T1ρ contrast in functional magnetic resonance imaging

T1ρ Contrast in Functional Magnetic Resonance Imaging

Justin Hulvershorn1,2, Arijitt Borthakur2, Luke Bloy2, Eugene E. Gualtieri2, RavinderReddy2, John S. Leigh2, and Mark A. Elliott2,*1Department of Biochemistry and Biophysics, University of Pennsylvania, Philadelphia,Pennsylvania, USA.2MMRRCC, Department of Radiology, University of Pennsylvania, Philadelphia, Pennsylvania,USA.

AbstractThe application of T1 in the rotating frame (T1ρ) to functional MRI in humans was studied at 3 T.Increases in neural activity increased parenchymal T1ρ. Modeling suggested that cerebral bloodvolume mediated this increase. A pulse sequence named spin-locked echo planar imaging (SLEPI)that produces both T1ρ and T2* contrast was developed and used in a visual functional MRI (fMRI)experiment. Spin-locked contrast significantly augments the T2* blood oxygen level-dependent(BOLD) contrast in this sequence. The total functional contrast generated by the SLEPI sequence(1.31%) was 54% larger than the contrast (0.85%) obtained from a conventional gradient-echo EPIsequence using echo times of 30 ms. Analysis of image SNR revealed that the spin-locked preparationperiod of the sequence produced negligible signal loss from static dephasing effects. The SLEPIsequence appears to be an attractive alternative to conventional BOLD fMRI, particularly when longecho times are undesirable, such as when studying prefrontal cortex or ventral regions, where staticsusceptibility gradients often degrade T2*-weighted images.

Keywords3 T; spin locking; cerebral blood volume; BOLD contrast; modeling

The most frequently used sequence in functional MRI (fMRI) studies, gradient-echo (GE)echo-planar imaging (EPI), is particularly sensitive to changes in the oxygenation state ofhemoglobin, commonly referred to as blood oxygenation level-dependent (BOLD) contrast(1). BOLD contrast results from a complex interaction of physiologic changes, includingchanges in cerebral blood flow, cerebral blood volume (CBV), and the cerebral metabolic rateof oxygen consumption (2). Recently introduced functional sequences detect neural activityby directly or indirectly measuring these or other physiologic changes (3–6). Generally, thesesequences allow for quantitation of physiologic parameters or more accurate localization ofneuronal activity, but are less sensitive in detecting activation than conventional BOLD fMRI.

An alternative method for detecting neural activity using MRI is via changes in T1ρ, the spin–lattice relaxation time in the rotating frame. T1ρ is sensitive to physical processes withcorrelation times (τc) close to the reciprocal of the applied spin-locking frequency (typicallyon the order of several milliseconds). Processes with relevant τc include dipolar fields createdby slowly tumbling molecules (7), the diffusion of spins through susceptibility gradients (8),

© 2005 Wiley-Liss, Inc.*Correspondence to: Mark Elliott, University of Pennsylvania, B1 Stellar Chance Labs, 422 Curie Boulevard, Philadelphia, PA19104-6100, USA. [email protected].

NIH Public AccessAuthor ManuscriptMagn Reson Med. Author manuscript; available in PMC 2010 May 18.

Published in final edited form as:Magn Reson Med. 2005 November ; 54(5): 1155–1162. doi:10.1002/mrm.20698.

NIH

-PA Author Manuscript

NIH


NIH


and chemical exchange processes (9). The potential of spin-locking to reveal functionalchanges in the brain lies in the fact that the T1ρ of blood increases with increasing hemoglobinoxygen saturation and is longer than the T1ρ of brain tissue, which is reported to be insensitiveto changes in blood-induced susceptibility gradients (10).

T1ρ preweighting can be prepended to a GE EPI sequence with the resultant signal (assuminga 90° flip angle) taking the form

[1]

where x is the volume fraction of the ith compartment (e.g., gray matter, blood, CSF), TR isthe repetition time, TSL is the duration of the spin-lock pulse, TE is the echo time of the gradientecho, and T1ρ is the relaxation in the rotating frame for a given spin-locking frequency. In thebrain, blood and tissue compose the largest fractions of activated voxels. At 3 T, T1 is longer,and T2* is shorter, in blood compared to brain tissue. Consequently, local increases in bloodvolume fraction due to neural activation (reported to be on the order of 20–40% (11–13))generate negative contrast in conventionally weighted MRI, according to the T1 and T2* termsof Eq. [1]. This contrast subtracts from any positive BOLD contrast caused by blood saturationeffects on T2*. Like T1, T1ρ is longer in blood than it is brain tissue (~110 ms versus ~75 at 3T). However, because T1ρ decay acts like T2* in Eq. [1], increases in CBV will cause overallparenchymal (perfused tissue: tissue + blood) T1ρ to increase and thus create positive contrastto neural activation. Because the positive T1ρ (spin-lock) contrast due to increased CBV addsto positive T2* BOLD (EPI) contrast, a sequence that derives functional contrast from bothT1ρ and T2* contrast may have certain advantages over a traditional EPI sequence.

A major distinction between T1ρ and T2* contrast is a markedly different sensitivity to staticsusceptibility gradients. The lengthy echo times used in the traditional EPI sequences, requiredto produce adequate T2* BOLD contrast, result in significant signal loss in areas with largestatic susceptibility gradients due to intravoxel dephasing (14). A variety of methods have beenimplemented to reduce susceptibility-induced signal losses in EPI data, including staticgradient shimming (15), modified pulse sequences (16–18), and susceptibility matching (19).In practice, however, appreciable signal losses remain in EPI data, and fMRI sensitivity inbrain regions near tissue–air boundaries is adversely affected. In contrast, T1ρ is not verysensitive to static susceptibility gradients. Spin-locking acts on the transverse magnetizationin a manner analogous to a Carr–Purcell echo train (20) with a very short interecho interval,equal to the reciprocal of the spin-locking frequency. Therefore, static susceptibility effectsare efficiently refocused, or “spin-locked”, and T1ρ contrast can be obtained with little staticsusceptibility-induced signal loss.

This work presents fMRI results at 3 T using a T1ρ-modified EPI sequence to measure activationin the visual cortex. This new sequence produces a positive CBV-based contrast, augmentsBOLD contrast, and reduces static susceptibility-induced signal loss while maintaining highdetection sensitivity. A model is presented that estimates the CBV and blood oxygenationcontributions to the T1ρ contrast, and the model predictions are compared to the experimentalresults from the visual fMRI study.

METHODSExperiments were conducted on a Siemens Trio 3 T whole-body scanner (Siemens, Erlangen,Germany) using a birdcage head coil for RF transmission and reception. Four healthyvolunteers participated in the study, approved by the Institutional Review Board of theUniversity of Pennsylvania, after providing informed consent.

Hulvershorn et al. Page 2

Magn Reson Med. Author manuscript; available in PMC 2010 May 18.

NIH


NIH


NIH


Spin-Lock EPI Pulse SequenceIn the spin-lock EPI (SLEPI) pulse sequence (Fig. 1), a nonselective π/2 pulse excites spinsthat are then spin-locked in the transverse plane by the application of two phase-alternating(±90° phase-shifted from the phase of the first π/2 pulse) spin-lock (SL) pulses. The alternatelyphased SL pulses implement the self-compensating method of Charagundla et al. (21). Theduration of the SL pulses is denoted TSL. A second nonselective π/2 pulse restores the spin-locked magnetization to the longitudinal axis. A large amplitude dephasing gradient (indicatedby the filled square) is subsequently applied to destroy any residual transverse magnetization.After the SL portion of the SLEPI sequence, the “T1ρ-prepared” longitudinal magnetization,described by

[2]

is subsequently imaged using a GE-EPI (22) sequence with rectilinear k-space sampling.

Functional MRI AcquisitionfMRI sessions with photic stimulation were performed twice on each of four subjects. Eachsession consisted of three GE and three SL functional trials, acquired with the followingparameters: TR = 2 s, flip angle = 90°, FOV = 240 mm, voxel size = 3.75 × 3.75 × 5.00 mm,slices = 1, repetitions = 144, axial slice orientation. The SL trials had echo times (TE) of 20,30, and 50 ms, and the GE trials had TE = 30, 50, and 90 ms. The acquisition order of sequencetype and TE was randomized across subjects and sessions. The SL pulse amplitude was 500Hz, with TSL = 50 ms. The image location was prescribed from a coronal localizer to positionthe slice in the center of the visual cortex, aligned with the calcarine fissure.

Visual StimulationPhotic stimulation was accomplished using a high-contrast reversing black and whitecheckerboard alternating at 8 Hz, with a central fixation cross presented in the rest condition.The stimulus was pseudo-randomly presented 32 times within each functional trial, with eachpresentation lasting 2 s. The total presentation period lasted 288 s, and the mean intervalbetween events was 6.6 s. The stimulus timing sequence was derived using the Optseq2software package (http://surfer.nmr.mgh.harvard.edu/opt-seq/). While blocked designs withlonger stimulus durations yield larger signal changes, event-related paradigms allow for alarger number of stimulus events, resulting in increased precision in the estimation of thehemodynamic response function (hrf). Consequently, the timing paradigm was designed tomaximize the statistical efficiency of the experiment, at the expense of the power of activationdetection (23).

Determination of T1ρ in Blood and Brain ParenchymaIn order to quantify T1ρ in blood and brain parenchyma, single repetition SLEPI images wereobtained in the neck and occipital cortex of two subjects using multiple TSL times of 10, 20,30, 40, 50, 70, and 100 ms. The spin-lock frequency was 500 Hz. The other imaging parameterswere TR = 2 s, TE = 12 ms, FOV = 240 mm, voxel size = 3.75 × 3.75 × 5 mm, averages = 6.Following the SLEPI acquisitions, a 3D-MPRAGE image of the whole head was acquired withthe following parameters: TR = 1630 ms, TE = 4 ms, TI = 1100 ms, FOV = 256 mm, voxelsize = 1 × 1 × 1 mm, slices = 160. The SLEPI images were coregistered to the MPRAGE, andregions of interest (ROI) were identified for voxels within the internal carotid and vertebralarteries and the internal jugular veins. Blood T1ρ was determined by fitting the mean signalintensity within the ROIs to an exponential decay versus TSL, as described by Eq. [2]. T1ρ inbrain parenchyma was derived in a similar fashion, using an ROI containing voxels activated



NIH


NIH


NIH


http://surfer.nmr.mgh.harvard.edu/opt-seq/

by the visual stimulus. These ROIs were also used to mask tissue segmentation maps in orderto determine the fractional composition of the gray matter (GM), white matter (WM), andcerebrospinal fluid (CSF) in the parenchymal voxels. The segmentation maps were derivedfrom the MPRAGE images using the automated segmentation routine in SPM99.

Functional Data AnalysisThe image data from each functional trial were processed using standard techniques in SPM2(Wellcome Department of Cognitive Neurology, London, UK) in the following order: (1)Spatial smoothing was applied to each EPI slice using a Gaussian kernel with a full width athalf maximum of 6 mm; (2) Correlation of voxel time course data to the stimulus paradigmwas carried out using the general linear framework (24). The design matrix was constructedusing a canonical hemodynamic response function plus its first derivative (25). The regressionswere examined for statistical significance, and only voxels exceeding a threshold of z = 3.09(P < 0.001), corrected for multiple comparisons, were selected for further analysis. For eachactivated voxel, the peak activation contrast was derived from the fitted response of theregression. The contrast was converted to percentage from baseline (i.e., ΔS/S × 100) by scalingto the regression parameter used to model the DC signal component. Peak contrast was takenas the maximum amplitude of the fitted response, where the canonical basis functions (hrf plusderivative), were defined to 10 ms temporal resolution. This peak contrast represents themaximum signal change from the average response to a single stimulus event (i.e., a 2-sduration photic stimulation).

RESULTSVisual Cortex Activation

Robust activation was observed in the primary visual cortex for both sequences, in all subjectsand at all TEs. Figure 2 depicts images of the SLEPI-fMRI and EPI-fMRI obtained functionalcontrast for the activated voxels from trials involving three of the subjects. The EPI contrastmaps from each depicted session are in the top row, with the corresponding SLEPI resultsbelow. The contrast maps are overlaid on the first image from the time series. The barscaleidentifies the contrast, in percentages, assigned to each map color. In each image pair, themaximum SLEPI contrast was greater than the maximum EPI contrast, and the regions ofactivation were similar in extent, but they were not identical.

For each trial, the mean functional contrast was computed from all the significantly activatedvoxels. The average response of the mean activated contrast, categorized by sequence type andTE, is reported in Fig. 3, with error bars expressed as the SE across trials. It should be notedthat the EPI contrast reported here is lower than some reported literature values for similarvisual stimuli. As expected, both SLEPI and EPI contrast increased with longer echo time. TheSLEPI contrast was significantly (P < 0.001) larger than the EPI contrast at equivalent TEs.Specifically, the SLEPI contrast was 55% (1.31 versus 0.85) and 24% (1.66 versus 1.33) largerthan the EPI contrast at echo times of 30 and 50 ms, respectively.

T1ρ Contrast MagnitudeBOLD contrast has been shown to increase linearly with echo time (26,27). The contrastsgenerated from the SLEPI and EPI trials in this study were linearly regressed against TE, withthe results plotted in Fig. 4. For both sequence types, the linear fits were extremely good, withr = 0.996 and 0.999 for the SLEPI and EPI results, respectively. The y-intercepts of theseregressions indicate the residual contrast at TE = 0 and correspond to the predicted contrast inthe absence of T2* BOLD effects. Similar to the observations of others (27), the gradient echofunctional contrast had a near-zero intercept (0.12 ± 0.08%). The intercept of the SLEPI data,0.69 ± 0.06%, was significantly (P < 0.001) greater than zero and provides an estimate of the



NIH


NIH


NIH


T1ρ contrast in the visual cortex for the given spin-lock parameters (amplitude = 500 Hz; TSL= 50 ms) and stimulation paradigm. This additional contrast due to the spin-locked preparationperiod is additive to the BOLD contrast, which accrues during the echo time.

Static Susceptibility SensitivitySpin-locked sequences can provide superior sensitivity in regions of large static susceptibilitygradients. For example, the prefrontal cortex has large static susceptibility gradients producedby the air–tissue interfaces around the frontal sinuses. Upon analyzing ROIs in the prefrontaland occipital cortices across multiple scans, it was found that the EPI sequence had significantsignal loss in the prefrontal cortex compared to the occipital cortex. Specifically, at TE = 50ms, the SNR in the prefrontal ROI (28.8) was only 50% of the SNR in the occipital cortex(53.7). In contrast, the SLEPI sequence with TSL = 50 ms and TE = 20 ms had comparableSNR to the T2*-weighted EPI sequence with TE = 50 ms in both the occipital cortex (52.3)and the prefrontal cortex (50.2). The preserved SNR in the prefrontal cortex provides increaseddetection sensitivity in the SLEPI compared to the EPI.

Blood Oxygen Saturation Dependence of T1ρWorking with blood phantoms at 4.7 T, Kettunen et al. demonstrated that the T1ρ relaxationtime in blood is linearly related to oxygen saturation (Y) (10). In order to estimate the saturationdependence at 3 T, T1ρ was calculated from ROIs identified in the veins (internal jugular) andarteries (internal carotid and vertebral) of two subjects. The results of these calculations fromone subject are shown in Fig. 5a. The arterial ROIs are depicted in red and the venous ROIs inblue. For both ROIs, the signal dependence on TSL was well approximated by a singleexponential decay. In the first subject, the fitted T1ρ values were 115.3 and 98.7 ms for arterialand venous blood, respectively. The second subject produced similar values of 115.0 and 100.3ms. Extending on the work of Kettunen and colleagues and preserving the assertion that a linearrelationship exists between T1ρ and blood saturation in vivo, linear regressions of oxygensaturation versus T1ρ are shown in Fig. 5b. For these curves, oxygen saturation values of 70 ±5% in the jugular vein (28) and 97 ± 1% in the vertebral and internal carotid arteries (29) wereassumed. The solid line of Fig. 5b represents the best-fit line through the average T1ρ valuesfrom our two subjects, with the dashed lines showing the range of values given the uncertaintyin the oxygen saturation. The data from the Kettunen group (from Fig. 1a in (10)) are included(red dash–dot line) for comparison. Also included is an approximation of the saturationdependence of blood T2* (lower curve). The T2* relationship was computed using typicalliterature values assuming

[3]

where Y is saturation, A and C are constants, and B was set to 0 based on literature reports(30,31). As is evident from the Figure 3, T2* increases more rapidly than T1ρ given equalincreases in the oxygen saturation of hemoglobin, especially in the physiologic range (>55%saturation), and explains the exquisite sensitivity of GE-EPI to changes in blood oxygenation.

BOLD and CBV Contribution to T1ρ ContrastThe measurement of T1ρ in blood as a function of oxygen saturation allows the effects ofactivation-induced changes in blood volume and saturation to be estimated. A two-compartment model was used to describe the signal contrast expected from the T1ρ-weightedpreparation period, namely



NIH


NIH


NIH


[4]

where xblood and xtissue are the blood and tissue volume fractions, respectively. The bloodvolume fraction can be further divided into an arterial and a venous pool, each comprising 50%of the total blood volume. For simplicity, the capillary contribution is assumed to be includedwithin these two pools. The T1ρ of brain tissue (gray and white matter) was calculated fromthe measured resting parenchymal T1ρ and subtracting the blood contribution assuming an 8%blood volume fraction. The tissue segmentation analysis revealed that the voxels used todetermine the parenchymal T1ρ were predominately composed of GM, although a substantialfraction of WM contribution existed. Specifically, the average fractional composition of allactivated voxels across all subjects was 62.0, 28.9, and 7.2% for GM, WM, and CSF,respectively. The contribution of oxygen saturation changes to T1ρ signal contrast duringactivation was calculated by allowing the T1ρ to change as a function of hemoglobinoxygenation, using the relationship established in Fig. 5b, while keeping the volume fractions(xtissue and xblood) constant. The contribution to T1ρ signal contrast due solely to CBV changeswas estimated by allowing the blood and tissue volume fractions to change, while keepingsaturation (and hence T1ρ) constant. Combined effects allowed both T1ρ and the volumefractions to change from the resting to the activated state. A summary of the model parametersused in the simulation is presented in Table 1. The simulation results indicate that both theblood volume and the blood oxygenation induced contributions to T1ρ contrast were positive.The majority (93%) of the modeled T1ρ contrast was due to the increase in CBV duringactivation. The total signal change predicted using the two-compartment model (0.68%) wasin excellent agreement with the isolated SLEPI contrast (0.69%) obtained from Fig. 4.

The contrast estimate in the model was highly dependent on the resting blood volume fractionand the percentage change in the CBV. To illustrate this dependence, the modeled SLEPIcontrast (ΔST1ρ/S) for a range of physiologically relevant blood volume fractions (1–10%) andactivation-induced CBV changes (20–50%) is presented in Fig. 6. The isolated SLEPI contrastestimate obtained from the fMRI trials is included for comparison (dash–dot line), along withits error bounds (dotted lines). In order to agree with our experimental results, any combinationof resting blood volume fraction and ΔCBV that intersect with the dash–dot line can be usedin the two-compartment model. However, over the range of these values that are consistentwith the data, the relative contribution of saturation and CBV changes to the total contrast isonly moderately affected. For example, the CBV-derived component of the SLEPI contrastaccounted for 87, 91, 93, and 95% of the total modeled contrast when using ΔCBV values of20, 30, 40, and 50%, respectively. Similarly, choosing a resting blood fraction anywhere in therange of 4–8%results in an estimate of the ΔCBV-based component of the total SLEPI contrast,which is always in excess of 90%. Consequently, the two-compartment model suggests thatthe SLEPI-erived functional contrast in the visual cortex is primarily due to increases in CBVwith activation.

DISCUSSIONFunctional Activation

SLEPI fMRI produced a positive contrast in response to functional activation of the visualcortex that was additive with (GE) EPI BOLD contrast. Figure 3 suggests that the magnitudeof the isolated spin-lock contrast is about 0.7% using the current parameters (amplitude = 500Hz; TSL = 50 ms). This contrast is about 80% of the conventional BOLD contrast elicited bythe visual stimulus at TE = 30 ms, a commonly used echo time in fMRI experiments at 3 T(32). As is evident in Fig. 1, the spin-lock and gradient echo EPI are independent components



NIH


NIH


NIH


of the pulse sequence, and thus the TSL and TE can be independently adjusted to allow theratio of T1ρ and T2* contrast to vary over a wide range. T2* sensitivity provides BOLD basedcontrast that is sensitive to static susceptibility effects, while T1ρ sensitivity provides a CBV-based contrast that is relatively insensitive to static susceptibility effects.

The SLEPI sequence allows for the relative proportion of the T2* and T1ρ contrasts to be tailoredto suit the specific criteria of an experiment. For example, when investigating areas withminimal static susceptibility gradients, such as the occipital or motor cortex, a long echo timeand a long TSL can be used to maximize the combined contrast from saturation and CBVchanges. In ventral and prefrontal regions where signal loss from static susceptibility effectsare problematic for BOLD studies, a shorter echo time and long TSL will generate less totalcontrast but with reduced signal dropout. An optimized combination of TSL and TE may wellproduce a larger contrast-to-noise ratio (CNR) than possible with standard GE BOLD in suchareas. This possibility is supported by the current study, where the SLEPI data at TE = 30 msproduced equivalent contrast to the EPI sequence with TE = 50 ms (~1.3%), but the SLEPIhad significantly higher SNR (38.5) in the orbitofrontal cortex than did the EPI (28.8) at thesesame echo times. Assuming that the ratio of T1ρ to T2* functional contrast is similar withactivation of the frontal cortex, this would lead to CNRs of 0.51 and 0.37 from the SLEPI andEPI sequences, respectively. Future work will attempt to verify this assumption throughapplication of the SLEPI sequence to activation studies of the frontal cortex.

SLEPI Contrast ModelIn the current model, changes in blood volume and intravascular T1ρ effects from bloodoxygenation were used to account for the observed functional spin-locked contrast.Extravascular oxygenation-dependent T1ρ changes were not included. This is in contrast tocurrent GE EPI models, which include both intravascular and extravascular (EV) sources ofBOLD contrast. At 3 T, approximately 40% of the GE BOLD signal has been reported to arrivefrom EV space (33). This EV contrast is due to static dephasing (T2′ effects) around bloodvessels. Spin echo (SE) BOLD contrast is less sensitive to these EV changes, because spinechoes refocus much of this static dephasing (34). Because spin-locking acts on the transversemagnetization in a manner analogous to a Carr–Purcell (SE) echo train with short TE, minimalEV T1ρ contrast was anticipated, and thus this potential source of contrast was excluded fromthe model.

An intravascular component of oxygenation level dependent T1ρ contrast was included in themodel. The T1ρ of blood shows a mild dependence on oxygenation, but it is much less than theT2* dependence (see Fig. 5b). The correlation time of water diffusion across and around redblood cells has been reported to be between 0.5 and 5 ms. This is close to the 2 ms τc to whicha 50-Hz spin-lock amplitude would be sensitive, and it is proposed that these diffusionprocesses are responsible for the observed oxygenation dependence of blood T1ρ. Work byKettunen et al. (10) in the rat brain found minimal parenchymal T1ρ changes after theadministration of paramagnetic susceptibility agents. This finding is in agreement with thepredictions of the model that show that oxygenation-dependent changes in T1ρ are not largeenough to explain the observed spin-lock contrast.

Consequently, the model presented here attributes functional increases in CBV as the primarysource of the observed spin-lock contrast. The magnitude of this CBV-based contrast within avoxel is critically dependent on its blood volume fraction and the percentage increase in CBV.The current model was able to match the contrast found in the experimental data usingphysiologically plausible values (e.g., 8% blood volume; 40% CBV increase). This bloodvolume fraction is somewhat larger than the reported range of 4–6% (35,36), while the CBVincrease is within the 30–40% range that has been previously reported (11–13). However, itshould be noted that while the CBV values reported in the literature represent average blood



NIH


NIH


NIH


volume fractions across the whole brain, it is likely that the local blood volume fraction variessignificantly across tissue types and brain regions. BOLD activation is well known to bepreferentially sensitive to the intra- and extravascular venous compartments. Lee et al. foundthat venous CBV in rats comprises 70 to 80% of total CBV (37). Therefore, the local, venouscompartment CBV (as opposed to the lower, global value of CBV) may provide a moreappropriate parameter for the model.

An alternative rationale for our slightly elevated estimate of CBV is the selection of onlysignificantly activated voxels for analysis. Because the blood volume fraction generates thefunctional activation in the SLEPI sequence, highly activated voxels are expected to containeither a large resting CBV or to undergo large changes in CBV during activation. It is thereforenot surprising that the blood volume fraction in these activated voxels is higher than the averageblood volume fraction across all parenchyma. However, the possibility that a portion of ourobserved T1ρ contrast may be due to a source other than blood volume changes cannot at presentbe ruled out. In such a case the model, which does not account for additional contrast sources,would underestimate the observed contrast. For example, using a lower range of CBV values(5% blood volume and 40% ΔCBV), the model only accounts for about 65% of the observedSLEPI contrast. In order to more rigorously verify the accuracy of the model, an analysis oflocal blood volume fraction and activation-specific ΔCBV changes is required.

The combination of SLEPI and EPI contrast, possibly with higher spatial resolution thanobtained here, offers the potential to separate CBV and saturation changes during neuralactivity. The current data have a significant GE BOLD component, due to the relatively longecho times used, and prevent a more precise characterization of spin-lock contrast. Future work,wherein the echo time is minimized to decrease GE BOLD effects, will provide a betteropportunity to analyze the spatial and temporal properties of the isolated spin-lock contrastand to verify that CBV indeed mediates the T1ρ contrast, as predicted by the current two-compartment model.

Sequence DevelopmentBoth the duration and the amplitude of the spin-lock preparation can be varied to adjust thefunctional T1ρ contrast. Higher spin-lock amplitudes will further increase the magnitude of theT1ρ difference between blood and tissue, thus increasing functional spin-lock contrast. Thecurrent sequence, using a spin-lock pulse amplitude of 500 Hz, approaches specific absorptionrate (SAR) limitations at TSL longer than 65 ms. Sequence development is currently underwaythat is less SAR intensive and will allow for higher spin-lock amplitudes. The relationshipbetween spin-locking amplitude and duration, as well as the resultant functional contrast, willbe another focus of future work. Finally, the current implementation is limited to a single-sliceacquisition, due to the spatially nonselective spin-locking pulse. Recent work has demonstratedthe successful in vivo implementation of spin-locking in a multislice spin-echo imagingsequence (38). However, a serious impediment to the development of a multislice EPI sequenceis the additional increase in SAR that would be incurred.

CONCLUSIONThis work demonstrated that neural activity increases parenchymal T1ρ. Modeling suggeststhat cerebral blood volume mediates this increase. However, the two-compartment model onlyconsidered changes in oxygenation and blood volume, and it is possible that other mechanismscontribute to the observed contrast. A T1ρ modified GE-EPI sequence, SLEPI, that producesT1ρ-weighted contrast in addition to T2* contrast was introduced for use in fMRI studies.Experiments showed that SLEPI contrast is additive to traditional BOLD-based contrast andis particularly useful in brain regions with large static susceptibility gradients.



NIH


NIH


NIH


AcknowledgmentsGrant Sponsor: National Institute of Health; Grant Numbers: 5-T32-HL07614, RR02305, and MH64045.

REFERENCES1. Ogawa S, Lee TM, Kay AR, Tank DW. Brain magnetic resonance imaging with contrast dependent

on blood oxygenation. Proc Natl Acad Sci USA 1990;87:9868–9872. [PubMed: 2124706]2. van Zijl PC, Ulug AM, Eleff SM, Ulatowski JA, Traystman RJ, Oja JM, Kauppinen RA. Quantitative

assessment of blood flow, blood volume and blood oxygenation effects in functional magneticresonance imaging. Nat Med 1998;4:159–167. [PubMed: 9461188]

3. Song AW, Woldorff MG, Gangstead S, Mangun GR, McCarthy G. Enhanced spatial localization ofneuronal activation using simultaneous apparent-diffusion-coefficient and blood-oxygenationfunctional magnetic resonance imaging. Neuroimage 2002;17:742–750. [PubMed: 12377149]

4. Darquie A, Poline JB, Poupon C, Saint-Jalmes H, Le Bihan D. Transient decrease in water diffusionobserved in human occipital cortex during visual stimulation. Proc Natl Acad Sci USA 2001;98:9391–9395. [PubMed: 11459931]

5. Stroman PW, Tomanek B, Krause V, Frankenstein UN, Malisza KL. Functional magnetic resonanceimaging of the human brain based on signal enhancement by extravascular protons (SEEP fMRI).Magn Reson Med 2003;49:433–439. [PubMed: 12594745]

6. Richter W, Richter M, Warren WS, Merkle H, Andersen P, Adriany G, Ugurbil K. Functional magneticresonance imaging with intermolecular multiple-quantum coherences. Magn Reson Imaging2000;18:489–494. [PubMed: 10913709]

7. Brown RD, Koenig SH. 1/T1 rho and low-field 1/T1 of tissue water protons arise from magnetizationtransfer to macromolecular solid-state broadened lines. Magn Reson Med 1992;28:145–152. [PubMed:1331697]

8. Engelhardt RT, Johnson GA. T1 rho relaxation and its application to MR histology. Magn Reson Med1996;35:781–786. [PubMed: 8722830]

9. Duvvuri U, Goldberg AD, Kranz JK, Hoang L, Reddy R, Wehrli FW, Wand AJ, Englander SW, LeighJS. Water magnetic relaxation dispersion in biological systems: the contribution of proton exchangeand implications for the noninvasive detection of cartilage degradation. Proc Natl Acad Sci USA2001;98:12479–12484. [PubMed: 11606754]

10. Kettunen MI, Grohn OH, Silvennoinen MJ, Penttonen M, Kauppinen RA. Effects of intracellular pH,blood, and tissue oxygen tension on T1rho relaxation in rat brain. Magn Reson Med 2002;48:470–477. [PubMed: 12210911]

11. Wu G, Luo F, Li Z, Zhao X, Li SJ. Transient relationships among BOLD, CBV, and CBF changesin rat brain as detected by functional MRI. Magn Reson Med 2002;48:987–993. [PubMed: 12465108]

12. Kennan RP, Scanley BE, Innis RB, Gore JC. Physiological basis for BOLD MR signal changes dueto neuronal stimulation: separation of blood volume and magnetic susceptibility effects. Magn ResonMed 1998;40:840–846. [PubMed: 9840828]

13. Stefanovic, B.; Pike, BG. Quantitative dynamic measurement of cerebral blood volume changes viafMRI. Proceedings of the 10th Annual Meeting of ISMRM; Honolulu. 2002. p. 119

14. Ogawa S, Lee TM, Nayak AS, Glynn P. Oxygenation-sensitive contrast in magnetic resonance imageof rodent brain at high magnetic fields. Magn Reson Med 1990;14:68–78. [PubMed: 2161986]

15. Mao H, Kidambi S. Reduced susceptibility artifacts in BOLD fMRI using localized shimming.Neuroimage 2000;11:S553.

16. Constable R, Spencer D. Composite image formation in z-shimmed functional MR imaging. MagnReson Med 1999;42:110–117. [PubMed: 10398956]

17. Frahm J, Merboldt KD. Direct flash MR imaging of magnetic field inhomogeneities by gradientcompensation. Magn Reson Med 1988;6:474–480. [PubMed: 3380007]

18. Cho ZH, Ro YM. Reduction of susceptibility artifact in gradient-echo imaging. Magn Reson Med1992;23:193–200. [PubMed: 1734179]

19. Wilson JL, Jenkinson M, Jezzard P. Optimization of static field homogeneity in human brain usingdiamagnetic passive shims. Magn Reson Med 2002;48:906–914. [PubMed: 12418007]



NIH


NIH


NIH


20. Carr HY, Purcell EM. Effects of diffusion on free precession in nuclear magnetic resonanceexperiments. Phys Rev 1954;94:630–638.

21. Charagundla SR, Borthakur A, Leigh JS, Reddy R. Artifacts in T(1rho)-weighted imaging: correctionwith a self-compensating spin-locking pulse. J Magn Reson 2003;162:113–121. [PubMed:12762988]

22. Ordidge RJ, Mansfield P, Coupland RE. Rapid biomedical imaging by NMR. Br J Radiol1981;54:850–855. [PubMed: 7296215]

23. Dale AM. Optimal experimental design for event-related fMRI. Human Brain Mapp 1999;8:109–114.

24. Worsley KJ, Friston KJ. Analysis of fMRI time-series revisited—again. Neuroimage 1995;2:173–181. [PubMed: 9343600]

25. Josephs O, Turner R, Friston K. Event-related fMRI. Human Brain Mapp 1997;5:243–248.26. Menon RS, Ogawa S, Tank DW, Ugurbil K. 4 Tesla gradient recalled echo characteristics of photic

stimulation-induced signal changes in the human primary visual cortex. Magn Reson Med1993;30:380–386. [PubMed: 8412612]

27. Stroman P, Krause V, Frankenstein UN, Malisza KL, Tomanek B. Spin-echo versus gradient-echofMRI with short echo times. Magn Reson Imaging 2001;19:827–831. [PubMed: 11551723]

28. Cormio M, Valadka AB, Robertson CS. Elevated jugular venous oxygen saturation after severe headinjury. J Neurosurg 1999;90:9–15. [PubMed: 10413150]

29. Guyton, AC. Textbook of medical physiology. Philadelphia: Saunders; 1976. p. 57430. Silvennoinen MJ, Clingman CS, Golay X, Kauppinen RA, van Zijl PC. Comparison of the dependence

of blood R2 and R2* on oxygen saturation at 1.5 and 4.7 Tesla. Magn Reson Med 2003;49:47–60.[PubMed: 12509819]

31. Wright GA, Hu BS, Macovski A. 1991 I.I. Rabi Award. Estimating oxygen saturation of blood invivo with MR imaging at 1.5 T. J Magn Reson Imaging 1991;1:275–283. [PubMed: 1802140]

32. Schonwiesner M, von Cramon DY, Rubsamen R. Is it tonotopy after all? Neuroimage 2002;17:1144–1161. [PubMed: 12414256]

33. Song, AW.; Wong, EC.; Jesmanowicz, A.; Tan, SG.; Hyde, JS. Diffusion weighted fMRI at 1.5T and3T. Proceedings of the 3rd Annual Meeting of ISMRM; Nice, France. 1995. p. 457

34. Hulvershorn J, Bloy L, Gualtieri EE, Leigh JS, Elliott MA. Spatial sensitivity and temporal responseof spin echo and gradient echo bold contrast at 3 T using peak hemodynamic activation time.Neuroimage 2005;24:216–223. [PubMed: 15588613]

35. Oja JM, Gillen J, Kauppinen RA, Kraut M, van Zijl PC. Venous blood effects in spin-echo fMRI ofhuman brain. Magn Reson Med 1999;42:617–626. [PubMed: 10502748]

36. Phelps ME, Grubb RL Jr, Ter-Pogossian MM. Correlation between PaCO2 and regional cerebralblood volume by x-ray fluorescence. J Appl Physiol 1973;35:274–280. [PubMed: 4198712]

37. Lee SP, Duong TQ, Yang G, Iadecola C, Kim SG. Relative changes of cerebral arterial and venousblood volumes during increased cerebral blood flow: implications for BOLD fMRI. Magn ResonMed 2001;45:791–800. [PubMed: 11323805]

38. Wheaton AJ, Borthakur A, Charagundla SR, Reddy R. Pulse sequence for multislice T1rho-weightedMRI. Magn Reson Med 2004;51:362–369. [PubMed: 14755662]

39. Hoogenraad FG, Pouwels PJ, Hofman MB, Reichenbach JR, Sprenger M, Haacke EM. Quantitativedifferentiation between BOLD models in fMRI. Magn Reson Med 2001;45:233–246. [PubMed:11180431]



NIH


NIH


NIH


FIG. 1.The SLEPI pulse sequence. Two nonselective π/2 pulses are separated by a pair of spin-lockingpulses (SL) with opposite phase. A crusher gradient (shaded) is used to destroy any residualtransverse magnetization after the spin-locking preparation period. A conventional gradient-echo EPI sequence acquires the image data.



NIH


NIH


NIH


FIG. 2.Paired EPI (top) and SLEPI (bottom) contrast maps for three subjects. Bar-scale is in units ofpercentage contrast (ΔS/S × 100). The leftmost images were obtained with TE = 30 ms. Thecenter and rightmost images used TE = 50 ms. For each image pair, the SLEPI had higher peakand average contrast. The regions of activation were similar in extent.



NIH


NIH


NIH


FIG. 3.Average functional contrast in the visual cortex produced by the SLEPI and EPI sequences.The contrast is expressed as the percentage of signal enhancement during activation, comparedto the baseline condition. The SLEPI contrast was obtained by application of a 500-Hzamplitude spin-locking pulse for a duration of 50 ms, followed by a standard EPI sequencewith TE of 20, 30, or 50 ms. At TEs of 30 and 50 ms, the SLEPI sequence produced 55 and24% higher functional contrast than the EPI sequence, respectively. Both increases werestatistically significant (* P < 0.001).



NIH


NIH


NIH


FIG. 4.Linear regression of the functional contrast produced by the SLEPI and EPI sequences versusecho time. The y-intercepts indicate the residual contrast at TE = 0 and correspond to thepredicted contrast in the absence of T2* BOLD effects. The EPI contrast had an intercept 0.12± 0.08%, which was not significantly greater than zero (P > 0.05). The SLEPI intercept, 0.69± 0.06%, was significantly (P < 0.001) greater than zero.



NIH


NIH


NIH


FIG. 5.(a) Calculation of T1ρ in venous and arterial blood. ROIs were defined in the carotid andvertebral arteries (red) and the jugular veins (blue). Using data obtained with the SLEPIsequence at TSL = 10, 20, 30, 40, 50, 60, 70, and 100 ms, the mean signal intensity from theseROIs were fit to a single exponential function. The results from one subject are shown, withresulting T1ρ estimates of 115.3 and 98.7 ms in arterial and venous blood, respectively. (b)Modeled dependence of blood T1ρ and T2* on hemoglobin oxygen saturation. The solid linerepresents a linear regression of the measured values of T1ρ in arterial and venous blood. Thedashed lines account for the uncertainty in the assumed saturation levels of arterial and venousblood. The dash–dot line (red) reproduces the results from Kettunen et al. (10) at 4.7 T. Thebottom line (green) shows an estimate of blood T2* saturation dependence taken from Refs.(30,31) (see text).



NIH


NIH


NIH


FIG. 6.Plots investigating the dependence of the modeled contrast (y-axis) on changes in the restingblood fraction (x-axis) and ΔCBV (family of curves). The experimentally measured contrast(dash–dot line) and its error bounds (dotted lines) are included for comparison. Intersectionsof the modeled contrast with these horizontal lines represent blood fraction/ΔCBV pairs thatare congruent with the fMRI data.



NIH


NIH


NIH


NIH


NIH


NIH



TABLE 1

Parameters Used for the Two-Compartment T1ρ Simulation

TissueT1ρ

Yveinrest

Yveinactive

Yarteryrest

Yarteryactive

76 msa 0.54b 0.68b 0.97 0.97

aCalculated by subtracting 8% blood contribution from the measured resting T1ρ in brain (79 ms).

bValues from Hoogenraad et al. (39).


T1ρ contrast in functional magnetic resonance imaging

Documents