to more posteriorof the EPI images, although the information is kept in the primary visual cortex and

3.2. Functional results

A BOLD response was demonstrated in our anesthetized mon-keys with the visual stimulation paradigm used (Fig. 4). Asillustrated in this figure, we found an effect on blood oxygena-tion levels following presentation of an alternating checkerboardstimulus. For both monkeys, the functional response is expressedas t-values of the significance of the change on statistical contrastactivation maps, thresholded at a voxel-wise level of p < 0.05. Thecorrection was used in the SPM family-wise error rate (FWE) testto avoid the multiplicity problem: multiple false positive resultsthat could occur when multiple hypothesis tests are performedsimultaneously (multiple individual voxel hypotheses).

The activation was localized exclusively to the V1 area of thevisual cortex along the calcarine sulcus (Fig. 4). The images show agood signal to noise ratio with little artifact from motion or mag-netic susceptibility differences. Although EPI image techniques aresensitive to magnetic field inhomogeneities and sometimes pro-duce geometric distortions in the resultant images, we were ableto obtain enough close spatial anatomical approximation with theGM images to make out the visual V1/V2 region in coronal, axialand sagittal views (see data from sessions 8 and 4 of cases 0303and 0390 respectively in Fig. 4 obtained at two different coronallevels and one axial level). For case 0303, BOLD activation wasidentified in the primary visual cortex (V1) leading to the borderof V2 and all along the continuity of the calcarine fissure. Similarresults were obtained for case 0390, although to a lesser extentalong the calcarine fissure. Although we obtained a relative stan-dard deviation range between 1.15% and 3.88% under the conditionstested, which represented the general noise, the mean time brain

monkey, with a maximal t-score value of 15.01 for case 0303 and11.67 for case 0390. It may be concluded that these signals are

92 P. Pró-Sistiaga et al. / Journal of Neuroscience Methods 205 (2012) 86– 95

Fig. 4. Example of BOLD functional maps of both 0303 (session 8) and 0390 (session 4) cases. Visual primary areas are activated in posterior V1 cortex and along the calcarinefissure. We show in (A) different coronal and axial views of the V1 activation after bilateral visual stimulation (described in Fig. 1). In (B) we present the 3D T1 weightedanatomical images of the corresponding approximate levels. Statistical analysis at the subject level was performed using the General Linear Model (GLM) as implementedi e at eaw lted inl e is co

at

3

aotat1sTaBTtibw

4

r

n SPM5; maps were generated by cross-correlating the postconvoluted time coursere performed using t-tests on the resulting beta weighs. This GLM analysis resu

evel of p < 0.05 and a cluster-extent of 10 voxels. The magnitude of the t-score valu

cceptable for detecting statistically significant functional activa-ion in anesthetized monkeys.

.3. Analysis of signal frequencies

When data from all the repetitions of the paradigm were aver-ged, the physiological noise remained present, because pollutionf high amplitude and low frequency was detected in the signal ofhe temporal profile of the cumulated acquisitions; low frequencynalysis of the BOLD signal retrieved is shown in Fig. 5. Given thathe respiratory frequency was fixed at 0.3 Hz, a TR shorter than600 ms corresponding to >0.6 Hz would be sampled without pos-ible frequency aliasing interference. In the very first sessions aR of 2000 ms was used (0.5 Hz) and frequency aliasing was notvoided. Nevertheless, it was possible to retrieve a sufficiently highOLD signal when TR of 1000 and 1250 ms were used (see Fig. 4).hus, there was no need to perform more advanced signal analysiso remove the effect of respiration. Nonetheless, for future studiest may be advisable to retrieve the origin of this frequency if possi-le at the time of acquisition, or to filter the resulting mean signalith an adapted custom-built frequency filter.

. Discussion

We have shown that the anesthetized M. mulatta brain caneadily activate functional pathways that are identified robustly

ch voxel with a boxcar model of the stimulus presentation protocol. Comparisons a FWE corrected statistical contrast activation maps, thresholded at a voxel-wiselor coded as shown by the key inset.

by commercially available magnets, coils and sequences with 3 TfMRI. We have provided evidence of pathway functionality in thevisual system, the simplest pathways to stimulate using a MRIdevice. Beyond reporting the feasibility of spatially resolved fMRI inhigher primary visual areas of M. mulatta, the present results maybe of relevance to future functional studies with high magnetic fieldstrength in the central nervous system of anesthetized primates.

In the present study, we used commercially made coils avail-able for human clinical MRI studies in an effort to simulate clinicalresearch in humans and NHP. In this way, reproducible studiescan be performed also using other commercially available Achievaquasar dual 3 T clinical scanners, with available coils and sequencesand with no particular constraints.

4.1. Differences between anesthetized and non-anesthetizednon-human primates

Our results at 3 T provide, therefore, a satisfactory signal to noiseratio with limited standard deviation range of the signal, in compar-ison with the results of studies in alert macaques (Stefanacci et al.,1998). It is noteworthy that in anesthetized macaques (using 0.3%isoflurane and fentanyl 3 �g/kg/h), other authors have reported a

maximal signal to noise ratio of 23.2 in high resolution EPI slices at4.7 T (0.5 mm3 voxel size) (Logothetis et al., 1999) or a mean changein the signal intensity of 0.5 or 1 standard units (Logothetis et al.,1999; Stefanacci et al., 1998). In anesthetized baboons, a recent

P. Pró-Sistiaga et al. / Journal of Neuroscience Methods 205 (2012) 86– 95 93

Fig. 5. Visually evoked BOLD response. The BOLD response was approximately 0.6 s.d. unit for case 0303 example (session 8, 1.15% standard deviation range) and 0.4 s.d. unitfor case 0390 example (session 4, 1.23% standard deviation range). A low respiratory frequency peak (0.3 Hz) is detected without influencing the retrieval of BOLD signal. (A,C) Mean time course of BOLD response observed in the local maximum voxel of the V1 region of interest, after the SPM5 GLM analysis thresholded at a voxel-wise level ofp < 0.05 corrected FWE test, for 0303 case (A), t = 15.01 and 0390 case (C), t = 11.67. BOLD signal was normalized for each event trial subtracting the mean of a 30 s baselinepreceding the onset of the stimulation and dividing by the standard deviation of the baseline and reported to standard deviation units. The signal retrieved is expressed inunits of the standard deviation (s.d.) of the activity. The visual stimulus is presented for 30 s. (B, D) Low frequency analysis of the signal retrieved, for the 0303 and 0390 casesrespectively (in A and C). s.d. units: standard deviation units; a.u.: arbitrary units.

Fig. 6. Graphic of the RSD data retrieved in the different sessions conditions, from 1 to 11 in both 0390 and 0303 monkeys.Data with the tested FlexM and FlexS surface coils, with the different repetition times used (1000, 1250 and 2000 ms), voxel size (3.84, 5.2, and 9.68 mm3) in the two casesused. See Table 1 for details.

9 euros

sabvrtnwaas

4

mp(aiebabstLWcwmi

cvbtrudm

4

4

ttbsasassedotiS

gii

4 P. Pró-Sistiaga et al. / Journal of N

tudy has shown 0.2–0.3% BOLD signal at 3 T with a 20 mm3 voxel,fter visual stimulation (Wey et al., 2011). Given that the baboonrain is bigger than the macaque brain, these authors found acti-ation with a higher voxel size but no analysis of signal to noiseatio was performed. In the present study, the minimal voxel sizeested was 3.96 mm3 at 3 T. Nevertheless, the maximal signal tooise ratio was achieved when a 9.57 mm3 voxel size was usedith a mean time brain signal between 0.5 and 1.0 standard devi-

tion units. Importantly, this gave a similar mean time brain signalnd higher signal to noise ratio than other authors at 0.5 mm3 voxelize and 4.7 T (Logothetis et al., 1999).

.2. BOLD signal with visual stimulation in monkeys

Although there are a few reports that show activation in theonkey’s visual cortex after visual stimulation using MRI, it has

roved difficult to obtain satisfactory results with brain fMRILogothetis et al., 2001). Attempts to image the visual cortex oflert monkeys have failed to produce a sufficiently clear stimulus-nduced focal activation (Murnane and Howell, 2010; Stefanaccit al., 1998). Although the use of anesthetics is not recommendedecause it can suppress the signals measured with fMRI (Murnanend Howell, 2010), there is some controversy because it haseen shown that primary sensory systems, such as the visual oromatosensory systems, can elicit recognizable signals in anes-hetized NHP: in macaques at 4.7 T (Goense and Logothetis, 2008;ogothetis et al., 2001) and in baboons at 3 T (Wey et al., 2011).

hile good quality fMRI data can be collected by utilizing these pro-edures, the successful goal of the present work was to investigatehether good quality BOLD fMRI data could be collected in a com-ercially available 3 T magnet with standard coils and sequences

n anesthetized M. mulatta monkeys.As already identified in retinotopic studies for a rotating

heckerboard visual stimulus, the primary visual area V1 is acti-ated, importantly, in the most caudal occipital pole of the macaquerain (Tootell et al., 1988; Van Essen et al., 1984 for general retino-opic overview; Warnking et al., 2002). In agreement with theseesults, we have shown primary visual activation after visual stim-lation, as published for other fMRI studies of visual stimulation atifferent magnet fields or anesthetic conditions but not with a 3 Tagnet (Logothetis et al., 1999; Stefanacci et al., 1998).

.3. Methodological commentaries

.3.1. Choice of sequencesThe DIR sequence used in this study utilizes two inversion pulses

o isolate gray matter information by arranging the timing of thewo inversion pulses so that the magnetization from the cere-rospinal fluid and white matter passes through the null pointimultaneously. The resulting anatomical images allow us to obtainnatomical markers of the gray matter to localize in 3D the corticaltructures studied, with the same slice choice as during functionalcquisition. In the literature, different types of inversion recoveryequence are used to improve signal suppression from different tis-ue types, such as white matter or cerebrospinal fluid (Sudhyadhomt al., 2009). As described by these authors, such sequences allowelimitation of structures that could be difficult to recognize inther classic sequences and easy acquisition of references for cor-ical information. This technique is also used in brain imaging tomprove the study of different pathologies (Rugg-Gunn et al., 2006;imon et al., 2010).

Given that the slice EPI images were acquired with the sameeometry as the DIR and GRE sequences, no further imaging reg-stration was necessary. Nevertheless, because the EPI sequences very sensitive to magnetic field inhomogeneities, geometrical

cience Methods 205 (2012) 86– 95

deformation of the slices is possible. However, a volumetricshimming method localized to the visual cortex can reduce theseinhomogeneities drastically. Thus, using volumetric shimming andcarefully chosen parameters for EPI sequence, we expect a verysmall or negligible geometrical deformation between functionaland anatomical images. These precautions allowed us to achievethe correct imaging registration of our functional maps andanatomical images, as seen in the control realized for all functionalsessions (Fig. 2). It is important to note that the standard gradientecho sequence (GRE) used for correct imaging registration verifi-cation, was performed with the EPI geometry (center of volumeprescribed, angulation and thickness of the slices) but not in thesame number of slices, as it could be recognized in the sagittaland axial orientations in Fig. 2, where functional images (A), GREimages (with both echo times; (B) TE1: 6.9 ms, (D) TE2: 30 ms) andthe GM images (C) are correctly aligned. In addition, and in orderto visually check the correct overlay of GM and EPI images, thegray level isocontour of all the GM images was traced over theEPI images at different antero-posterior coronal levels from moreanterior (−58) to more posterior (−36) (Fig. 6). The contour of theanatomical images showed a signal loss in the external borders ofthe EPI images, although the information is retained in the primaryvisual cortex and all along the continuity of the calcarine fissure.The BOLD activation map was then finally overlaid on the GManatomical images (Fig. 4).

4.3.2. Noise sourcesA noise of low-frequency fluctuation or drift is observed com-

monly in fMRI data acquired using BOLD contrast (Bandettini et al.,1993). It is well known that BOLD data suffer from high noise atlow frequencies. This noise appears as a slow drift of the signal,due to scanner instabilities with contribution from physiologicalnoise (Aguirre et al., 2002). Other authors pointed out that this low-frequency drift is assumed to have a physiological origin (Yan et al.,2009) and may arise from systemic sources such as cardiac or res-piratory cycles through aliasing (Yan et al., 2009). In our study, wefixed the respiratory frequency to 18 cycles per minute (0.3 Hz), giv-ing a frequency two times higher than the repetition time of the EPIsequence that was used (TR 1000 ms (1 Hz) or even 1250 ms (0.8 Hz)as recommended by the Nyquist–Shannon sampling theorem). Thistheorem states that the sampling frequency of a signal has to be atleast twice the maximum frequency contained in the signal in orderto be analogous to digital signal conditioning, in order to avoid alias-ing pollution when identifying frequencies. This precaution allowsus to reduce the statistical importance of the physiological noisepresent in our images, as already confirmed by other authors (Yanet al., 2009), but makes it necessary to increase the number of rep-etitions of the functional paradigm (Sultan et al., 2007) in orderto counteract the fact that a short TR retrieves little information ineach run. Given that physiological frequency noise did not interferein the recognition of our visual activation, we required no addi-tional correction algorithms to remove it, as recommended recentlyby other authors to avoid the impact of well-known physiologicalnoise correction algorithms (Vogt et al., 2011).

In agreement with other studies (Logothetis et al., 1999), wefound that controlled, regular respiration and correct positioningof the visual stimulus were critically important for eliminatingmovement artifacts and for obtaining high-resolution results inanesthetized monkeys. B0 drift is known to induce a slow trans-lation of the imaged object along the phase-encoding axis (Benneret al., 2006). An option called ‘dynamic stabilization’ is also pro-posed by the magnet manufacturer to correct the drift of B0

prospectively during each dynamic acquisition by measuring andreadjusting the center frequency at each TR. This option includesand corrects spurious large movements along the dorso-ventralaxis about several pixels. After some tests, we concluded that this

eurosc

mmtoabt

5

pfiMchuut

osm

A

aNc

D

R

A

B

B

B

FG

G

P. Pró-Sistiaga et al. / Journal of N

anufacturer option was not adapted for imaging studies withacaque brains and we switched off the function. In addition,

he sphinx position of our monkeys allowed for higher stabilityf the B0 and fewer inhomogeneity artifacts, as described by otheruthors (Valette et al., 2006), because B0 homogeneity within therain is expected to vary significantly depending upon the orienta-ion of the head, as shown previously in humans.

. Conclusions

The anesthetized macaque brain can readily activate functionalathways that are robustly identified by commercially available 3 TMRI. We provide methodological evidence of functional pathwaysn the visual system, the simplest way to quantify activation in an

RI. We have used commercially made coils available for humanlinical MRI studies in an effort to get nearer clinical research inumans and NHP. The general interest of this paper is that it opensp the possibility of fMRI 3 T studies in anesthetized macaquessing other more invasive approaches such as electrical stimula-ion, which is both easier and more precise than behavioral tasks.

Our methodology could be employed to study other pathwaysr to use other stimulation approaches, or even for longitudinaltudy of models of neurodegenerative diseases to investigate theodification of these pathways.

cknowledgements

The authors wish to thank Tho Hai Nguyen and Laurent Chaz-lviel for their technical support. We also wish gratefully to thankicolas Delcroix for his availability and his great help with statisti-al analysis.

This study was supported by the French ANR MNP 2008yProBag program.

eferences

guirre GK, Detre JA, Zarahn E, Alsop DC. Experimental design and the relativesensitivity of BOLD and perfusion fMRI. Neuroimage 2002;15:488–500.

andettini PA, Jesmanowicz A, Wong EC, Hyde JS. Processing strategies for time-course data sets in functional MRI of the human brain. Magn Reson Med1993;30:161–73.

edell BJ, Narayana PA. Implementation and evaluation of a new pulse sequencefor rapid acquisition of double inversion recovery images for simultaneous sup-pression of white matter and CSF. J Magn Reson Imaging 1998;8:544–7.

enner T, van der Kouwe AJ, Kirsch JE, Sorensen AG. Real-time RF pulse adjustmentfor B0 drift correction. Magn Reson Med 2006;56:204–9.

SL, FSLView version 3.1 www.fmrib.ox.ac.uk.oense JB, Logothetis NK. Neurophysiology of the BOLD fMRI signal in awake mon-

keys. Curr Biol 2008;18:631–40.oense JB, Whittingstall K, Logothetis NK. Functional magnetic resonance imaging

of awake behaving macaques. Methods 2010;50:178–88.

ience Methods 205 (2012) 86– 95 95

Goode SD, Krishan S, Alexakis C, Mahajan R, Auer DP. Precision of cerebrovascularreactivity assessment with use of different quantification methods for hyper-capnia functional MR imaging. AJNR Am J Neuroradiol 2009;30:972–7.

Ishizawa Y. Mechanisms of anesthetic actions and the brain. J Anesth2007;21:187–99.

Joseph JE, Powell DK, Andersen AH, Bhatt RS, Dunlap MK, Foldes ST, et al. fMRI in alert,behaving monkeys: an adaptation of the human infant familiarization noveltypreference procedure. J Neurosci Methods 2006;157:10–24.

Kida I, Hyder F. Physiology of functional magnetic resonance imaging: energeticsand function. Methods Mol Med 2006;124:175–95.

Logothetis NK, Guggenberger H, Peled S, Pauls J. Functional imaging of the monkeybrain. Nat Neurosci 1999;2:555–62.

Logothetis NK, Pauls J, Augath M, Trinath T, Oeltermann A. Neurophysiological inves-tigation of the basis of the fMRI signal. Nature 2001;412:150–7.

Murnane KS, Howell LL. Development of an apparatus and methodology for con-ducting functional magnetic resonance imaging (fMRI) with pharmacologicalstimuli in conscious rhesus monkeys. J Neurosci Methods 2010;191:11–20.

Ogawa S, Lee TM, Kay AR, Tank DW. Brain magnetic resonance imaging with contrastdependent on blood oxygenation. Proc Natl Acad Sci U S A 1990;87:9868–72.

Rugg-Gunn FJ, Boulby PA, Symms MR, Barker GJ, Duncan JS. Imaging the neocortex inepilepsy with double inversion recovery imaging. Neuroimage 2006;31:39–50.

Simon B, Schmidt S, Lukas C, Gieseke J, Traber F, Knol DL, et al. Improved in vivodetection of cortical lesions in multiple sclerosis using double inversion recoveryMR imaging at 3 Tesla. Eur Radiol 2010;20:1675–83.

Sloan TB, Erian R. Effect of vecuronium-induced neuromuscular blockade on corticalmotor evoked potentials. Anesthesiology 1993;78:966–73.

SPM5, www.fil.ion.ucl.ac.uk/spm.Stefanacci L, Reber P, Costanza J, Wong E, Buxton R, Zola S, et al. fMRI of monkey

visual cortex. Neuron 1998;20:1051–7.Sudhyadhom A, Haq IU, Foote KD, Okun MS, Bova FJ. A high resolution and high

contrast MRI for differentiation of subcortical structures for DBS targeting:the Fast Gray Matter Acquisition T1 Inversion Recovery (FGATIR). Neuroimage2009;47(Suppl. 2):T44–52.

Sultan F, Augath M, Logothetis N. BOLD sensitivity to cortical activation inducedby microstimulation: comparison to visual stimulation. Magn Reson Imaging2007;25:754–9.

Tootell RB, Switkes E, Silverman MS, Hamilton SL. Functional anatomy of macaquestriate cortex. II. Retinotopic organization. J Neurosci 1988;8:1531–68.

Valette J, Guillermier M, Boumezbeur F, Poupon C, Amadon A, Hantraye P, et al.B(0) homogeneity throughout the monkey brain is strongly improved in thesphinx position as compared to the supine position. J Magn Reson Imaging2006;23:408–12.

Van Essen DC, Newsome WT, Maunsell JH. The visual field representation in stri-ate cortex of the macaque monkey: asymmetries, anisotropies, and individualvariability. Vision Res 1984;24:429–48.

Vogt KM, Ibinson JW, Schmalbrock P, Small RH. The impact of physiologic noisecorrection applied to functional MRI of pain at 1.5 and 3.0 T. Magn Reson Imaging2011;29:819–26.

Warnking J, Dojat M, Guerin-Dugue A, Delon-Martin C, Olympieff S, Richard N, et al.fMRI retinotopic mapping – step by step. Neuroimage 2002;17:1665–83.

Wey HY, Wang DJ, Duong TQ. Baseline CBF, and BOLD, CBF, and CMRO2 fMRIof visual and vibrotactile stimulations in baboons. J Cereb Blood Flow Metab2011;31:715–24.

Yan L, Zhuo Y, Ye Y, Xie SX, An J, Aguirre GK, et al. Physiological origin of low-frequency drift in blood oxygen level dependent (BOLD) functional magneticresonance imaging (fMRI). Magn Reson Med 2009;61:819–27.

The Weatherall report: the use of non-human primates in research. A working

group report chaired by Sir David Weatheall. FRS FMedSci. Academy ofMedical Sciences; Medical Research Council; The Royal Society; WellcomeTrust; December 2006. http://www.bprc.nl/BPRCE/L4/newsdownloads/The%20use%20of%20non-human%20primates%20in%20research%20-%20The%20Weatherall%20Report.pdf.