Functional magnetic resonance imaging based on …

1 Functional magnetic resonance imaging based onChanges in Vascular Space OccupancyHanzhang Lu,1 3*Xavier Golay,1,3 James J. Pekar,1,3and Peter van Zijl1,3*During brain activation, local control of oxygen delivery is facil-itated through microvascular dilatation and constriction. A newfunctional MRI (fMRI) methodology is reported that is sensitiveto these microvascular adjustments. This contrast is accom-plished by eliminating the blood signal in a manner that isindependent of blood oxygenation and flow. As a consequence,changes in cerebral blood volume (CBV) can be assessedthrough changes in the remaining extravascular water signal( , that of parenchymal tissue) without need for exogenouscontrast agents or any other invasive procedures.

2 The feasibil-ity of this vascular space occupancy (VASO)-dependent func-tional MRI (fMRI) approach is demonstrated for visual stimula-tion, breath-hold (hypercapnia), and hyperventilation (hypocap-nia). During visual stimulation and breath-hold, the VASO signalshows an inverse correlation with the stimulus paradigm, con-sistent with local vasodilatation. This effect is reversed duringhyperventilation. Comparison of the hemodynamic responsesof VASO-fMRI, cerebral blood flow (CBF)- based fMRI, andblood oxygenation level-dependent (BOLD) fMRI indicates botharteriolar and venular temporal characteristics in VASO.

3 The effectof changes in water exchange rate and partial volume contami-nation with CSF were calculated to be negligible. At the common-ly-used fMRI resolution of 5mm3, the contrast-to-noise-ratio (CNR) of VASO-fMRI was comparable to that of CBF- based fMRI, but a factor of 3 lower than for supporting a better gray matter localization for theVASO-fMRI approach compared to BOLD are provided. MagnReson Med 50:263 274, 2003. 2003 Wiley-Liss, words: cerebral blood volume; fMRI; visual stimulation;microvascular; water relocation; exchange; hyperventilation;breath-hold; hemodynamic response; BOLD; cerebral bloodflow; VASOD uring the last decade, Functional MRI (fMRI) has revolu-tionized the field of cognitive neuroscience.

4 This is a con-sequence of its capability to provide noninvasive spatialmapping of the hemodynamic response to neuronal activ-ity (1 3). To date, virtually all fMRI studies in humansubjects have been based on the measurement of localchanges in the oxygenation state of hemoglobin, whichaffects the MRI signal by changing the local magnetic fieldin the image volume elements (voxels). This so-calledblood oxygenation level-dependent (BOLD) phenomenon(1,4) reflects the combined effect of many physiologicalparameters, including oxygen extraction ratio, cerebralmetabolic rate of oxygen, cerebral blood flow (CBF) andvolume (CBV), hematocrit, and initial arterial oxygenationfraction (4 6).

5 One drawback of the BOLD approach is thatthe altered oxygenation state of hemoglobin not only in-fluences MRI signals in and around the microvessels closeto the sites of neuronal activation, but also in and aroundthe large veins draining from these areas (7 10). This prob-lem is especially apparent at low magnetic field strengths( Tesla), at which most fMRI studies are conducted, andpersists at intermediate (3 4 Tesla) and higher fields (9).Although the availability of improved spatial resolution athigher magnetic field strengths allows the actual areas ofactivation to be better localized, it would be extremelyuseful to have a method in which the contrast would arisepredominantly from the microvessels that are expandingdue to local neuronal brain vasculature is under dual mechanistic control(11).

6 The diameters of larger vessels of the cerebral inflowtract are under sympathetic regulation, and are consideredto be unimportant for local regulation of flow during acti-vation. On the other hand, the diameters of small intrapa-renchymal vessels (100 200 m) are influenced by thedemand for homeostasis of the microevironment. Suchlocal neurovascular coupling is mediated by vasoactivecompounds, such as CO2, NO, prostaglandin, and K (11 13). An fMRI approach that reflects activation-related mi-crovascular regulation is expected to show effects in localparenchyma (tissue microvasculature), but not in largevessels.

7 To achieve contrast that is dependent on vascularspace occupancy (VASO), we designed a technique inwhich the blood signal is selectively nulled. This wasaccomplished by taking advantage of the fact that MRIradiofrequency (RF) pulses can invert the longitudinalequilibrium magnetization of water from parallel with re-spect to the magnetic field to antiparallel, after which theinverted magnetization returns exponentially to equilib-rium with the longitudinal relaxation time constantT1(Fig. 1). BecauseT1differs between blood and tissue, thetimes at which their magnetizations cross zero differ, and,when images are acquired at the time of blood nulling,sufficient gray matter tissue magnetization remains forMRI detection (Fig.)

8 1).Here we demonstrate the feasibility of this VASO-fMRIapproach during vasodilatation (visual stimulation andbreath-hold) and vasoconstriction (hyperventilation) ma-nipulations in humans at T. In addition, an event-related visual stimulation experiment was performed tocompare the hemodynamic and contrast-to-noise-ratio(CNR) characteristics of VASO with that of conventionalBOLD-fMRI and a CBF- based fMRI approach using pulsed1 Department of Radiology, Johns Hopkins University School of Medicine,Baltimore, of Biomedical Engineering, Johns Hopkins University School ofMedicine, Baltimore, Kirby Research Center for Functional Brain imaging , Kennedy KriegerInstitute, Baltimore, sponsor: NIH; Grant number: NS37664 (NINDS); Grant sponsor: Na-tional Center for Research Resources.

9 Grant number: RR15241.*Correspondence to: Peter van Zijl or Hanzhang Lu, Dept. of Radiology,Johns Hopkins University School of Medicine, 217 Traylor Bldg., 720 RutlandAve., Baltimore, MD 21205. E-mail: 5 December 2002; revised 10 March 2003; accepted 30 online in Wiley InterScience ( ). magnetic resonance in Medicine 50:263 274 (2003) 2003 Wiley-Liss, spin labeling (PASL-fMRI). Finally, fMRI data wereacquired at high resolution to compare the spatial speci-ficities of the VASO and BOLD of the MethodFigure 1 shows the basic MRI pulse sequence for VASO-fMRI. After nonselective inversion of the starting longitu-dinal magnetization,M(0), from parallel to the magneticfield to antiparallel, the parenchymal components relaxback to equilibrium.

10 BecauseT1differs between blood andtissue, the point at which the magnetization crossesthrough zero differs. Longitudinal magnetization cannotbe detected, but when, at a timet TI, such an inversionexperiment is followed by excitation of transverse magne-tization (which is detectable), the effects of the inversionpreparation determine the starting signal magnitude ( ). This signal decays exponentially with the transverserelaxation time, which is described byT*2for gradient-echo imaging andT2for spin-echo imaging . When a gra-dient-echo pulse sequence is used for image acquisitionwith echo time TE, the signal dependence is:S M TI e TE/T*2 M 0 1 2e TI/T1 e TR/T1 e TE/T*2[1]in which TR is the time needed for a single MRI acquisi-tion (Fig.)