Methods 50 (2010) 125–135

Contents lists available at ScienceDirect

Methods

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

Diffusion tensor and perfusion MRI of non-human primates

Timothy Q. Duong *

Research Imaging Center and Departments of Ophthalmology, Radiology and Physiology, University of Texas Health Science Center at San Antonio, 8403 FloydCurl Drive, San Antonio, Texas 78229, USASouthwest National Primate Research Center, Southwest Foundation for Biomedical Research, South Texas Veterans Health Care System, Department of Veterans Affairs, USA

a r t i c l e i n f o

Article history:Accepted 1 August 2009Available online 7 August 2009

Keywords:Magnetic resonance imagingCerebral blood flowNeuroimagingStrokeNHPDTIBOLD fMRIArterial spin-labelingDWI

1046-2023/$ - see front matter � 2009 Elsevier Inc. Adoi:10.1016/j.ymeth.2009.08.001

Abbreviations: ADC, apparent diffusion coefficienBOLD, blood-oxygenation-level-dependent; cASL, contCBF, cerebral blood flow; DWI, diffusion weighted imimaging; EPI, echo-planar imaging; FA, fractional anisFOV, field of view; Gd-DTPA, gadolinium-diethyleneMRI, magnetic resonance imaging; NHP, non-humanSNR, signal-to-noise ratio; T1, spin–lattice relaxation ttime.

* Address: Research Imaging Center, 8403 Floyd CuUSA. Fax: +1 210 567 8152.

E-mail address: duognt@uthscsa.edu.

a b s t r a c t

This paper reviews recent non-human primate (NHP) neuroimaging literature using MRI in macaque,baboon and chimpanzee. It describes general challenges and limitations for NHP MRI studies, and reviewsrecent applications of anatomical, diffusion tensor, cerebral blood flow MRI. Applications to NHP stroke isdiscussed in some detail.

� 2009 Elsevier Inc. All rights reserved.

1. Introduction

Non-human primates (NHPs) are important animal models be-cause of their overall similarities to humans, resulting in betterrecapitulation of many human diseases compared to rodent mod-els. Research using NHP models have significantly advanced ourunderstanding of neuroscience, neurodegenerative disorders, agingand development. NHP models have played a vital role in vaccine,AIDS and infectious disease research.

Magnetic resonance imaging (MRI) has become a powerful re-search and clinical imaging tool because it can provide non-inva-sive anatomical, physiological and functional images at highspatial and temporal resolution with excellent soft tissue contrast.Many NHP studies could benefit significantly using non-invasiveMRI. A few non-human primate research centers have built theirown imaging operations or are closely affiliated with MRI centers.

ll rights reserved.

t; ASL, arterial spin-labeling;inuous arterial spin-labeling;aging; DTI, diffusion-tensor

otropy; fMRI, functional MRI;-tri-amine-penta-acetic acid;primate; RF, radiofrequency;ime; T2, spin–spin relaxation

rl Dr., San Antonio, TX 78229,

Importantly, MRI protocols developed using NHP subjects can bemore readily extended to humans because clinical scanners are of-ten used, and the size and complexity of NHP brain are moresimilar to those of human brain.

This paper reviews recent NHP MRI neuroimaging literature inmacaque, baboon and chimpanzee. It starts by describing somegeneral limitations and challenges in NHP MRI studies, followedby a review of recent applications of anatomical, diffusion tensor,cerebral blood flow MRI. Applications to NHP stroke are discussedin some detail.

2. Challenges and limitations

Some of the potential challenges and limitations in carrying outNHP MRI studies include the need for animal support, custom-made RF coils for NHP to use on human scanners, hardware con-straints on spatial resolution, and magnetic susceptibility artifacts.

2.1. Animal support

The majority of NHP MRI studies are carried out under anesthe-sia, which requires veterinary support to provide care for animalsduring anesthesia and recovery. The common choices of anesthesiaare sevoflurane, isoflurane, ketamine and propofol. The choice ofanesthetics and dosages vary depending on the types of MRI exper-iments (i.e., anatomical MRI versus functional MRI (fMRI)) and the

126 T.Q. Duong / Methods 50 (2010) 125–135

duration of the MRI exams. Stimuli for fMRI experiments involvinganesthetized NHPs are mostly limited to primary sensory systems(such as visual and somatosensory) because anesthetics suppresshigh-order cognitive functions, precluding fMRI studies of cogni-tive and behavioral task paradigms. The use of paralytics alongwith a combination of low anesthetic dose and opiates (such asfentanyl or morphine) have been successfully applied for NHP fMRIstudies [1,2].

It is necessary to monitor the physiological status of anesthe-tized NHPs during MRI experiments. Common physiologicalparameters include heart rate, respiration rate, arterial oxygen sat-uration, end-tidal CO2, blood pressure and temperature. The mon-itoring equipment used for these purposes must be MRI-compatible, should yield reliable recordings, and should not intro-duce noise or artifacts in the MR images. A set of MRI-compatiblemonitoring equipment that is cost effective and meets the de-mands we have encountered include a Surgivet’s capnograph(end-tidal CO2, inspired O2 and CO2 and respiratory waveforms)and blood pressure monitor and DigiSense temperature monitor/regulator. Standard MRI-compatible electrode-based EKG systemsfrom human MRI scanners work well on NHP subjects. Optical-based oximetry to derive arterial oxygen saturation and heart rateis unreliable on NHP in our experience likely because of theopaqueness of NHP skin. Multifunctional, MRI-compatible moni-toring devices that can monitor EKG, oximetry, end-tidal CO2, in-spired gas concentration, respiration waveforms and rate, bloodpressure and temperature in a single device are commerciallyavailable for human MRI studies. However these units are signifi-cantly costlier than systems built from selected individualcomponents.

While the majority of NHP MRI studies are carried out underanesthetized conditions, some are carried out under awake andbehaving conditions [1,3,4]. Awake NHP studies allow mappingof activation sites, response latencies, synchronicity of activationand behavioral correlates involved in higher order functions. Thephysiological monitoring need not be as extensive as in anesthe-tized preparations because awake subjects can better regulatetheir own physiology. Awake NHP studies have been reportedusing macaque [3,5–7] and marmoset [8,9].

2.2. MRI hardware

MRI scanners specifically designed for imaging primates aresparse and there are only a few in operation to date [5,6]. Someanimal scanners offer gradients with a clear bore of 15–25 cm in-ner diameter which could accommodate a small anesthetized ma-caque. Human scanners however, are more widely used forimaging large (adult macaque and larger species) NHPs [3,4,10].The large scanner size and the ease of use for many commonlyavailable MRI protocols make human MRI systems more facile thansmall-animal imaging systems for translational aspects of NHPimaging research.

The smaller NHP brain volume, compared to human brain vol-ume, make it necessary to acquire images with higher spatial res-olution in order to achieve comparable parcellation of anatomicaland functional structures. Brain weights for different species aresummarized in Table 1 [11,12]. Assuming cubic volume, we esti-mated the relative spatial resolution (voxel size) for each of the

Table 1Comparison of brain sizes and comparable pixel resolution of different species.

Brain Human Chimpanzee

Weight (g) 1400 420Pixel resolution (lm) 1000 670

species in the table below. Achieving these spatial resolutions forNHP on clinical scanners could be challenging, especially for proto-cols other than anatomical MRI. For example, a1000 � 1000 � 1000 lm voxel for human brain MRI occupies acomparable fractional brain volume to a 460 � 460 � 460 lm vox-el for baboon. Achieving allometric spatial resolution for a range ofNHP species using a clinical scanner presents a unique set ofchallenges.

Strong applied magnetic field gradients are necessary to achievehigh spatial resolution (i.e., to increase matrix sizes, reduce FOVand decrease slice thickness). Gradient strengths achievable on hu-man scanners are typically 20–60 mT/m (per axis), compared to400–1500 mT/m (per axis) commonly available on small-animalMRI systems. In addition to spatial resolution, some imaging proto-cols (such as diffusion-tensor imaging and echo-planar imaging)on human are limited by the maximum achievable gradientstrength. MRI vendors do not have financial incentive to make cus-tom gradients dedicated for NHP studies. The majority of MRI cen-ters do not have the capability to make custom gradient inserts.However, with the increasing need to push for higher spatial reso-lution in humans, head-only gradient systems with capability toproduce �80 mT/m magnetic field gradients are becoming com-mercially available. NHP MRI could also benefit significantly withthese head-only gradient inserts. It is expected that this trend willfacilitate future NHP MRI research.

Another factor that makes high spatial resolution NHP MRIchallenging is the available signal-to-noise ratio (SNR). To achieveoptimal SNR, RF coils should have the maximum possible fillingfactor – that is the RF coils should be the smallest possible thatcan cover the organs of interest. A limitation of NHP MRI studieson human scanners is that RF coils of different sizes may not bereadily available. The few standard human RF coils may not beideal for various NHP sizes. Thus, centers with a substantial NHPMRI research portfolio may consider developing their own RF coilfabrication capability.

Magnetic susceptibility and signal drop off around the frontallobe and ear canal are generally worse for NHP brain than humanbrain because NHP brain has small volume-to-surface ratio. Theshimming capability on human MRI scanners is poorer comparedto animal scanners. Therefore, susceptibility-induced signal lossand distortion around the frontal lobe and ear canals, for example,are found to be substantially worse in DTI study of chimpanzeebrain compared to human brain using typical imaging protocols,such as single-shot echo-planar imaging (EPI) with a matrix sizeof 64 � 64. We explored a few approaches that successfully mini-mize distortion and signal drop off, and these are briefly describedbelow. High-resolution field mapping improves EPI quality [13,14].Segmented EPI could be used to reduce readout time and echo timewhich reduce distortion and signal drop off, respectively, but at theexpense of longer scanner time and increased segmentation arti-facts. Reconstructing two sets of images acquired with oppositephase encoding directions [15] is helpful in removing spatial dis-tortion but it doubles the imaging time. However, if signal averag-ing is needed, this is not an issue. Multiple thinner imaging slicescan be employed to reduce intravoxel dephasing (and thus signaldrop off) and these slices can be summed to recover SNR [15]. Thisapproach may be confounded by slice selection imperfection, andcould lengthen TR, and thus, total acquisition time. Finally, parallel

Baboon Macaque Marmoset Rat

137 95 7 2.2460 410 170 116

T.Q. Duong / Methods 50 (2010) 125–135 127

imaging [16] to reduce EPI readout time is helpful at the expense ofSNR. These approaches can also be used in combination for severalapplications.

3. Anatomical and diffusion tensor MRI

Anatomical contrast in MRI arise from differences in spin den-sity, spin–lattice relaxation time (T1), spin–spin relaxation time(T2) and properties of translational diffusion of water in tissue.These MRI parameters are dependent on the local tissue environ-ment including water content, cellular structure, macromoleculecontent, and ion concentrations. Many diseases alter these bio-physical parameters, causing visible changes in image contrast.These changes may occur in early stages of disease progression,prior to clinical symptom onset. Anatomical NHP MRI has beenused for surgical planning, to guide stereotaxic brain injection, vol-umetric studies of brain anatomy and anatomical changes associ-ated with development, aging, sex and species differences, aswell as several other applications. Fig. 1 shows an assortment ofhigh-resolution anatomical images of rhesus and chimpanzeebrains. T1-weighted, T2-weighted, FLAIR (fluid-attenuated inver-sion-recovery) and fractional anisotropy MRI provide excellentcontrast both in vivo and post-mortem. Anatomical MRI is re-viewed elsewhere in this special issue and will not be further dis-cussed here.

Diffusion-tensor imaging (DTI) [17,18] has become a widelyused tool for non-invasive imaging of anatomical connectivity[19–22] and brain microstructural morphology [23–25]. DTI con-trast arises from barriers (such as cell membranes of axons and oli-godendrocytes) that hinder water diffusion in some orientationsmore than others, giving rise to anisotropic diffusion. In white mat-ter, for example, water diffusion perpendicular to fiber tracts ismore restricted than that parallel to the fiber tracts [26,27]. DTIhas been used for detecting changes in myelination in the develop-ing brain [28–30] and in demyelinating diseases [31,32].

Fig. 1. Anatomical images of (Top) rhesus and (bottom) chimpanzee brains. T1W, T1-weianisotropy.

Although there have been numerous DTI studies in humans androdents, DTI studies of NHP are sparse by comparison. Only a smallnumber of NHP studies have been reported [33–42] and many ofthose utilized post-mortem brain tissue. Fig. 2 shows a typicalDTI tractography data from our laboratory obtained on a macaquebrain at 3T.

D’Arceuil and co-workers [43] investigated very high-SNR andhigh-resolution DTI of fixed macaque brain (Fig. 3). They modifiedthe water relaxation properties within fixed tissue to obtain condi-tions most compatible with constraints imposed by the MRI systemhardware. Relaxivity measurements in gray and white matter al-lowed optimization of Gd-DTPA (gadolinium diethylenetriaminepenta-acetic acid) concentration with respect to SNR, resulting in a2-fold improvement. Fractional anisotropy (FA) changed little withGd concentrations of up to 10 mM, although the ADC was reducedby the presence of 5–10 mM contrast agent. Comparison of in vivo,fresh ex vivo and fixed brains showed no significant FA changesbut reductions in ADC of about 50% in fresh ex vivo, and 64% and80% in fixed gray and white matter respectively. They also studiedthe temperature dependence of diffusion in these tissues and thedata suggest that a 30� increase in sample temperature may yieldan improvement of up to 55% in SNR for a given diffusion weighting.

Cerebral cortical development involves complex changes in cel-lular architecture and connectivity that occur at regionally varyingrates. Kroenke et al. [40] investigated gestational development infixed post-mortem baboon brain using DTI (Fig. 4) and characterizedregional changes in diffusion anisotropy using surface-based visual-ization methods. They concluded that anisotropy values vary acrossthe thickness of the cortical wall, being higher in superficial layers.Further, cortical diffusion anisotropy exhibits a regional spatialdependence, and this pattern potentially parallels expansion of cor-tical surface, which does not occur uniformly in all regions.

Rilling et al. [44] used DTI to perform comparative studies on hu-man, chimpanzee and macaque (Fig. 5). They focused on the arcuatefasciculus, a white matter fiber tract which, in humans, is involved inhuman language processing. They found that the prominent tempo-ral lobe projection of the human arcuate fasciculus was much

ghted; T2W, T2-weighted; FLAIR, fluid-attenuated inversion-recovery; FA, fractional

Fig. 2. Fiber tracking of the corpus callosum (A) and whole-brain (B) of a macaque at 3T.

Fig. 3. Orthogonal planes through high resolution (425 l) trace ADC maps, direction encoded Fractional Anisotropy (FA) maps and gradient echo FLASH (175 l) images of afixed macaque brain acquired at 4.7 Tesla. Voxel color in FA maps encode principle eigenvector directions in the usual way (red = left–right, green = anterior–posterior,blue = Superior–inferior). [43].

128 T.Q. Duong / Methods 50 (2010) 125–135

smaller or absent in non-human primates. They concluded that thishuman specialization may be relevant to the evolution of language.

Liu et al. [36] reported an in vivo macaque DTI study at highspatial resolution (1 � 1 � 1 mm) and evaluated variants on theiracquisition protocol, susceptibility distortion correction method,and image resolution. They concluded that in vivo high-resolutionmonkey brain DTI on a clinical 3-Tesla scanner can be achievedwithin practical time. Other studies also explored more advancedDTI tracking algorithms in NHP, such as diffusion spectrum imag-ing tractography of crossing fibers [45].

In sum, NHP DTI studies offer unique perspectives on develop-ment and comparative white matter connectivity across species.

4. Blood flow MRI

Non-invasive cerebral blood flow (CBF) measurements usingMRI are widely used to study normal physiology and pathophysi-ology. Quantitative CBF can be obtained at high temporal and spa-tial resolution. Functional MRI based on changes in CBF is spatially

Fig. 4. Heterogeneity in isocortical diffusion anisotropy at E125. In each image plane, the intersection of the right hemisphere isocortical surface is shown in blue. (a), Laminar(light blue arrow) and regional (compare regions near yellow and red arrows) patterns of anisotropy variation are observable. Regional variation is also apparent in thecoronal views (b, c; positions of these image planes are indicated by dashed lines in a, d and e). For example, anisotropy caudal/inferior to the STS is larger than rostral/superior to the STS. The dotted blue line in e indicates the medial extent of the parahippocampal gyrus.

Fig. 5. Color maps of principal diffusion direction in one in vivo human, a post-mortem chimpanzee and a post-mortem macaque macaque brain (sagittal view). Yellow arrowpoints to red, mediolaterally oriented fibers in chimpanzee brain [44].

T.Q. Duong / Methods 50 (2010) 125–135 129

more specific to the site of increased neural activity than tradi-tional blood-oxygenation-level-dependent (BOLD) methods, capa-ble of resolving cortical columns [46], which makes CBF-baseddata easier to interpret than the BOLD fMRI signals. CBF fMRI isalso less susceptible to pathologic perturbation, and therefore isassociated with less inter-subject and cross-day variability [47].Combined cerebral blood flow and BOLD fMRI measurements offerthe means to estimate the stimulus-evoked changes in cerebralmetabolic rate of oxygen metabolism [48–50]. The main draw-backs of CBF fMRI measurements are relatively low temporal reso-lution, arising from low SNR per unit time achievable for thistechnique, and greater vulnerability to motion artifacts [47,51].

CBF can be measured by using an exogenous intravascular con-trast agent or by magnetically labeling the endogenous water inblood [47,51]. The former is efficient but it is incompatible withdynamic fMRI experiments because the long half-life of the con-trast agent allows only one CBF measurement per bolus injection.Arterial spin-labeling (ASL) techniques, on the other hand, utilizeendogenous compounds, and the labeled molecular tracer (magne-tization-prepared water) has a short half-life (�blood water T1)making it possible to perform multiple repeated measurements,which can be used to augment spatial resolution and/or signal-to-noise ratio. ASL is compatible with dynamic CBF fMRI studies.

ASL can be performed using pulsed labeling [52–54] or contin-uous labeling [55–57]; both are capable of multislice and whole-

brain imaging. Continuous ASL (cASL) can be achieved with thesame radiofrequency (RF) coil used for imaging or a separate neckcoil. cASL with a separate neck coil is generally more sensitive rel-ative to the single-coil technique [52,53,58], particularly in smallanimals such as rodents which have a short arterial transit time[59,60]. Another advantage of utilizing a separate neck coil is thatpotential confounds associated with magnetization-transfer effectsare eliminated, provided the coils are properly decoupled. This re-sults in a larger signal difference between labeled and non-labeledimages, and thus improved CBF SNR. RF power deposition is local-ized to the neck area and unlabeled images can be acquired with-out labeling RF, reducing the specific absorption rate (SAR) [61].While the cASL technique for measuring quantitative basal CBFand CBF-based MRI is more readily available on animal scannersfor rodent imaging [59,60], similar studies on humans and largenon-human primates are sparse because clinical scanners gener-ally lack the necessary hardware and software. cASL using a sepa-rate neck coil has been reported on General Electric scanners forhuman studies [56,57]. The typical spatial resolution of basal ASLCBF measurements on human scanners was 70 mm3 [56,57] whichmay be inadequate for NHP MRI. CBF-based fMRI using cASL with aseparate neck coil on human scanners remains to be demonstrated.

We implemented a three-coil arterial spin-labeling techniqueon a Siemens 3-Tesla Trio clinical scanner for non-human primate(macaque monkey) studies at high spatial resolution up to 3.4 mm3

Fig. 7. Labeling efficiency. DS/S images at different labeling RF powers and theangiograms are shown. The ROI’s of the arteries were used to quantify labelingefficiency. The optimal labeling power was 2 W which yielded 92% labelingefficiency [10].

130 T.Q. Duong / Methods 50 (2010) 125–135

[10]. To overcome the difficulty and safety concerns in re-configur-ing hardware clinical MRI system, a stand-alone hardware unit forcASL using a separate neck coil was constructed. Hardware compo-nents which included an external RF amplifier, control electronics,optical-electrical relays, active decoupling circuits and radiofre-quency probes were built, interfaced and tested on a Siemens 3-Te-sla Trio. Fig. 6 shows the schematic layout of the stand-alone ASLsetup, placement of the RF coils and labeled arteries.

Fig. 7 shows the labeling efficiency curve. The optimal labelingpower was 2 W, labeling efficiency was 92 ± 2%. Through a series ofvariable post-labeling delay measurements, the optimal post-labeling delay was determined to be 0.8 s. Whole-brain averageDS/S was 1.0–1.5%. Fig. 8 shows the multislice quantitative CBFimages obtained in 3 min with 1.5-mm isotropic resolution. GMCBF was 104 ± 3 ml/100 g/min (n = 6, SD) and WM CBF was45 ± 6 ml/100 g/min in isoflurane-anesthetized macaque monkeys,with the CBF GM/WM ratio of 2.3 ± 0.2.

Combined CBF and BOLD (blood-oxygenation-level-dependent)fMRI associated with hypercapnia and hyperoxia were made with8-s temporal resolution (Fig. 9). CBF fMRI responses to 5% CO2 were59 ± 10% (GM) and 37 ± 4% (WM); BOLD fMRI responses were2.0 ± 0.4% (GM) and 1.2 ± 0.4% (WM). CBF fMRI responses to 100%O2 were �9.4 ± 2% (GM) and �3.9 ± 2.6% (WM); BOLD responseswere 2.4 ± 0.7% (GM) and 0.8 ± 0.2% (WM). Hypercapnia is knownto cause vasodilation which increases both CBF and BOLD signals.Hyperoxia is known to cause vasoconstriction which decreasesCBF but increase BOLD signals.

The CBF GM to WM ratio was 2.3 which is within the ranges re-ported previously of 1.7 [57], 2.7 [62] 3 [63] by MRI, and 2.0 by PET[64]. The GM CBF was 104 ± 3 ml/100 g/min and WM CBF was45 ± 6 ml/100 g/min in isoflurane-anesthetized monkeys. Dynamiccontrast-enhanced MRI only reported a CBF GM:WM ratio in anes-thetized monkeys [63]. PET reported quantitative CBF of 23–43 ml/

Fig. 6. (A) Schematics of a stand-alone unit for the three-coil arterial-spin-labeling techimage of a monkey head and neck using a ‘‘volume coil” for demonstration of the approxcarotid and vertebral arteries. Neck coil was used as the transceiver [10].

100 g/min (whole-brain) in propofol-anesthetized monkeys [65],and GM CBF value of 56–68 ml/100 g/min in the cerebral cortexand a WM CBF value of 34 ml/100 g/min in ketamine-anesthetizedmonkeys [66]. Our CBF values in isoflurane-anesthetized monkeyswere high compared to those in awake humans and anesthetizedmonkeys. The major cause of such difference is likely due to isoflu-rane anesthetic which is an established potent vasodilator. Simi-larly high CBF under isoflurane [67,68] relative to otheranesthetics [60] and awake conditions [68] have been well docu-mented in rodents (isoflurane:awake CBF ratio in rat was 1.48

nique. (B) A picture of the head-holder apparatus with neck and brain coils. (C) MRimate positions of the coils only. (D) Cross-sectional image of the neck showing the

Fig. 8. Quantitative CBF images using the optimized parameters [10].

T.Q. Duong / Methods 50 (2010) 125–135 131

[60,68] and the isoflurane:a-chloralose CBF ratio was 1.84 [68]).NHP CBF measurements have also been reported on a Bruker scan-ner using the built-in second RF channel which is offered by thevendor [69,70].

The use of a separate neck coil for spin-labeling significantly in-creased CBF signal-to-noise ratio and the use of small receive-onlysurface coil significantly increase image signal-to-noise ratio andspatial resolution. ASL CBF measurements are helpful to manyNHP studies such as stroke MRI as described below.

Fig. 9. Single-animal BOLD and CBF fMRI of (A) hypercapnic and (B) hyperoxic challeheterogeneous with larger changes in the gray matter and smaller changes in the white

5. NHP MRI studies of stroke

5.1. Stroke in the rhesus macaque

Stroke is the third leading cause of death and the leading causeof long-term disability. It has been estimated that 5.8 millionAmericans have permanent neurological deficits from stroke, andthat 71% of these stroke survivors cannot return to work. TheAmerican Heart Association estimated that $65.5 billion will be ex-pended on the care of stroke patients in 2008. The cost is steadilyrising because the conditions that put people at risk for stroke(such as heart disease and obesity) are also steadily on the rise. De-spite the tremendous effort invested in ischemic stroke research,our ability to minimize neurological deficit by protecting ischemicneurons in stroke patients remains extremely limited.

While numerous neuroprotective drugs have shown positiveresults on experimental rodent stroke models, none has provento be effective clinically. One widely accepted view to accountfor this disappointing outcome is that the rodent stroke modelsimply does not adequately reflect the complexity of humanstroke. The Stroke Therapy Academic Industry Roundtable (STAIR),a committee commissioned by the American Heart Association hasrecommended that clinically relevant non-human primate strokemodels be established for developing and assessing neuroprotec-tive drugs.

Few MRI studies have focused on experimental NHP stroke, andmost studies that have been performed used conventional MRI(such as T2 and FLAIR) to characterize the non-acute phase of in-jury, at a single time point [71–73], often at an endpoint to quan-tify lesion volume. A MRS NHP stroke study has been reported [74].Multimodality MRI to serially image acute NHP stroke is onlyreported recently [75]. In these studies, ischemia was induced in

nge. Color bar indicates cross-correlation coefficients. BOLD and CBF changes arematter [10].

Fig. 10. (a) Serial MRI data (1 slice) for an animal with a 3-h middle cerebral artery occlusion as a function of time after occlusion. T2-weigthed MRI shows subtlehyperintensity soon after reperfusion, while ADC is clearly reduced hyperacutely. (b) Serial MRI data (1 slice) for an animal with permanent middle cerebral artery occlusion,showing minimal hyperacute T2 elevation but more severe and longer duration acute ADC decrease [75].

Fig. 12. MR images of the chimpanzee brain before and during the stroke. (A–D): An anterior–posterior series of coronal, T1-weighted MR images made 10 years prior tostroke. (E–H): T2-weighted MR images captured after the symptoms had emerged, with the location of affected tissue in the left hemisphere indicated by asterisks. (A T2-weighted scan was not run prior to the stroke.) [80].

Fig. 11. (A) Anatomy, CBF and ADC at different time after permanent occlusion (embolic model). Resolution: 750 � 750 � 1500 l.

132 T.Q. Duong / Methods 50 (2010) 125–135

Fig. 13. A naturally occurring stroke in a chimpanzee. High intensity lesion wasdetected over left middle cerebral artery territory. This includes left temporal(inferior and middle temporal artery), left frontal (operculum artery) and left basalganglia (perforators from M1).

T.Q. Duong / Methods 50 (2010) 125–135 133

macaque in which stroke is induced using an endovascular tech-nique. Fig. 10 shows the T2, FLAIR, ADC and FA images from 3-hand permanent occlusion in macaque. These results demonstratethat stroke can be reproducibly studied in a clinically more rele-

Fig. 14. Tau and Ab pathology in the aged chimpanzee. (A): Tau-immunoreactive somatathe left prefrontal cortex. Numerous tau-immunoreactive processes (threads) occupy thein area CA1 of the right hippocampus; note the integrity of the nearby pyramidal cells. Antau (CP13/nickel-DAB, black, developed first) and Ab (R398/DAB, brown) immunoreacpyramidal neuron (arrowhead) are present. A focus of Ab-immunoreactive CAA is indicblood vessels) and a senile plaque (arrow) in the right temporal neocortex. Antibody 6E

vant animal model in which the precise stroke onset is known.NHP stroke model combined with clinically relevant MRI ap-proaches could positively impact stroke research.

We have performed similar studies, with the addition of serialperfusion MRI using arterial spin-labeling technique and echo-pla-nar imaging to image the perfusion–diffusion mismatch. The per-fusion–diffusion mismatch approximates the ischemic penumbra.We employed an embolic stroke model using endovascular tech-nique similar to that described in [75]. Fig. 11 shows a perfusion(CBF) and diffusion (ADC) image from one animal (coronal view).The CBF image was obtained using the arterial spin-labeling tech-nique [10,53]. The region of perfusion deficit after stroke inductioncould be clearly delineated. CBF-defined lesion volume did notchange over time for up to 6 h (permanent occlusion) and thus onlythe CBF image at 1.5 h after stroke is shown. In contrast, the ADC-defined lesion grew progressively until it matched the CBF deficitvolume, in good agreement with observations in rats [76,77] andhumans [78,79] albeit the temporal evolution of the perfusion–dif-fusion mismatch differred among the three species. Differences inischemic evolution, as well as others species-dependent character-istics, underscore the importance of NHP stroke model.

5.2. Stroke in a chimpanzee (a natural stroke with Alzheimer’s disease-like deposits)

A 41-year-old, socially housed female chimpanzee (Pan troglo-dytes; CO494) spontaneously developed acute lethargy and rapidlyprogressive unilateral motor dysfunction, indicative of stroke ofthe middle cerebral artery territory. The animal was anesthetized,intubated and MRI was performed. Physiological parameters heartrate, respiration rate and arterial O2 saturation were within normalphysiological ranges and were similar to other normal animals. AT2-weighted MRI revealed a massive, left-hemispheric lesion

(two indicated by arrowheads) and neuritic tau plaques (one indicated by arrow) inintervening neuropil. Antibody AT 8. (B): Tau-immunoreactive neuron (arrowhead)tibody AT 8, hematoxylin counterstain. (C) Double-immunostained section showingtivity in the left prefrontal cortex. A tau plaque (black arrow) and a taupositive

ated by the gray arrow. (D): Cerebral amyloid angiopathy (arrowheads denote two10. Scale bars 100 lm in A, C; 50 lm in B; 200 lm in D. [80].

134 T.Q. Duong / Methods 50 (2010) 125–135

involving mainly the temporal, parietal, and occipital lobes with asubstantial shifted in the midline (Fig. 12) [80]. In contrast, T1-weighted MRI of the brain performed 10 years earlier on the sameanimal showed no apparent abnormalities. Diffusion-weightedMRI lesion volume obtained using echo-planar imaging matchedthose of the T2-weighted images (Fig. 13) and did not evolve overthe duration of the MRI exam (�2 h). Post-mortem staining for tis-sue viability using 2,3,5 triphenyl tetrazolium chloride (TTC) ofroughly the same slices confirmed infraction. Together with thetiming of discovery of clinical symptoms and behavioral assess-ment, the lesion was estimated to be 6–12 h old at the time ofMRI. Gross examination of the post-mortem brain confirmed a sub-stantial region of ischemic, non-hemorrhagic necrosis in the lefthemisphere. The right hemisphere was grossly normal. The previ-ous medical history was unremarkable except for a chronic systolicheart murmur first diagnosed at 15 years of age, moderate obesity(weight at death 61.5 kg), and high serum cholesterol [80].

Interestingly, in addition to stroke pathology, histological anal-ysis showed tauopathy with paired helical filaments in an agedchimpanzee (Fig. 14) [80]. The neurodegenerative tauopathies aretypified by the intracellular aggregation of hyperphosphorylatedmicrotubule-associated protein tau and the dysfunction and deathof affected neurons. The neurofibrillary tangles consisted of tau-immunoreactive paired helical filaments with a diameter and heli-cal periodicity indistinguishable from those of Alzheimer’s disease.A moderate degree of Ab deposition was present in the cerebralvasculature and in senile plaques. The co-presence of paired helicalfilaments and Ab-amyloidosis indicates that the molecular mecha-nisms for the pathogenesis of the two canonical Alzheimer lesions– neurofibrillary tangles and senile plaques – are present in agedchimpanzees. This is the first documented case of tauopathy withhelical filaments in NHP. Given Alzheimer’s disease is consideredan uniquely human disorder, these findings compel us to recon-sider the assumption that humans are the only primates to mani-fest Alzheimer-like tauopathy with age.

6. Concluding remarks

This review summarizes recent development and application ofMRI to study NHP brains. Non-invasive MRI can clearly provideclinically useful structural, physiological and functional data in asingle setting at very high spatiotemporal resolution. There arehowever some significant challenges in carrying out MRI studieson NHP. The main challenges are associated with (1) the need foranesthetized preparations which preclude many fMRI studies ofcognitive functions, and (2) the use of human scanners which aresuboptimal for NHP due to size differences. We anticipate thatcontinuing advances in MRI technologies, such as improving cus-tom-made RF coil detectors and gradient capabilities, will improveand expand NHP MRI applications.

Acknowledgments

This work is funded in part by the American Heart Association’sEstablished Investigator Award EIA 0940104N. The author wouldlike to thank Drs. Frank Tong, Yoji Tanaka, Tsukasa Nagaoka fortheir contributions to the stroke project from which Fig. 11 was de-rived. I would also like to thank Drs. Helen D’Arceuil and Alex deCrespigny for numerous discussions on the NHP stroke project.

References

[1] N.K. Logothetis, J. Pauls, M. Augath, T. Trinath, A. Oeltermann, Nature 412(2001) 150–157.

[2] N.K. Logothetis, Philos. Trans. R. Soc. Lond. B Biol. Sci. 357 (2002) 1003–1037.[3] A.H. Andersen, Z. Zhang, T. Barber, W.S. Rayens, J. Zhang, R. Grondin, P. Hardy,

G.A. Gerhardt, D.M. Gash, J. Neurosci. Methods 118 (2002) 141–152.

[4] F.P. Leite, D. Tsao, W. Vanduffel, D. Fize, Y. Sasaki, L.L. Wald, A.M. Dale, K.K.Kwong, G.A. Orban, B.R. Rosen, R.B. Tootell, J.B. Mandeville, Neuroimage 16(2002) 283–294.

[5] N.K. Logothetis, H. Guggenberger, S. Peled, J. Pauls, Nat. Neurosci. 2 (1999)555–562.

[6] P.D. Gamlin, M.K. Ward, M.S. Bolding, J.K. Grossmann, D.B. Twieg, Methods 38(2006) 210–220.

[7] W. Vanduffel, D. Fize, J.B. Mandeville, K. Nelissen, P. Van Hecke, B.R. Rosen, R.B.Tootell, G.A. Orban, Neuron 32 (2001) 565–577.

[8] C.F. Ferris, C.T. Snowdon, J.A. King, T.Q. Duong, T.E. Ziegler, K. Ugurbil, R.Ludwig, N.J. Schultz-Darken, Z. Wu, D.P. Olson, J.M. Sullivan Jr, P.L.Tannenbaum, J.T. Vaughan, Neuroreport 12 (2001) 2231–2236.

[9] C.F. Ferris, C.T. Snowdon, J.A. King, J.M. Sullivan Jr., T.E. Ziegler, D.P. Olson, N.J.Schultz-Darken, P.L. Tannenbaum, R. Ludwig, Z. Wu, A. Einspanier, J.T.Vaughan, T.Q. Duong, J. Magn. Reson. Imaging 19 (2004) 168–175.

[10] X. Zhang, T. Nagaoka, E.J. Auerbach, R. Champion, L. Zhou, X. Hu, T.Q. Duong,Neuroimage 34 (2007) 1074–1083.

[11] H. Stephan, H. Frahm, G. Baron, Folia Primatol. (Basel) 35 (1981) 1–29.[12] H. Stelkis, J. Erwin, in: Comparative Primate Biology, Alan R. Liss, New York,

1988.[13] F. Lamberton, N. Delcroix, D. Grenier, B. Mazoyer, M. Joliot, A.N. Priest, E. De

Vita, D.L. Thomas, R.J. Ordidge, P.J. Reber, E.C. Wong, R.B. Buxton, L.R. Frank, J.Magn. Reson. Imaging 26 (2007) 747–755.

[14] P.J. Reber, E.C. Wong, R.B. Buxton, L.R. Frank, Magn. Reson. Med. 39 (1998)328–330.

[15] G. Liu, S. Ogawa, J. Magn. Reson. Imaging 24 (2006) 683–689.[16] K.P. Pruessmann, Top Magn. Reson. Imaging 15 (2004) 237–244.[17] P.J. Basser, J. Mattiello, D. LeBihan, Biophys. J. 66 (1994) 259–267.[18] P.J. Basser, J. Mattiello, D. LeBihan, J. Magn. Reson. B 103 (1994) 247–254.[19] K. Mori, Curr. Opin. Neurobiol. 5 (1995) 467–474.[20] R. Xue, P.C. van Zijl, B.J. Crain, M. Solaiyappan, S. Mori, Magn. Reson. Med. 42

(1999) 1123–1127.[21] T. Nakada, H. Matsuzawa, Neurosci. Res. 22 (1995) 389–398.[22] N. Makris, A.J. Worth, A.G. Sorensen, G.M. Papadimitriou, O. Wu, T.G. Reese, V.J.

Wedeen, T.L. Davis, J.W. Stakes, V.S. Caviness, E. Kaplan, B.R. Rosen, D.N.Pandya, D.N. Kennedy, Ann. Neurol. 42 (1997) 951–962.

[23] J. Zhang, P.C. van Zijl, S. Mori, Neuroimage 15 (2002) 892–901.[24] S. Mori, R. Itoh, J. Zhang, W.E. Kaufmann, P.C. van Zijl, M. Solaiyappan, P.

Yarowsky, Magn. Reson. Med. 46 (2001) 18–23.[25] S. Mori, R. Itoh, W.E. Kaufmann, P.C.M. van Zijl, M. Solaiyappan, P. Yarowsky,

Magn. Reson. Med. 46 (2001) 18–23.[26] D. Le Bihan, R. Turner, P. Douek, Neuroreport 4 (1993) 887–890.[27] C. Beaulieu, P.S. Allen, Magn. Reson. Med. 31 (1994) 394–400.[28] D.M. Wimberger, T.P. Roberts, A.J. Barkovich, L.M. Prayer, M.E. Moseley, J.

Kucharczyk, J. Comput. Assist. Tomogr. 19 (1995) 28–33.[29] D. Prayer, T. Roberts, A.J. Barkovich, L. Prayer, J. Kucharczyk, M. Moseley, A.

Arieff, Neuroradiology 39 (1997) 320–325.[30] J.J. Neil, S.I. Shiran, R.C. McKinstry, G.L. Schefft, A.Z. Snyder, C.R. Almli, E.

Akbudak, J.A. Aronovitz, J.P. Miller, B.C. Lee, T.E. Conturo, Radiology 209 (1998)57–66.

[31] A.C. Guo, J.R. Petrella, J. Kurtzberg, J.M. Provenzale, Radiology 218 (2001) 809–815.

[32] E.M. Larsson, E. Englund, M. Sjobeck, J. Latt, S. Brockstedt, Dement. Geriatr.Cogn. Disord. 17 (2004) 316–319.

[33] H. D’Arceuil, A. de Crespigny, Neuroimage 36 (2007) 64–68.[34] J. Dauguet, S. Peled, V. Berezovskii, T. Delzescaux, S.K. Warfield, R. Born, C.F.

Westin, Neuroimage 37 (2007) 530–538.[35] S. Hofer, K.D. Merboldt, R. Tammer, J. Frahm, Cereb. Cortex 18 (2008) 1079–

1084.[36] X. Liu, T. Zhu, T. Gu, J. Zhong, Magn. Reson. Imaging 27 (2009) 335–346.[37] J.D. Schmahmann, D.N. Pandya, R. Wang, G. Dai, H.E. D’Arceuil, A.J. de

Crespigny, V.J. Wedeen, Brain 130 (2007) 630–653.[38] C.D. Kroenke, G.L. Bretthorst, T.E. Inder, J.J. Neil, Neuroimage 25 (2005) 1205–

1213.[39] C.D. Kroenke, G.L. Bretthorst, T.E. Inder, J.J. Neil, Magn. Reson. Med. 55 (2006)

187–197.[40] C.D. Kroenke, D.C. Van Essen, T.E. Inder, S. Rees, G.L. Bretthorst, J.J. Neil, J.

Neurosci. 27 (2007) 12506–12515.[41] N. Makris, G.M. Papadimitriou, A. van der Kouwe, D.N. Kennedy, S.M.

Hodge, A.M. Dale, T. Benner, L.L. Wald, O. Wu, D.S. Tuch, V.S. Caviness,T.L. Moore, R.J. Killiany, M.B. Moss, D.L. Rosene, Neurobiol. Aging 28(2007) 1556–1567.

[42] G.J. Parker, K.E. Stephan, G.J. Barker, J.B. Rowe, D.G. MacManus, C.A. Wheeler-Kingshott, O. Ciccarelli, R.E. Passingham, R.L. Spinks, R.N. Lemon, R. Turner,Neuroimage 15 (2002) 797–809.

[43] H.E. D’Arceuil, S. Westmoreland, A.J. de Crespigny, Neuroimage 35 (2007) 553–565.

[44] J.K. Rilling, M.F. Glasser, T.M. Preuss, X. Ma, T. Zhao, X. Hu, T.E. Behrens, Nat.Neurosci. 11 (2008) 426–428.

[45] V.J. Wedeen, R.P. Wang, J.D. Schmahmann, T. Benner, W.Y. Tseng, G. Dai, D.N.Pandya, P. Hagmann, H. D’Arceuil, A.J. de Crespigny, Neuroimage 41 (2008)1267–1277.

[46] T.Q. Duong, D.-S. Kim, K. Ugurbil, S.-G. Kim, Proc. Natl. Acad. Sci. USA 98 (2001)10904–10909.

[47] E.L. Barbier, L. Lamalle, M. Decorps, J. Magn. Reson. Imaging 13 (2001) 496–520.

T.Q. Duong / Methods 50 (2010) 125–135 135

[48] T.L. Davis, K.K. Kwong, R.M. Weisskoff, B.R. Rosen, Proc. Natl. Acad. Sci. USA 95(1998) 1834–1839.

[49] S.-G. Kim, K. Ugurbil, Magn. Reson. Med. 38 (1997) 59–65.[50] R.D. Hoge, J. Atkinson, B. Gill, G.R. Crelier, S. Marrett, G.B. Pike, Proc. Natl. Acad.

Sci. 96 (1999) 9403–9408.[51] F. Calamante, D.L. Thomas, G.S. Pell, J. Wiersma, R. Turner, J. Cereb. Blood Flow

Metab. 19 (1999) 701–735.[52] E.C. Wong, R.B. Buxton, L.R. Frank, Magn. Reson. Med. 39 (1998) 702–708.[53] J. Wang, Y. Zhang, R.L. Wolf, A.C. Roc, D.C. Alsop, J.A. Detre, Radiology 235

(2005) 218–228.[54] J.A. Detre, W. Zhang, D.A. Roberts, A.C. Silva, D.S. Williams, D.J. Grandis, A.P.

Koretsky, J.S. Leigh, NMR Biomed. 7 (1994) 75–82.[55] A.C. Silva, W. Zhang, D.S. Williams, A.P. Koretsky, Magn. Reson. Med. 33 (1995)

209–214.[56] G. Zaharchuk, P.J. Ledden, K.K. Kwong, T.G. Reese, B.R. Rosen, L.L. Wald, Magn.

Reson. Med. 41 (1999) 1093–1098.[57] S.L. Talagala, F.Q. Ye, P.J. Ledden, S. Chesnick, Magn. Reson. Med. 52 (2004)

131–140.[58] S.-G. Kim, Magn. Reson. Med. 34 (1995) 293–301.[59] A. Silva, S.-P. Lee, C. Yang, C. Iadecola, S.-G. Kim, J. Cereb. Blood Flow Metab. 19

(1999) 871–879.[60] T.Q. Duong, A.C. Silva, S.-P. Lee, S.-G. Kim, Magn. Reson. Med. 43 (2000) 383–

392.[61] W. Zhang, A. Silva, D. Williams, A. Koretsky, Magn. Reson. Med. 33 (1995) 370–

376.[62] F.Q. Ye, J.A. Frank, D.R. Weinberger, A.C. McLaughlin, Magn. Reson. Med. 44

(2000) 1236–1249.[63] M. Pedersen, M. Klarhofer, S. Christensen, J.C. Ouallet, L. Ostergaard, V.

Dousset, C. Moonen, J. Magn. Reson. Imaging 20 (2004) 930–940.[64] F.Q. Ye, K.F. Berman, T. Ellmore, G. Esposito, J.D. van Horn, Y. Yang, J. Duyn,

A.M. Smith, J.A. Frank, D.R. Weinberger, A.C. McLaughlin, Magn. Reson. Med. 44(2000) 450–456.

[65] N. Kudomi, T. Hayashi, N. Teramoto, H. Watabe, N. Kawachi, Y. Ohta, K.M. Kim,H. Iida, J. Cereb. Blood Flow Metab. 25 (2005) 1209–1224.

[66] M. Enlund, J. Andersson, P. PHartvig, J. Valtysson, L. Wiklund, ActaAnaesthesiol. Scand. 41 (1997) 1002–1010.

[67] K.M. Sicard, T.Q. Duong, NeuroImage 25 (2005) 850–858.[68] K. Sicard, Q. Shen, M.E. Brevard, R. Sullivan, C.F. Ferris, J.A. King, T.Q. Duong, J.

Cereb. Blood Flow Metab. 23 (2003) 472–481.[69] A.C. Zappe, J. Reichold, C. Burger, B. Weber, A. Buck, J. Pfeuffer, N.K. Logothetis,

Magn. Reson. Imaging 25 (2007) 775–783.[70] A.C. Zappe, J. Pfeuffer, H. Merkle, N.K. Logothetis, J.B. Goense, J. Cereb. Blood

Flow Metab. 25 (2007) 775–783.[71] W.J. Mack, R.J. Komotar, J. Mocco, A.L. Coon, D.J. Hoh, R.G. King, A.F. Ducruet,

E.R. Ransom, M. Oppermann, R. DeLaPaz, E.S. Connolly Jr., Neurol. Res. 25(2003) 846–852.

[72] A.L. D’Ambrosio, M.E. Sughrue, J. Mocco, W.J. Mack, R.G. King, S. Agarwal, E.S.Connolly Jr., Methods Enzymol. 386 (2004) 60–73.

[73] J. Huang, J. Mocco, T.F. Choudhri, A. Poisik, S.J. Popilskis, R. Emerson, R.L. DelaPaz,A.G. Khandji, D.J. Pinsky, E.S. Connolly Jr., Stroke 31 (2000) 3054–3063.

[74] A.L. Coon, F. Arias-Mendoza, G.P. Colby, J. Cruz-Lobo, J. Mocco, W.J. Mack, R.J.Komotar, T.R. Brown, E.S. Connolly Jr., AJNR Am. J. Neuroradiol. 27 (2006)1053–1058.

[75] Y. Liu, H.E. D’Arceuil, S. Westmoreland, J. He, M. Duggan, R.G. Gonzalez, J.Pryor, A.J. de Crespigny, Stroke 38 (2007) 138–145.

[76] Q. Shen, X. Meng, M. Fisher, C.H. Sotak, T.Q. Duong, J. Cereb. Blood Flow Metab.23 (2003) 1479–1488.

[77] X. Meng, M. Fisher, Q. Shen, C.H. Sotak, T.Q. Duong, Ann. Neurol. 55 (2004)207–212.

[78] S. Warach, J. Dashe, R. Edelman, J. Cereb. Blood Flow Metab. 16 (1996) 53–59.[79] C.S. Kidwell, J.R. Alger, J.L. Saver, Stroke 34 (2003) 2729–2735.[80] R.F. Rosen, A.S. Farberg, M. Gearing, J. Dooyema, P.M. Long, D.C. Anderson, J.

Davis-Turak, G. Coppola, D.H. Geschwind, J.F. Pare, T.Q. Duong, W.D. Hopkins,T.M. Preuss, L.C. Walker, J. Comp. Neurol. 509 (2008) 259–270.