Anatomical and functional MR imaging in the macaque monkey using a vertical large-bore 7 Tesla setup

Transcription

1 Magnetic Resonance Imaging 22 (2004) Research articles Anatomical and functional MR imaging in the macaque monkey using a vertical large-bore 7 Tesla setup Josef Pfeuffer a, *, Hellmut Merkle b, Michael Beyerlein a, Thomas Steudel a, Nikos K. Logothetis a a Department Physiology of Cognitive Processes, Max-Planck Institute for Biological Cybernetics, Tübingen, Germany b Laboratory of Functional and Molecular Imaging, NINDS/NIH, Bethesda, MD , USA Received 16 August 2004; accepted 8 October 2004 Abstract Functional magnetic resonance imaging (MRI) in the nonhuman primate promises to provide a much desired link between brain research in humans and the large body of systems neuroscience work in animals. We present here a novel high field, large-bore, vertical MR system (7 T/60 cm, 300 MHz), which was optimized for neuroscientific research in macaque monkeys. A strong magnetic field was applied to increase sensitivity and spatial resolution for both MRI and spectroscopy. Anatomical imaging with voxel sizes as small as Am 3 and with high contrast-to-noise ratios permitted the visualization of the characteristic lamination of some neocortical areas, e.g., Baillarger lines. Relaxation times were determined for different structures: at 7 T, T1 was 2.01/1.84/1.54 s in GM/GM-V1/ WM, T2 was 59.1/54.4 ms in GM/WM and T2* was 29 ms. At 4.7 T, T1 was 25% shorter, T2 and T2* 18% longer compared to 7T. Spatiotemporally resolved blood-oxygen-level-dependent (BOLD) signal changes yielded robust activations and deactivations (negative BOLD), with average amplitudes of 4.1% and 2.4%, respectively. Finally, the first high-resolution (500 Am in-plane) images of cerebral blood flow in the anesthetized monkey are presented. On functional activation we observed flow increases of up to 38% (59 to 81 ml/100 g/min) in the primary visual cortex, V1. Compared to BOLD maps, functional CBF maps were found to be localized entirely within the gray matter, providing unequivocal evidence for high spatial specificity. The exquisite sensitivity of the system and the increased specificity of the hemodynamic signals promise further insights into the relationship of the latter to the underlying physiological activity. D 2004 Elsevier Inc. All rights reserved. Keywords: Functional imaging; Monkey brain; High-field MR system; Cerebral blood flow 1. Introduction Understanding the distributed, synergistic activity of large neural populations requires more than single microelectrode-based measurements of neuron spiking, as the latter provide very little information on spatiotemporal cooperativity and global, associational operations in a given brain structure. Single cell recordings, large electrode or tetrode-array recordings, monitoring of action potentials and slow waves, and neuroimaging must all be employed to obtain the information required for studying the brain s capacity to generate various behaviors. Instrumental in bringing about the integration of such information has been the continuous development and improvement of anatomical, physiological and imaging techniques. * Corresponding author. Tel.: ; fax: address: josef.pfeuffer@tuebingen.mpg.de (J. Pfeuffer). Perhaps the most exciting developments have been those permitting the imaging of activity in the human brain. Currently, many different noninvasive functional neuroimaging techniques exist that can be divided into two fundamentally different approaches: (1) electromagnetic approaches, including electroencephalography (EEG) and magnetoencephalography (MEG), both providing a high temporal resolution but poor spatial information (nonuniqueness of the inverse problem); and (2) hemodynamic metabolic approaches based on the fact that neuronal activity is coupled to energy metabolism and the subsequent changes in cerebral blood flow and volume. The most commonly applied neuroimaging technique used in the second approach is functional magnetic resonance imaging (fmri), different variants of which can be employed to measure different aspects of the hemodynamic response. The BOLD contrast method is currently the mainstay of brain fmri studies because of its high X/$ see front matter D 2004 Elsevier Inc. All rights reserved. doi: /j.mri

2 1344 J. Pfeuffer et al. / Magnetic Resonance Imaging 22 (2004) sensitivity and its wide applicability [1 5], and it has already provided a wealth of information regarding the human brain, ushering in new and exciting concepts, but also raising novel and important questions about cortical function. Its chief advantage so far has been its capacity to provide us with information regarding activity patterns at good spatial resolution generated when humans are subjected to sensory stimulation or undertake particular tasks. Its disadvantage is its limited temporal resolution as well as the incapacity of the surrogate hemodynamic signal to reveal the actual neural events which generate the metabolic and hemodynamic responses [6 9]. Gradient- echo sequences, like those used routinely in human or animal fmri, are sufficiently sensitive to study both small and large vessels [10,11]. However, the significant contribution of the large vessels can lead to erroneous mapping of the activation site, as the flowing blood will generate BOLD signal changes downstream of the actual tissue with increased metabolic activity. Thus, the extent of activation will appear to be larger than it really is. The contribution of large vessels depends on both the field strength and the parameters of the pulse sequences [10,12 14], and can be de-emphasized in stronger magnetic fields, because the strength of the BOLD effect that originates in the parenchyma (extravascular) increases more rapidly for small vessels than for large ones (for review, see Ref. [15]). Thus, with a sufficiently high signal-to-noise ratio (SNR), signals originating from the capillary bed are discernable at strong magnetic fields and better reflect the site of actual neural activation [16 18]. The recent application of BOLD fmri in the nonhuman primate [19 21], and the development of the technology permitting the simultaneous acquisition of intracortical recording and fmri data in anesthetized and alert monkeys [22,23], or the in vivo study of connectivity [24], may prove useful for overcoming the aforementioned limitations. In an effort to obtain the best possible spatiotemporal resolution at the greatest spatial specificity, we set out to extend our initial experimental setup (4.7 T/40 cm MR system) by the installation of a second large bore MR scanner (7 T/60 cm). The higher magnetic field was chosen in order to increase the SNR, as well as the spatial resolution and specificity for functional imaging and to improve the spectral dispersion and quality [25] for MR spectroscopy and chemical shift imaging (CSI) using 1 H, 13 C, 17 O and other nuclei. The large magnet bore of 60-cm diameter (33 cm after gradient coil and acoustic shielding) was necessary to be able to conduct experiments with awake monkeys, trained to participate in a number of psychophysical tests. An upright (vertical) magnet, affording natural positioning of the monkeys and minimum discomfort, was designed to expedite the training process and ensure longer cooperation during the demanding psychophysical experiments. Here we report the first MR structural and functional imaging results in anesthetized monkeys at 7 T. Emphasis is given to the description of the radiofrequency (rf) system, the sensitivity of which plays a critical role in image quality, namely, in establishing the best possible signal- and contrast-to-noise ratios [26]. All rf coil designs were optimized for applications at 300 MHz. The anatomical, BOLD and functional-cbf images acquired with the system at 7 T have exquisite spatial specificity promising deeper insights into the relationship of these signals to the underlying neural activity. Preliminary results have been published in abstract form [27 31]. 2. Methods 2.1. Dedicated monkey 7-T MR system A novel vertical 7 T/60 cm MR system (Bruker BioSpin, Ettlingen, Germany) was set up, in which multimodal MRI, MR spectroscopy and simultaneous electrophysiology can be performed in the anaesthetized and the awake, behaving monkey (Fig. 1). The system is housed in a tower and extends over three floors with an overall height of 6.40 m. The room under the bore opening is quality shielded for low frequencies and has floating ground to ensure noise-free electrophysiological recording of both local field and action potentials inside and outside the magnet bore. The technical details of the 7-T (300 MHz) magnet are as follows: 60 cm bore, distance to isocenter 180 cm, four magnet stands 2.35 m (floor to base of steel shield), 0.5 mt magnetic stray field at 8.5 m axial and 5.3 m radial, empty weight 12 tons for the magnet and 67 tons for the iron shielding. The gradient insert has an opening of 38 cm (BGA-38) and is driven by a Siemens gradient power amplifier with peak voltage of 700 V and current of 500 A. A maximum of 80 mt/m gradient strength per channel can be achieved in less than 200 As. Shim coils are integrated in the gradient coil. Inside the gradient, a passive sound insulation with 33-cm inner diameter provides a 30- db noise reduction to db for demanding single-shot EPI sequences. The MR system is controlled by a Bruker BioSpec console under Linux on the ParaVision 3 platform. Three rf transmission channels with 4- kw, 400-W and 2.5-kW amplifiers allow experiments on all common nuclei like 1 H, 13 C, 15 N, 17 O and 19 F as well as dual 1 H channel transmission. Four receiving channels allow for parallel imaging and simultaneous multinuclear spectroscopy Subjects, transport systems and visual stimulation For all experiments, anaesthetized rhesus monkeys (Macaca mulatta) with body weights of 4 5 kg were used. All procedures and experiments were approved by the local authorities (Regierungspr7sidium) and were in full compliance with the guidelines of the European Community (EUVD 86/609/EEC) for the care and use of laboratory animals. A prototype chair for primates was custom- designed and built to accommodate the positioning of the electrophysiology apparatus, rewards for the awake animal, stimulus presentation and control of unwanted movement. The chair

3 J. Pfeuffer et al. / Magnetic Resonance Imaging 22 (2004) Fig. 1. Newly installed 7 T/60 cm vertical Bruker Biospec MR system. In this dedicated monkey system, simultaneous electrophysiology and MR imaging and spectroscopy can be performed in the anaesthetized as well as the awake, behaving animal. (A, B) The 7-T magnet is housed in a tower and extends over three floors with an overall height of 6.40 m. (C) Chair for experiments with anaesthetized animals, which is set in the horizontal position for preparation and then moved into the vertical position with the help of the support swing for the experiment. (D) Chair for experiments with awake monkeys below the magnet bore and setup for fmri and electrophysiology. is driven into the magnet with a vertical transport system using 2-m spindle drives and magnetically screened motors with programmed controlled gears and an emergency remove feature. All cables connecting the proximal end of the recording and monitoring devices to the main equipment were fed through a filter panel at the lower end of the chair. Surgical procedures, anesthesia and details of the presentation of visual stimuli were described recently [23]. A full-field visual stimulus was presented with a custombuilt 8 Hz flickering LED array or with rotating checkerboard using a fiber-optic system (Avotec, Silent Vision, FL) Radiofrequency coils The 1 H Larmor frequency in a 7.05-T magnetic field is 300 MHz, which is in the ultra-high frequency (UHF) domain of the rf spectrum. At this frequency one has to be aware that

4 1346 J. Pfeuffer et al. / Magnetic Resonance Imaging 22 (2004) the wavelength l in vacuum is 1 meter. Because the relative dielectric constant (e r ) of pure water is 81, the wavelength is shortened by a factor of 9 resulting in l of 11 cm. The shortening in fatty tissue is less; however, it is still significant. As a consequence, dielectric modes can be present in 1 H MRI applications for structures in the order of a fraction of 11 cm in size, which is appropriate for the human as well as in the monkey brain. The dielectric mode also modulates the B 1 distribution [32 35]. The majority of MR techniques for imaging or spectroscopy utilize the nuclear spin of the 1 H atom, which is abundant in organic matter, e.g., water, fat, proteins, peptides and smaller molecules including neurotransmitters. Within the present communication, for reasons of clarity, we limit ourselves to describing the rf coils at the Larmor frequency of 1 H. Radiofrequency coils are the most crucial front-end elements within the rf chain of an MR system and considered a key locus for overall sensitivity [36]. The large variety of scan types used in our research (e.g., volume imaging, imaging with implanted coils, spectroscopy, perfusion imaging) require diverse designs of rf coils. What follows describes typical examples of such coils and coil assemblies in different operation modes Transceive mode This uses the same coil for rf transmission and reception (transceive) utilizing the system s transmit receive (TR)

5 J. Pfeuffer et al. / Magnetic Resonance Imaging 22 (2004) switch. The setup is simple and best performance is typically achieved with coils having a homogeneous field (B 1 ). Local coils, e.g., surface coils, generally offer better sensitivity. However, their B 1 field generated by such coils is inhomogeneous and its use requires additional precautions in pulse sequence design and execution TORO mode This approach (commonly used within clinical scanners) utilizes a combination setup with separate transmission-only and receive-only (TORO) coils or arrays [37]. In that mode both transmit and receive coils need to be isolated from each other so that rf B 1 coupling is avoided, which would otherwise result in nonlinear excitation and reception with compromised SNRs. There are several methods to overcome coupling: (a) geometric placement of the coils so that their B 1 is orthogonal. (b) Partial overlap of the coils so that their B 1 interaction is nullified. (c) Electronic decoupling (detuning) with additional hardware such as pin diode switches. The particular coil is then activated while the other coil is detuned and vice versa at complementary segments of the application pulse sequence. (d) Utilization of low-noise lowimpedance preamplifiers in case of receive arrays. We used a combination of several methods in order to achieve sufficient decoupling with our homebuilt coils. In addition, we utilized also a high-performance coil (a 30-mm electronic surface coils for reception) supplied by the system vendor. The following designs were adapted to work with the standard active decoupling hardware of the 7-T system Volume coils (transceive) High sensitivity and homogeneity are important for anatomical imaging. In addition, minimization of space requirement is necessary to allow the positioning of the visual stimulation equipment and other apparatus used for electrophysiology or the maintenance of anesthesia (in the anesthetized-animal experiments). Coils of different sizes were constructed, each used to fit a group (size range) of animals. A typical example of a saddle coil is given in Fig. 2A. The transceiver coil consists of a pair of barrelshaped loops of silver tubing (5 mm outer diameter, 4 mm inner diameter) or from 1.5-mm pure silver sheets that were cut into the required shape. High-voltage ceramic chip capacitors type 100E (American Technical Ceramics, Huntington Station, NY) were added in-line/in-loop for increased performance and to fulfill resonance conditions at 300 MHz. The loops were driven in parallel by the tune match network made from variable capacitors type NMNT20HV (Voltronics, Denville, NJ). The equipment was mounted onto a cylindrical structure of a 1-mm-thick fiberglass epoxy (G10) tube that was open at the anterior side to allow space for the extended mount/nose of the animal and for stimulus presentation. Several sized coils with diameters ranging from 105 to 135 mm inner diameter (id) were constructed Surface coils (transceive) Surface coils are made from a single loop of wire mounted on a G10 substrate and connected to a tune match network. A selection of surface coils and a generic schematic are shown in Fig. 2B. Smaller loops (18 25 mm) were made Fig. 2. Radiofrequency coils for 1 H experiments at 7 T (300 MHz) for transceive mode (A C) and transmit-only/receive-only mode (E F). (A) 11-centimeterdiameter saddle coil (transceiver) for whole brain studies using silver stripes. (B) Diverse surface coils with 30-, 40- and 60-cm loops (transceiver) for studies with a localized FOV. (C) Two-loop (6/6 cm) quadrature coil. (D) 12.5-cm-diameter actively decoupled saddle coil (transmit) for whole excitation. (E) Receive-only surface coil with 30-cm loop for studies with a localized FOV. (F) Pluggable circuit for an implanted 20-mm coil with active decoupling. Circuit descriptions: Saddle coil (transceive) (A) for relatively homogeneous coverage of the whole primate brain. The inductor loops A and B with in-loop capacitors (Ci) are each respectively connected in parallel to the tuning capacitor (Ct) at points (a) and (b), thereby forming the resonating structure of the coil which is connected to match (Cm) and balance capacitors (Cb), respectively, to the transmission line. Surface coil (transceive) (B) used for studies of local areas of interest. The transmission line center wire is fed to a variable capacitor (Cm) in order to match the impedance of the coil circuit to the characteristic impedance of the transmission line of 50 V (double-shielded coaxial cable). The Cb facilitates the return pathway for the rf of the resonator which is composed of a variable capacitor (Ct) and an inductor (loop of wire with Ci). In-loop capacitors are not necessarily needed for smaller coils (e.g., 18 to 25 mm diameter). One to three Cis are typically used for surface coils with a diameter between 22 and 80 mm diameter. Quadrature surface coil (transceive) (C) used for studies of local areas of interest. The transmission line center wire is fed to a variable capacitor (Cm) in order to match the impedance of the coil circuit to the characteristic impedance of the transmission line. The Cb facilitates the return pathway for the rf of the resonator which is composed of a variable capacitor (Ct) and an inductor (loop of wire with Ci). One Ci is used for each loop of 65-mm diameter. bdouble-dq-shaped coil (transmit-only) (D) for relatively homogeneous coverage of more than two thirds of the primate brain. The inductor loops A and B with Ci are connected in parallel to the Ct at points (A) and (B), respectively, thereby forming the resonating structure of the coil which is connected to Cm and Cb to the transmission line. The in-line pin diode (Dp) in loop A is a switch which is activated by a DC current during the rf transmission period. DC is applied via two rf chokes (rfc) at (b) and (c) that prevent rf leakage. A filled box is used for the symbol of the rfc or for an inductor. The resistor (R) limits the current that is applied to Dp at constant voltage (resistor symbol is an open box). Surface coil (receive-only) (E) used for studies of local areas of the brain. The transmission line center wire is fed to a variable capacitor (Cm) in order to match the impedance of the coil circuit to the characteristic impedance of the transmission line. The Cb facilitates the return pathway for the rf of the resonator which is composed of a variable capacitor (Ct) and an inductor (loop of wire with Ci). Detuning is achieved by shorting the tuning capacitor with the aid of the Dp using a DC current supplied at points (a) and (b) via rfcs. At least one Ci is necessary in order to avoid loop DC bias current. During the signal reception period, a negative voltage is applied at control port (D). Implanted surface coil (receive-only) (F) used for the study of a local area of the brain utilizing an implanted isolated loop of silver wire. The transmission line center wire is fed to a variable capacitor (Cm) in order to match the impedance of the coil circuit to the characteristic impedance of the transmission line. The Cb facilitates the return pathway for the rf of the resonator which is composed of a variable capacitor (Ct), two Cis and an inductor (permanently implanted loop of silver wire). The impedance tune- and match network is connected to the loop using slotted copper beryllium tubes (St). Detuning during rf transmission is achieved by shorting the tuning capacitor with the aid of the pin diode (Dp) using a DC current supplied at points (a) and (b) via rfc.

6 1348 J. Pfeuffer et al. / Magnetic Resonance Imaging 22 (2004) Table 1 Longitudinal relaxation times T1 (s) (standard deviation in brackets) in the macaque monkey brain at 4.7 and 7 T measured by saturation recovery MSME Study a Gray matter (not V1) Gray matter (V1) White matter Thalamus Basal ganglia 7 T/300 MHz mm 3 P02.i (0.20) 1.73 (0.18) 1.53 (0.24) 1.63 (0.20) 1.78 (0.16) I02.iO (0.17) 1.91 (0.18) 1.61 (0.13) 1.66 (0.12) 1.83 (0.17) 7 T/300 MHz mm 3 H02.lf (0.29) 1.88 (0.18) 1.49 (0.19) 4.7 T/200 MHz mm 3 B97.ve (0.22) 1.49 (0.17) 1.08 (0.14) 1.13 (0.10) 1.38 (0.15) a Code before decimal point reflects different animals like dh02t, dp02t; code after decimal point reflects different study dates. from 2- or 3-mm-thick silver wire. Larger coils (25 80 mm) were made from 5-mm silver tube and one to three ceramic capacitors (4 7 pf) built-in to increase performance and to meet resonance conditions at 300 MHz. Utilization of surface coils is straightforward and has an operational benefit and increased SNR because of its simplicity. However, the steep B 1 profile also requires great attention in the transmit pulse selection [38] Implanted coils (transceive) Transceive implanted surface coils take advantage of closer proximity to the area of interest resulting in increased sensitivity as compared to surface coils. Implanted coils for transceive consist of a single loop of silver wire heavily insulated with Teflon tubing that is permanently placed within the monkey s skull and terminated within a turret as described previously [23]. In addition, a tune match network is connected via a pair of slotted copper beryllium tubes, which is identical to the surface coil (transceiver) described above Quadrature surface coil pair (transceive) Quadrature coil assemblies typically exhibit increased sensitivity at deeper lying structures and require lower transmit power as compared to single coils. In addition, a B 1 profile which is less steep overall is achieved. Radiofrequency coupling of the two surface coil elements is neutralized by overlap [39,40] as shown in Fig. 2C. A drawback of this coil configuration is that each coil has to be tuned and matched individually. In addition, a quadrature hybrid (type , MHz, Anaren) tuned to the frequency of interest and an identical set of cables between coil and hybrid is required, which makes the setup somewhat more complicated Partial volume coil (transceive- or transmit- only) This coil is to cover a relatively large area of the brain with a lateral B 1 excitation field, exhibiting sufficient field homogeneity in fmri, MRI or MRS studies with the tradeoff of a moderate B 1 gradient. Its lateral field makes it an ideal partner for use in combination with surface coil reception (TORO) mode, because the transmission and reception coil B 1 fields are orthogonal. In addition, it can be used solely as a transceiver coil. The coil consists of a pair of D-shaped loops (Fig. 2D, A and B) of silver tubing (5 mm outer diameter, 4 mm inner diameter). In-line capacitors of type 100E were added for increased performance and to fulfill the resonance conditions at 300 MHz. The loops were driven in parallel by the tune match network made from variable capacitors of type NMNT23HV. A type UM 9415 or UM 9401 pin diode (Microsemi, Watertown, MA, USA) was inserted into loop A. When the resonator is in operation mode, a DC current is applied to the pin diode via two rf chokes. In the case of TORO mode operation, the pin diode is negatively biased at all times except during rf transmission, which then detunes the two-loop resonator. For our applications, the coils were mounted on a cylindrical structure of a 1-mm-thick G10 tube. The structure was chosen to be half of a cylinder and wide open (p) at the anterior side allowing rotation along the z-axis without interfering with the face of the animal and in order to cover desired areas of interest for better sensitivity, for example, within the lateral visual cortex of one hemisphere. Table 2 Transverse relaxation times T2 (ms) (standard deviation in brackets) in the macaque monkey brain at 4.7 and 7 T measured by MSME Study Gray matter (not V1) Gray matter (V1) White matter Thalamus Basal ganglia 7 T/300 MHz mm 3 O02.jy (6.5) 64.8 (7.1) 61.1 (6.0) 59.6 (3.2) 62.0 (5.3) D01.gG (10.9) 49.7 (7.6) a 55.5 (11.5) 51.7 (6.2) 56.2 (8.4) P02.i (5.3) 54.0 (6.1) 52.5 (6.8) 49.2 (3.4) 52.4 (4.9) I02.iO (4.5) 51.4 (4.6) 48.3 (4.4) 48.5 (2.6) 49.8 (3.6) 7 T/300 MHz mm 3 H02.lf (4.7) 48.0 (2.7) 43.1 (5.7) 4.7 T/200 MHz mm 3 C01.d (4.7) 63.1 (3.1) 63.2 (4.5) 63.7 (3.7) 65.7 (6.5) B02.d (6.0) 76.6 (8.6) 65.0 (6.9) 61.5 (4.9) 61.0 (7.1) B01.d (5.6) 65.1 (4.4) 63.9 (5.1) 64.3 (4.0) 65.1 (5.1) G03.kq (6.4) 64.3 (4.0) 63.1 (5.3) 62.7 (4.2) 65.7 (4.6) a T2 values decreased due to V1 chamber.

7 J. Pfeuffer et al. / Magnetic Resonance Imaging 22 (2004) Surface coils (receive- only) Receive-only surface coils have the advantage that reception of the NMR signal is independent from the excitation B 1 and can achieve higher SNR in the sensitive area. For proper operation, an orthogonal placement to the excitation coil is often not sufficient for decoupling and therefore additional electronic detuning is a requirement. As shown in the schematic (Fig. 2E), typically a pin diode is included in the circuit that detunes the resonance frequency during transmission of the excitation pulse Implanted coils (receive- only) This mode of operation takes advantage of an implanted loop of silver wire, which was identical to the implanted transceiver coil case (Fig. 2F). In addition, the tune match terminates with slotted copper beryllium tubes that are connected to the implant. It is particularly important that the implanted coil does not short-circuit the DC used to control the pin diode. Therefore, DC blocking capacitors were used in addition to the simple tune/match circuit Data acquisition A multislice spin-echo sequence (MSME) and a gradient- echo sequence (FLASH) was used for anatomical and susceptibility-weighted imaging. Specific acquisition parameters are given in the adjacent figure captions. To optimize our pulse-sequence in subsequent experiments, we first conducted a series of experiments aiming to accurately measure relaxation times for monkeys at 7 T. Quantitative measurements of T1, T2 and T2* in brain tissue were obtained from a series of spin-echo and gradientecho images. For T1, saturation recovery MSME series with varying repetition time TR was measured keeping the echo time TE minimal and constant. Typical parameters were field of view cm 2, matrix , five axial slices with 2 mm thickness, SW 50 khz, TE 13 ms, TR = s (logarithmic time steps, 16 images). For T2, MSME series with multiple refocused spin echoes (CPMG) were measured with a long repetition time between excitations. Similar parameters were used, but TR=5 s and TE= ms (linear time steps, 24 echoes). For T2*, FLASH series with different echo times were measured keeping TR constant. Similar parameters were used, but SW 62 khz, TR=450 ms and TE= ms (logarithmic time steps, 16 images). The studies on the 4.7- and 7-T systems used a different field-of-view (FOV), which resulted in different voxel size (see Tables 1 3). A segmented, blipped gradient-echo echo-planar imaging sequence (EPI) was used for functional imaging of BOLD (IMND BLIP_ EPI, Bruker). The EPI readout was a gradient waveform with alternating trapezoidal lobes. Linear sampling on the ramps was accounted for by regridding before image reconstruction. A navigator FID with the duration of two k-space lines was sampled before the EPI readout, which was used for segmented EPI images to monitor phase and frequency changes and correct the dynamic off-resonance in k-space (DORK) [41] and image segmentation artifacts. In the anaesthetized monkey setup, global frequency changes due to breathing were 1 2 Hz peak-to-peak (S.D Hz), corresponding to a 3% pixel shift in the EPI phase-encoding direction, if uncorrected. The global amplitude stability (center k-space point) in vivo was approximately % S.D./mean. Therefore, no amplitude correction was performed. For imaging of cerebral blood flow (CBF) and its changes upon functional activation, the most recent EPI sequence (PVM EPI, Bruker) provided by the vendor was modified and supplemented by preparation modules for arterial spin labeling and inversion-prepared EPI. Singleshot, blipped gradient-echo EPI was acquired with an inplane resolution of 500 Am and an asymmetric FOV of cm 2 (matrix 12848, sweep width 200 khz). A zoomed fast imaging approach was utilized here to achieve a high spatial resolution of 500 Am without segmentation, in a similar fashion as described previously [41,42]. The reduction of the FOV and the matrix size in the phasecoding direction was possible by using a small receive coil with 30 mm diameter and a single coronal slab of outervolume suppression placed anterior to the intended pseudoaxial image. The single-shot EPI had an acquisition time of 31 ms, and thereby negligible T2* blurring effects. A FAIR module was implemented for arterial spin labeling using alternating adiabatic slice-selective and non- slice- selective inversion with a 10- ms hyperbolic Table 3 Empirical transverse relaxation times T2* (ms) (standard deviation in brackets) in the macaque monkey brain at 4.7 and 7 T measured by gradient echo Study Whole image Gray matter (not V1) Gray matter (V1) White matter Thalamus Basal ganglia 7 T/300 MHz mm 3 P02.i (12.2) 29.1 (10.9) 20.5 (8.2) 28.3 (8.4) 35.7 (8.1) 36.4 (8.7) I02.iO (10.4) 35.8 (12.6) 36.0 (9.3) 29.1 (7.6) 40.9 (7.5) 7 T/300 MHz mm 3 H02.lf (11.5) 29.9 (12.9) 30.3 (10.9) 24.4 (10.1) 4.7 T/200 MHz mm 3 C01.d (16.1) 30.7 (16.0) 19.4 (12.5) 30.0 (13.1) 48.4 (12.0) 49.7 (14.0) B02.d (12.5) 32.0 (12.5) 23.9 (10.5) 30.0 (10.1) 45.4 (12.1) 33.2 (11.8) B96.ve (13.5) F03.kp (17.0)

8 1350 J. Pfeuffer et al. / Magnetic Resonance Imaging 22 (2004) Fig. 3. Anatomical spin-echo images of the monkey brain at 7 T obtained with an 11-cm diameter transceiver saddle coil. Four pseudo-axial slices are shown cutting through the cerebellum (a) and the ventricles (d). The image contrast is mainly determined by T2 and proton density: cerebrospinal fluid appears brightest, white matter dark and gray matter intermediate. Acquisition parameters (D01.gG1): multislice spin echo sequence, FOV cm 2, matrix , seven slices with 2 mm thickness, nominal resolution mm 3, 12 echoes (echoes 1 3 averaged), NA 2, SW 50 khz, TE ms, TR 4 s. secant pulse (R =20). Different inversion times (TI) were tested from 700 to 1500 ms. For the reported experiments, a TI of 1100 ms was used with an inversion slice thickness of 8 mm (image slice thickness was 3 mm), which provided the best compromise between CBF contrast and specificity for capillary contributions. A minimal possible echo time of 12 ms was chosen to optimize SNR and to minimize at the same time the residual BOLD contributions in the CBF data. Semiquantitative CBF was calculated from the difference of the slice- selective and non-slice-selective image intensities S [43], CBF=(S SS S NS )k{m 0 TI[2exp( TI/T1) exp( TR/T 1 )]} 1, with the tissue-to-blood partition coefficient k =0.9 ml/g [44], TI=1.1 s, TR=3 s and T1 assumed to be 1.9 s as determined for gray matter in monkey primary visual cortex V1 in a separate relaxation measurement. A fully relaxed M 0 image was measured with TR=10 s (five times T1) Estimation of relaxation times and analysis of functional data The data analysis was performed with self-written routines under MATLAB 6.5 (The Mathworks). The calculation of relaxation time maps was performed in three steps: voxelwise fitting to relaxivity curves, image segmentation to delineate different brain areas and calculation of mean and distribution widths of the relaxation times. Before processing, data were screened for consistency within a study and potential acquisition artifacts like voxel shifts, internal timing and scaling correctness, spikes, etc. In total, 24 series from 11 animals were analyzed Fit procedure The series of T1, T2 and T2* relaxation data were fitted to the following equations: (a) T1 series with varying t =TR: S =S0[1 a exp( t/t1)]. (b) T2/T2* series with varying t =TE: S =S0 exp( t/t2)+b. Nonlinear least-squares fitting of three parameters S0, T1/T2/T2* and a/b was done for each voxel with the Gauss Newton method (MATLAB function nlinfit). For each fitted parameter, the 95% confidence intervals were calculated (MATLAB functions nlparci, nlpredci) and used as an error estimate of the fitted relaxation times T1/T2/T2* and the proton density S0 (initial signal at t =0). The fit procedure resulted in parameter maps of T1, T2, T2*, S0 and error maps r T1, r T2, r T2*, r S0 corresponding to these parameters Segmentation procedure An inhomogeneity correction was applied to selected images to flatten out global spatial intensity variations before the tissue segmentation. For this purpose, a high-pass

9 J. Pfeuffer et al. / Magnetic Resonance Imaging 22 (2004) spatial frequency filter and subsequent Gaussian windowing were applied to eliminate high spatial frequencies. The resulting intensity profile of spatial B1 distributions was then used for correction. (1) Manual region of interest (ROI) selection was used to mask out skull and outer-brain areas and to determine areas such as the primary visual cortex, V1, the basal ganglia and portions of the thalamus. (2) A CSF mask was derived from T2-weighted MSME images with echo times greater than 200 ms (approximately three to four times the T2 of gray/ white matter) using inhomogeneity correction (as described above) and a manual threshold. (3) The S0 map from the T1 or T2 fit was considered as the reference and was segmented based on intensity thresholds to separate gray and white matter regions. First, an inhomogeneity correction was applied. Next, a manual upper threshold was chosen to target gray matter areas only. Finally, an upper and a lower threshold were implied to focus on white matter areas. This segmentation procedure was performed separately for each of the five slices to provide the masks for the different ROIs (CSF, white matter, gray matter, basal ganglia and thalamus). In addition, the gray matter mask was divided into V1 and other extrastriate cortical areas Calculation of mean and distribution width An iterative Gaussian fit was used to determine the mean and standard deviation of a distribution with outliers. For this purpose, a histogram of a distribution was fitted to a Gaussian to first estimate the mean and standard deviation. Then the tails of the distribution were discarded using a threshold of 3 S.D. A repeated fit proved to be robust and converged to the btrueq mean and Gaussian width of the core distribution not containing the outliers, which were observed as a result of the nonlinear fit of noisy voxels. Thresholding of the relaxation data was done twofold and for each of the T1, T2 and T2* maps separately: (1) based on the error maps r i for each of the parameters resulting from the nonlinear fit (single-voxel error thresholding), and (2) based on the distribution width of the relaxation parameters from a selected ROI ensemble (ensemble thresholding). A specific ROI mask was first applied to the error maps r i (single-voxel error). Then mean/width of the r i distribution (m i,sd i ) was estimated using an iterative Gaussian fit, which was used to discard voxels with single-voxel errors larger than m i +2 sd i. From the remaining voxels, the ensemble mean and Gaussian distribution width of corresponding relaxation times were determined, again using an iterative Gaussian fit. In this way, further outliers were discarded that were not found based on the bsinglevoxel error thresholding Q. Finally, the processing of the relaxation data resulted in region-specific T1, T2 and T2* values, and standard deviations, respectively Processing of functional data The analysis of functional imaging series was done with STIMULATE [45] and MATLAB routines. Statistical maps were generated from the fmri times series comparing intensities during the task vs. the resting condition (block design) with a Student s t test for populations with unequal variances (unpaired, two-tailed). Activated voxels were considered with t N2 (P b.05). Time courses were linearly detrended for each voxel. 3. Results and discussion 3.1. Anatomical imaging using spin- echo and susceptibility-weighted contrast Anatomical imaging of the whole primate brain was performed with the transceive volume coil (saddle coil, 11- cm diameter; Fig. 2A). Four selected spin-echo images are shown in Fig. 3, acquired with a multislice, multiecho sequence. In the pseudo-axial slices, the cortex, white matter structures, ventricles and cerebellum are outlined clearly. The image contrast is mainly determined by T2 and proton density, whereby cerebrospinal fluid appears brightest; white matter, dark; and gray matter, intermediate. The nominal voxel resolution was mm 3, which was interpolated in-plane to mm 2 for visualization purposes. The saddle coil had a relatively tight fit around the head of the animal with a frontal opening providing access to the visual stimulator. This design minimized the coil volume and provided good SNR together with a relatively homogeneous excitation over the whole brain. Some signal decreases due to smaller B1 are visible in most frontal brain areas, which seemed, however, to be a good compromise between design effort and coil performance. A combination coil setup (TORO mode) was used to address the demands for highest SNR in localized brain areas, which could then be translated into high spatial and/or high temporal resolution. Examples of high-resolution susceptibility-weighted images of the occipital cortex of one hemisphere are shown in Fig. 4. The 12.5-cm diameter, partial volume coil with active decoupling was used for rf transmission (saddle coil; Fig. 2D), which covered about two thirds of the posterior primate brain for rf excitation. An implanted 20-mm surface coil (Fig. 2F) was used for the highest receive sensitivity. The approximate position of the coil wires is indicated by the two circles in Fig. 4D. Scout images indicate the position of the axial and coronal high-resolution FLASH images with an acquisition resolution of up to Am 3. The slices were chosen to be parallel and perpendicular to the cortical surface. With the susceptibility weighting, which was achieved by use of a gradient echo with a relatively long echo time of 20 ms and a large flip angle, the signal from blood vessel areas was decreased and appeared dark. In a three-dimensional view of the whole coronal data set (Fig. 4C), the pial vessels could be clearly seen plunging from the surface into the cortex and branching to different areas. The small dark dots in Fig. 4C are intracortical blood

10 1352 J. Pfeuffer et al. / Magnetic Resonance Imaging 22 (2004) Fig. 4. Susceptibility-weighted gradient- echo images of the primary visual cortex using an implanted 20-mm receiver coil. The scout images [sagittal (A) and axial (B)] show the position of the pseudo-axial and pseudo-coronal slices perpendicular to the cortex surface. With susceptibility weighting using a relatively long echo time, the signal from blood vessels is decreased and appears dark. In the pseudo-coronal slice close to the cortical surface (C), large pial vessels on top of the cortex are depicted as well as the small intracortical vessels that run perpendicular to the cortex (small dots in (C)). They are similarly seen in the axial image as lines (D) that run exactly perpendicular to the cortical surface. The approximate position of the wires of the implanted coil is indicated by the two white circles. Acquisition parameters (I02.l61): multislice FLASH sequence: (A) pseudo-axial ( 308), pos 10.9 mm F (slice 13), FOV cm 2, matrix , 51 slices with 300 Am thickness, nominal resolution Am 3, NA 8 (1 h 10 min). (B) Pseudo- coronal (+308), pos 15.1 P (slice 21), FOV cm 2, matrix , 51 slices with 300 Am thickness, nominal resolution Am 3, NA 2 (15 min). SW 25 khz, a = 45, TE 20 ms, TR 1.8 s. vessels that run perpendicular to the cortical surface as seen in the axial slice in Fig. 4D Relaxation time parameter maps of T1, T2 and T2* Relaxation times T1, T2 and T2* in the monkey brain were determined for two magnetic field strengths, 4.7 and 7 T, from a series of spin-echo and gradient-echo experiments. The fit to mono-exponential relaxation decay curves provided multislice parameter maps of T1, T2 and T2*, and proton density S0 with corresponding error maps (see Methods section for details). Image segmentation was performed and provided region-specific values for the five areas of gray matter (separated into V1 and other cortex), white matter, thalamus and basal ganglia. The mean of the T1 values at 7 T (Table 1) was largest in gray matter (not including V1) with 2.01 s and smallest in white matter with 1.54 s. T1 in gray matter of the primary visual cortex V1 was smaller (1.84 s) than the other cortex (2.01 s), possibly due to higher myelination in layer IVb. We found an intermediate T1 in the subcortical structures, which was 1.65 s in the thalamus and 1.81 s in the basal ganglia. The ratio between different regions was very similar at 4.7 T, but the T1 values were shorter and only 75% compared to the T1 values at 7 T (Table 1). The mean of the T2 values at 7 T (Table 2) was largest in the gray matter (not including V1) with 59.1 ms and smallest in the thalamus with 52.3 ms. We found intermediate T2 of 55.0 ms in V1, 54.4 ms in the white matter and 55.1 ms in the basal ganglia. White matter T2 was about 10% decreased compared to gray matter T2, which was observed at both fields, 7 and 4.7 T. The ratio between different regions was again similar at 4.7 T, but T2 values were on average 18% longer at 4.7 T compared to 7 T. The T2* values (Table 3) were significantly more inhomogeneous compared to T1 and T2, most likely because of the greater effects that local susceptibility changes and varying field inhomogeneities have on the T2* values. The whole brain T2* values were about 34 ms at 4.7 T and were shorter at 7 T with 29 ms. However, there was a wide standard deviation of 10 ms at both fields. Inner brain regions such as the thalamus and basal ganglia showed significantly longer T2* values. Gray matter regions (not

11 J. Pfeuffer et al. / Magnetic Resonance Imaging 22 (2004) including V1) and white matter were very similar. A significant difference was found between V1 and the other gray matter cortex, which could be explained by the wellknown higher density of vascularization characterizing the primary visual cortex of the primate [46,47]. Notably, the large pial vessels produced strong local susceptibility artifacts, an observation that may account for the reduction of the average T2* value in V1 by 20 ms compared to the other cortical regions. The errors of the reported T1 values at 7 T were on average about 10%; the errors of the T2 values were about 11% at 7 T and 8% at 4.7 T. When comparing the voxelbased distribution of the error maps, which resulted from the single voxel fits of the decay curves, we found that the standard deviation of the ensemble means was similar and about 10% for T1. This indicated that our errors were directly and predominately determined by the accuracy of the MR measurement, which was directly related to the voxel resolution and SNR, and to the number of time points in the series measured for the saturation recovery decay. Similar conclusions hold for the T2 and T2* series. In this regard, it was important to measure T1 and T2* with logarithmic spaced steps to improve the fit error of the decay parameters. For the T2 multiecho decay this was not possible due to MR technical constraints and therefore measured in linear steps. The relaxation times we report here are similar to those reported for the human brain: at 7 T, T2 was 55/46 ms in GM/WM, and T2* was 25 ms in GM; at 4 T, T2 was 67/58 ms in GM/WM, and T2* was 41 ms in GM [16,48]. T1 was reported to be 1.91 s in GM and 1.43 s in WM at 7 T; T1 was 1.35 s in GM and 0.93 s in WM at 4 T [49,50] Stria of Gennari outlined with T2 contrast An important goal in neuroscientific research is to determine the functional specificity of cyto- or myeloarchitectonically distinct cortical regions. Development of methods that do not require a fixed brain, and thus the killing of the subject, is of obvious significance. With this aim, recent studies demonstrated the feasibility of in vivo structural mapping of the human cerebral cortex utilizing T1 contrast [51,52]. The spatial resolution afforded by the present study made it possible to also visualize structures that are commonly observed only in situ in histology microphotographs obtained following intravascular injections. For example, the CNR of the images was sufficient to identify characteristic myeloarchitectonic cortical features, such as the axonal plexuses, commonly referred to as Fig. 5. High-resolution spin- echo images and corresponding T1 and T2 maps of the monkey brain at 7 T. (A) T2-weighted image of the primary visual cortex shows bright CSF and dark white matter regions (average of echoes 1 3 with TE=13, 26 and 39 ms). Within gray matter V1 the stria of Gennari is visible, which has different contrast due to higher myelination of the horizontally cross-linked fiber connections. (B F) Fitted relaxation parameter maps from a saturation recovery series for T1 (B) and a multiecho series for T2 (C), proton density (D), R1=1/T1 (E), and R2=1/T2 (F). Obvious contrast is visible between gray and white matter in the proton density and R1 maps (D, E). In contrast, the stripe of Gennari is mostly distinguished in the R2 maps, where the structure (layer IVb) within the gray matter is visible (F). Acquisition parameters (H02.lf1): multislice spin echo sequence, FOV cm 2, matrix , five slices with 1 mm thickness, nominal resolution Am 3, NA 2, SW 50 khz, TE 13 ms, TR 6 s. T2 series with 20 echoes (TE= ms), T1 series with 16 different repetition times (TR=0.3 6 s in log steps).

12 1354 J. Pfeuffer et al. / Magnetic Resonance Imaging 22 (2004)

13 J. Pfeuffer et al. / Magnetic Resonance Imaging 22 (2004) Fig. 7. High-resolution maps of cerebral blood flow (CBF) using single-shot gradient-echo FAIR-EPI in the anaesthetized monkey with a 30-mm receive coil. (A C) Anatomical inversion-recovery EPI images with different inversion times showing (A) gray matter, (B) white matter, and (C) mixed contrast (image are smoothed for visualization). (D) Susceptibility-weighted gradient-echo image of the primary visual cortex (intensity corrected for the B1 decay of the surface coil). (E) Highest cerebral blood flow spots overlaid as contours on an IR-EPI image. (F) Quantitative CBF map of the occipital cortex with mm 3 resolution delineating high CBF in gray matter and blood vessels vs. white matter with little CBF. Acquisition parameters (O02.nV1): (A C, F) single-shot GE EPI (PVM) with IR/FAIR module (adiabatic slice-selective/non-slice-selective inversion) and outer-volume suppression, pseudo-axial, FOV cm 2, matrix 12848, one slice with 3 mm thickness and 8 mm inversion slice thickness, nominal resolution mm 3, SW 200 khz, TIR=700 ms (A), 900 ms (B), 1100 ms (C, F), TE 12 ms, TR 2*3 s, segment acquisition time 31 ms. (D) Multislice FLASH, FOV cm 2, matrix , 41 slices with 1 mm thickness, nominal resolution mm 3, NA 4 (34 min), SW 25 khz, a =45, TE 15 ms, TR 2 s. the Baillarger lines, on the basis of differences in the relaxation time. Specifically, we have acquired a full set of T1 and T2 series to measure quantitative T1, T2 and proton density values in subcortical structures of primate V1. An intermediate in-plane resolution of Am was used for relaxation time series covering the whole brain (see previous section). We performed one study with considerably higher resolution of Am 3 using a TORO setup with a 30-mm receive coil. The results are also reported in Tables 1 3 (study H02.lf1). The T2-weighted image of the primary visual cortex of the right hemisphere shows T2 contrast between bright CSF, intermediate gray matter and darker white matter regions (Fig. 5A). The fitted T1, T2 and S0 parameter maps are shown in Fig. 5B F, for demonstration purposes together with the inversed R1 and R2 maps (R1=1/T1, R2=1/T2). In Fig. 5F, the gray/white matter border is marked by a line for easier visualization. Different myelin concentrations along the depth of the cortex could be observed due to the high contrast-to-noise ratio in the T2 images and the good spatial resolution, which minimizes the partial volume effects of different tissue types within the same voxel. We have manually segmented the V1 cortex into high and low myelin sections, whereby the former was found to coincide with the outer Baillarger stripe. The Gennari line, as this stripe is called in the visual cortex, is known from histology to include myelinated axons of stellates and pyramidal cells running in parallel to the pial surface. The T2 of the Gennari line was shorter with 46.0 ms (2.2 S.D.) compared to the other gray matter in V1 with 47.8 ms (2.6 S.D.). The T1 values were only slightly different with 1.86 s (0.15 S.D.) for the Gennari stripe compared to the T1 in the other V1 gray matter with 1.92 s (0.27 S.D.). In conclusion, we have Fig. 6. Functional BOLD activation map from a single time series of an eight- segmented gradient- echo EPI in the anaesthetized monkey using an 80-mm transceive surface coil. Full-field visual stimulation with a rotating checkerboard was used in a block design with 48-s stimulus on and off (four periods). (A) Anatomical images of the same slices as used for fmri, which were obtained with an inversion-recovery RARE sequence using an adiabatic inversion pulse. Excellent T1 contrast is visible even in anterior areas where the B1 excitation profile of the surface coil is already strongly diminished. (B, C) t-value map (t N2) and time course of positive (red yellow) and negative (blue) activation. The relative BOLD percent changes were 4.1% (0.4) and 2.4% (0.5) for all activated voxels with P b.05. Acquisition parameters (I02.jn1): (A) multislice IR-RARE sequence for T1 anatomy, pseudo-axial, FOV cm 2, matrix , 13 slices with 2 mm thickness, nominal resolution mm 3, NA 4, SW 55 khz, TE 10 ms, TR 5 s, RARE factor 8. (B) Multislice GE-EPI fmri series, pseudo-axial, FOV cm 2, matrix , 13 slices with 2 mm thickness, nominal resolution mm 3, NA 1, NR 64, SW 150 khz, TE 20 ms, TR 750 ms, eight segments, segment acquisition time 13.7 ms.

14 1356 J. Pfeuffer et al. / Magnetic Resonance Imaging 22 (2004) shown that based on quantitative T1 and T2 mapping one can achieve sufficiently high T2- contrast to observe differences in myelination within the gray matter Functional imaging based on BOLD contrast BOLD fmri was conducted in the anaesthetized monkey using an 80-mm transceiver surface coil. Stimulation of the central 228 of the visual field was done by means of a rotating polar-transformed checkerboard pattern in a block design with 48-s stimulus on/off in four periods. Anatomical T1-weighted spin-echo images in pseudo- axial orientation of six out of a total of 13 slices are shown in Fig. 6A. All images had excellent T1 contrast and reasonable homogeneity in the T1 contrast despite the use of a surface rf coil for transceiving. This is mainly due to the use of an adiabatic inversion pulse, providing homogeneous and B1-independent inversion. In Fig. 6B, the stimulus-induced activation is shown as a t value map ( t N2, positive activation in red to yellow and negative activation in blue to purple, t =[2...8]) superimposed on the T1-weighted scans. The time series of activated voxels (Fig. 6C) show the magnitude of signal changes during the presentation of the stimulus. Relative signal increases were 4.1% (0.4% S.D.) for all activated voxels with t N2, total 5697 voxels (bpositive activationq, red squares). Correspondingly, relative signal decreases were 2.4% (0.5% S.D.), total 2310 voxels (b negative activationq, blue circles). Activation was observed along the entire representation of the stimulated visual field in primary and extrastriate visual cortices, e.g., V1, V2, V3 and V5 (MT). Activation was confined entirely within the gray matter with sharp edges at the gray white matter border (see for example the most posterior area V1) demonstrating the overall spatial selectivity of the signal. Nevertheless, highly significant signal modulations were often observed in locations, which on the basis of the anatomical images, correspond to vessels. The high magnetic field undoubtedly emphasizes extravascular BOLD, originating in the capillary bed, by further suppressing the intravascular contributions. Yet the latter are not entirely eliminated at 7 T field strength, and they may still play a significant role in the definition of the spatial response function of the neurovascular system examined with susceptibilitysensitive pulse sequences. Anticorrelated responses, often referred to as negative BOLD, were consistently observed in areas adjacent to the visual cortex that were not directly activated by the visual stimulus, as reported in recent studies [53 56]. The number of anticorrelated voxels was less than half of the positively activated ones. The average negative percent change was about 60% of the average positive percent change. A recent study [53] suggested that negative BOLD is consistently associated with decreases in CBF that were larger in amplitude than the accompanying decreases in oxygen consumption. Correlation of negative BOLD with decreases in cerebral blood volume has been shown in the cat [54]. Based on experiments with simultaneously measured fmri and electrical recordings, it has been shown that the negative activation in monkey V1 is associated with decreases in neuronal activity [57]. These decreases in neuronal activity were comparably large in both the local-field potential and the multiunit activity. The decreases in neuronal activity in V1 are likely to underlie at least part of the CBF decreases observed as anticorrelated hemodynamic responses. In the anesthetized monkey, a second component of CBF Fig. 8. Functional activation map deduced from changes in CBF (fcbf) using single- shot gradient-echo FAIR-EPI in the anaesthetized monkey with a 30- mm receive coil. Upon visual stimulation, CBF increased on average by 38% from 60.7 (3.0) ml/100 g/min at rest to 80.9 (5.6) ml/100 g/min during activation. CBF decreases during activation were 21% to 46.2 (2.8) ml/100 g/min. (A) Anatomical inversion-recovery EPI image showing gray matter. (B, C) t-value map ([2...8], red...yellow) and time course of positive (red yellow) and negative (blue purple) activation. Robust and consistent functional CBF changes are observed, which were excellently localized to gray matter only. Acquisition parameters (O02.nV1): single- shot GE EPI as described, NR 64, full-field visual stimulation with an 8-Hz flickering LED array, 48-s stimulus on/off in four periods.