Non-linear BOLD response to low-contrast Gabor elements

摘要

Many studies have demonstrated that the magnitude of the blood oxygenation level-dependent (BOLD) fMRI response in primary visual cortex increases with increasing stimulus contrast, and that the relationship between the magnitude of the BOLD response and the average firing rate in the V1 neural population is linear across a wide dynamic range. These results, however, are for BOLD responses to stimuli with a large angular subtent. We therefore tested the linearity of the relationship between the V1 BOLD response and the inferred V1 neural population response to isolated Gabor patches. Four Gabor patches (3 cycles/deg, 1 octave bandwidth), one in each visual quarterfield, were presented at 3 degrees eccentricity. One trial consisted of all four elements presented simultaneously at the same pedestal contrast in a two-interval forced choice paradigm (150 ms duration with 100 ms inter-stimulus interval); the subject's task was to detect an increment in contrast for one of the four Gabors on one of the two intervals. The Gabors were presented at 4 different contrasts (5%, 10%, 30% and 90%), randomly interleaved in an event-related design with an average (jittered) inter-trial interval of 4.5 s. Separate adaptive staircases adjusted the contrast increment to maintain performance for each pedestal contrast close 79% correct, thereby equating task difficulty across contrast. BOLD data were acquired for 5 subjects using both gradient echo (GE) and spin echo (SE) pulse sequences in the same scanning session at 7 Tesla with 2 mm (isotropic) spatial and 1.5 s temporal resolutions. The SE BOLD data showed the expected monotonic increase in amplitude with increasing contrast, but the GE BOLD response to 5% contrast Gabors was stronger than the GE BOLD response to 10% contrast elements. Because the SE BOLD technique minimizes contributions from large veins, this pattern of results suggests a vascular origin for the observed low-contrast non-linearity.