Using natural images to study the relationship between the activity of neural populations and the BOLD signal

Gregor Rainer, Mark Augath, Torsten Trinath and Nikos K. Logothetis

Max-Planck-Institute for Biological Cybernetics, Tübingen (Germany)

The relationship between the blood-oxygenation-level-dependent (BOLD) signal measured using functional-magnetic-resonance-imaging (fMRI) and the activity of neural populations remains incompletely understood. The recording of BOLD activity in monkeys permits us to formulate experimentally testable hypotheses about the relationship between these signals. Here, we used parametric sets of images that were generated by interpolation between natural images and noise patterns matched for luminance and spatial frequency. We thus assembled noise-tuning characteristics of the BOLD signal for image sets with natural second-order statistics. We found that these noise-tuning curves were non-monotonic in occipital visual areas, and based on this finding we formulate some predictions about the activity of the underlying neural ensembles. Future recordings of the electrical activity in monkeys will allow us to experimentally refute or verify these hypotheses, and thus promise to advance our understanding of the relationship between BOLD levels and neural ensemble activity.