Contrast adaptation and representation in human early visual cortex

Abstract

The human visual system can distinguish variations in image contrast over a much larger range than measurements of the static relationship between contrast and response in visual cortex would suggest. This discrepancy may be explained if adaptation serves to re-center contrast response functions around the ambient contrast, yet experiments on humans have yet to report such an effect. By using event-related fMRI and a data-driven analysis approach, we found that contrast response functions in V1, V2, and V3 shift to approximately center on the adapting contrast. Furthermore, we discovered that, unlike earlier areas, human V4 (hV4) responds positively to contrast changes, whether increments or decrements, suggesting that hV4 does not faithfully represent contrast, but instead responds to salient changes. These findings suggest that the visual system discounts slow uninformative changes in contrast with adaptation, yet remains exquisitely sensitive to changes that may signal important events in the environment.

Figure 1

Visual stimulation paradigm for event-related contrast adaptation experiment (see text for details).

Figure 2

Example timecourse from a single voxel in retinotopically defined V1. A) BOLD response as a function of time from the start of experiment. Yellow arrow marks the time at which the adaptation stimulus was first presented. Green and magenta arrows indicate times at which test contrasts 1 or 2 octaves above the adaptation contrast were presented. Purple and blue arrows mark test contrast presentations 1 or 2 octaves below the adaptation contrast. 0% BOLD is set to the mean level after the 60 sec adaptation period. B) Deconvolved response to each stimulus contrast as a function of time from the beginning of the presentation of each stimulus contrast.

Figure 3

The amount of variance accounted for by stimulus time locked events (r²) is a reliable indicator of activated voxels. A) Distribution of r² values obtained for the real data (green) and when the stimulus times were randomly shuffled (blue). Inset shows the whole distribution for all voxels in the volume, and the main graph shows only the tail of the distribution. Red arrowhead marks the r² cutoff value chosen based on the randomized distribution for this experiment (see text for details). B, C and D show examples of hemodynamic responses from voxels with r² values higher than the cutoff value which are in retinotopically expected areas and show classic hemodynamic responses (same color and symbol convention as Figure 2B).

Figure 4

Contrast response functions for different adaptation levels and in different visual areas. The top row of A, B and C show contrast response functions constructed as detailed in the text for voxels in retiontopically defined V1, V2 and V3 and averaged over all subjects. The bottom three rows show distributions of parameters of fits of contrast response functions performed on a voxel by voxel basis, where the second, third and fourth rows are the distribution of parameters for c₅₀, R_max and offset, respectively (see text for details). Arrows indicate mean values of distributions. As detailed in the text, 0% BOLD is set to the mean response during the baseline period of the experiment.

Figure 5

Change in contrast response functions with different r² cutoffs. Contrast response functions for V1, V2 and V3 are shown in A, B and C respectively. Curves are constructed with voxels exceeding a cutoff r² value the same as Figure 4 (most saturated colors and thickest lines) and with cutoff values 0.11, 0.1, 0.09, 0.08, 0.05 and 0 (in descending order of color saturation and line thickness). Note that the curve with cutoff of 0 is one that is equivalent to an ROI based approach because it includes all voxels in the ROI.

Figure 6

Analysis of the rate at which adaptation affects BOLD responses. The left column in A, B and C show the response averaged over subjects for the first 60 secs of BOLD response after the adaptation stimulus is first shown, for V1, V2 and V3, respectively. Black arrow indicates the time of initial response that is used to calculate contrast sensitivity shown in the right column (open black squares). The red arrow marks the end of the adaptation period. The open red circles in the right column plot the contrast sensitivity as the average response during the experiment.

Figure 7

Responses to contrast decrements in hV4 were positive rather than negative. A) Hemodynamic responses to test contrasts for a representative voxel in hV4 (same conventions as Figure 2B). B shows the average contrast response function constructed for voxels in hV4 with the same criteria as Figure 4 (B, p<0.001). C shows curves constructed from voxels with p < 0.001, 0.005, 0.01, 0.02, 0.05, 0.10 and for the full ROI (r² >= 0.136, 0.128, 0.124, 0.120, 0.114 and 0.109, and 0 respectively) in descending order of color saturation and line thickness.

Figure 8

Positive responses to contrast decrements in hV4 examined with different stimulus lengths. Each column represents the responses found to contrast increments (top row) or contrast decrements (middle row) presented as one half period of a sinusoidal modulation. Magenta, cyan, blue, red and black traces and symbols represent the responses for 3.125, 4.167, 6.25, 8.3 and 12.5 secs of stimulus duration, respectively. Insets in the first column display stimulus types. Bottom row replots the response to the longest stimulus duration (12.5 secs) for both contrast increments (black) and decrements (red), to facilitate comparison between the two (see text for details).