Positron emission tomography as a tool to study human vision and attention.

Abstract

The study of vision is probably the most advanced in systems neuroscience. Significant progress has been made over the last 30 years in understanding the anatomical, physiological, and computational organization of the visual system in the primate brain (1). Since the seminal work of David Marr (2), a major goal in vision research has been to characterize vision as a complex computational task and the visual system as an informationprocessing device. From this systems perspective, it is becoming more evident that the elementary processing components within the visual system are not isolated neurons, but neuronal assemblies within cortical or subcortical areas. The visual system (and brain systems in general) has been characterized as a distributed multilevel hierarchy of visual areas in which both serial and parallel processing occur simultaneously. Single neuron recordings in awakebehaving monkeys has been a powerful way to correlate neuronal activity and visual behavior. The above considerations suggest, however, that a more complete understanding of visual processing requires the analysis of neuronal activity from a wider spatial window. This analysis is now being provided by several techniques, such as 2-deoxyglucose autoradiography and optical imaging in primates and neuroimaging [PET and functional magnetic resonance imaging (fMRI)] methods in humans. The basic rationale for using PET to study human visual neurophysiology is that the performance of any task places specific information processing demands on the brain. These demands are met through changes in neural activity in various functional areas of the brain. Changes in neuronal activity produce changes in local blood flow (3, 4), and these variations in blood flow can be measured with PET. It is currently believed that changes in blood flow correspond to changes in neuronal activity at the level of one or a few adjacent cortical or subcortical functional areas. The spatial localization of the method-i.e., the accuracy of localizing a single source-is -2-5 mm (see, e.g., ref. 5), in the order of magnitude of most brain areas. The temporal resolution of a PET activation study, when using a short-lived tracer such as 150labeled water (half-life 2 min), is about 40 sec. A major drawback of the poor temporal resolution is the inability to discern the temporal relationships among multiple regions of activation. Furthermore, only processes that are repeated many times over the 40 sec can be imaged through appropriate experimental designs. It is possible that partial resolution of some of these issues will be provided by coupling PET recording with faster methods (magnetoencephalography, evoked response potentials, electrocorticography). In summary, PET allows monitoring of brain neural activity that has been averaged over time at the level of individual functional areas. PET spatio-temporal performance characteristics better suit experimental designs that look at the activation of distributed networks in the brain rather than at the fine mapping (e.g., retinotopical organization) of cortical areas in space or time.