Functional MRI in the Awake Monkey: The Missing Link

Functional imaging in humans, first with positron emission tomography and now using functional magnetic resonance imaging (fMRI), has become a major tool of neuroscientists in the study of cerebral systems. It allows in vivo mapping of human cerebral regions engaged in an endless variety of sensory, motor, and cognitive conditions. Because these imaging techniques provide indirect measurements of the activities of large populations of neurons, their interpretation can benefit tremendously from links to the wealth of information obtained in nonhuman primates using more invasive techniques such as the properties of single neurons, anatomical connections, and behavioral effects of controlled lesions or temporary inactivations. Unfortunately, this comparison has been hampered by the confound between differences in species and in techniques. Because it is difficult to record from single neurons in humans, the most logical step has been to develop fMRI in the monkey. The first to realize the importance of this approach were Stefanacci et al. (1998), who showed that blood oxygenation level-dependent (BOLD) fMRI was feasible in the awake monkey. Activity in voxels tentatively identified as belonging to the extrastriate cortex (V2) correlated with the visual stimulus presentation: a video presentation alternating with total darkness. Without their impetus, we, like others (E. DeYoe and C. Olson, cited in Stefanacci et al., 1998) who failed to obtain fMRI signals in anesthetized monkeys, would have abandoned the effort. A heavily attended historic session at the 1998 Society for Neuroscience meeting in Los Angeles revealed that three vision laboratories, at Caltech, Max-Planck-Institut Tubingen, and Katholieke Universiteit Leuven, were following suit. In a brief report, researchers from Caltech (Dubowitz et al., 1998) confirmed that a blocked ‘‘on–off’’ paradigm (25-sec movie alternating with complete darkness) evoked correlated fMRI activity, measurable with a standard 1.5-T magnet and a knee coil, in discrete areas of the visual cortex of a single awake monkey. In addition to the low signal-to-noise ratio of the BOLD signal and an absence of retinal position control, the main problem revealed by these two preliminary reports was brain motion during scanning. The Tubingen group eliminated motion artifact by anesthetizing the monkey and administering a muscle relaxant and increased the BOLD signal by using a high-field magnet (4.7 T). This allowed them to produce ‘‘focal, reproducible stimulus-induced MR changes’’ (Logothetis, Guggenberger, Peled, & Pauls, 1999). Rotating checkerboards alternating on and off evoked activity that could be attributed to the lateral geniculate nucleus, primary visual cortex, and extrastriate areas, including V4 and MT/ V5. Comparing faces to scrambled versions of the same images evoked activity in the superior temporal sulcus and amygdala. The same anesthetized preparation was also used in two somatosensory studies using 1.5-T magnets (Disbrow, Slutsky, Roberts, & Krubitzer, 2000; Hayashi, Konishi, Hasegawa, & Miyashita, 1999). The Hayashi study revealed that face and hand representations in somatosensory cortex (SI and SII) can be distinguished at 1.5 T. The second study compared in the same animals the single-celland fMRI-defined somatotopic maps. In some cases, the match between the two types of maps was good, but mislocalization of the fMRI signal up to 1 cm from the actual single-cell activity was observed. Directional asymmetries in the mislocalizations suggested that the BOLD signal likely originated near the draining veins rather than the neuronal source. Draining veins, particularly the sagittal sinus, were also the likely source of the dominant signal in the reports of Dubowitz et al. (1998) and Stefanacci et al. (1998). One potential benefit of developing fMRI in monkeys is that it will allow one to directly compare neuronal activity and MR activity, which should provide the much needed information about the physiological basis of the functional MR signal. There are at least three unanswered questions about the functional MR (BOLD or other) signals (for the first two, see also Heeger & Rees, 2002): (1) Where is the neuronal activity giving rise to the functional MR signal localized to? (2) What type of activity (single-unit, multiple-unit, or local field potentials) underlies the MR signal? (3) How does the activity level in a population measured by fMRI relate to the selectivity and tuning curves revealed in singleunit studies? Regarding the first two questions, only one study published thus far has simultaneously measured neural activity and BOLD signal. Logothetis, Pauls, Augath, Trinath, and Oeltermann (2001) demonstrated that at high field (4.7 T), the MR activity is colocalized with neuronal activity (although it is difficult to avoid a small susceptibility artifact at the very tip of the electrode). The BOLD signal was more closely correlated Katholieke Universiteit Leuven, Belgium