Whether we are picking out our friend's face in the middle of a crowd or driving down a dark and poorly marked road on a foggy night, we must often extract information from cluttered and ambiguous visual scenes. Often, the visual information that comes from our eyes is not enough. We must incorporate prior information we have about what color our friend's hat is or where the road is likely to turn to disambiguate noisy sensory evidence. We may use prior information to allocate cortical resources to heighten sensitivity to the location where we expect to see our friend or to the range of image contrasts that we have encountered in our night time drive. However, we understand very little of how the brain combines prior information with sensory evidence to form perceptions. Is prior information encoded in sensory areas of the brain? How does it modify noisy sensory evidence coming from the retina? The general goal of our research is to use human psychophysics and functional MRI to understand how different types of prior information is encoded and used by the brain to disambiguate noisy sensory information.

Pestilli, F., Carrasco, M., Heeger, D. J., Gardner, J. L. (2008) Does increased contrast-response in human V1 account for enhanced behavioral performance with attention? Society for Neuroscience Washington, D.C. Abstact

Human observers can deploy attention to a spatial location to improve their ability to discriminate visual information. While activity in human primary visual cortex (V1) is known to increase in a retinotopically specific manner with attention, it is unknown whether this increased activity can account for improved performance.

We measured contrast discrimination performance in 3 observers while attention was directed to the target location (an arrow at fixation cued the target) or when it was distributed over 4 locations (4 arrows cueing all locations). On each trial, flickering sinusoidal gratings (5º eccentricity, 4º diameter, 4 cycles/º, 5Hz) were presented in each visual quadrant. The contrast at each location was chosen randomly on each trial from 8 “pedestal” contrasts (range 0 to 84%). Stimuli were shown in two 600 ms intervals separated by a 200 ms blank interval. At 1 location (randomly counterbalanced across trials) the target stimulus had a near threshold contrast increase (chosen by an adaptive staircase, Quest) in 1 interval. The 3 non-target stimuli had different contrasts that remained unchanged across intervals. After stimulus offset, an arrow at fixation indicated the target location. Observers pressed 1 of 2 buttons to indicate the interval in which the target had higher contrast. Observers maintained central fixation throughout the experiment (monitored by an infrared eye tracker).

While observers performed this task, cortical activity was measured with fMRI (3T Siemens Allegra; 4 channel, phased- array surface coil; 14 slices perpendicular to calcarine; 3x3x3mm). Each observer participated in 5 sessions of the main experiment and in 1 retinotopic mapping session to localize visual cortical areas. fMRI data were pre-processed with standard procedures. Responses were computed using deconvolution, independently for each visual quadrant, pedestal contrast and attentional condition. Responses were then combined across quadrants.

Attention increased both contrast discrimination performance and V1 contrast responses. Contrast discrimination improved at all pedestal contrasts with attention. If we assume that differences in the magnitude of contrast response lead to perceptual discrimination, this predicts a contrast response curve with higher slope at all contrasts. Instead, we found that contrast response in V1 increased by a similar amount at all contrasts; its slope did not change. Our data suggest that standard models that link the shape of the contrast response curve with contrast discrimination performance are insufficient to predict the measured behavioral enhancement from the attentional modulation of human V1 activity.

Gardner, J. L., Sun, P., Waggoner, R. A., Ueno K., Tanaka, K., and Cheng K. (2005) Contrast adaptation and representation in human early visual cortex. Neuron 47:607-620 Abstact

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. pdf

The biology of the cortical vasculature system is fundamentally what enables functional MRI. Fresh oxygenated blood is delivered quite precisely to the location of recent neural activity – the signatures of these events are what functional MRI measures. Clearly the vasculature system is well organized enough to be able to respond to the metabolic demands of cortical activity. But, exactly how well organized is the vasculature around functional boundaries? For example, are cortical columns that function together served together by the same cortical vasculature? We aim to use high-resolution functional and structural MRI to understand the relationship between the organization of the cortical vasculature and the functional organization of the human cortex.

Gardner, J. L. , Sun, P., Tanaka, K., Heeger, D.J. and Cheng K. (2006) Classification analysis with high spatial resolution fMRI reveals large draining veins with orientation specific responses. Society for Neuroscience San Diego Abstact

Classifier analysis with conventional resolution fMRI (3x3x3 mm) has recently been used to decode the orientation of a grating stimulus from the fMRI responses of early visual cortex; thus demonstrating that some of the voxels in the analysis display reliable and repeatable orientation biased responses. These results suggest that conventional fMRI can probe a neural representation thought to be organized in cortical columns; presumably on the submillimeter scale. However, given that the large voxels used in the analysis would be expected to cover many orientation columns of different specificity, why such large voxels retain specificity for orientation is unknown. One possibility is that local inhomogenieties in the cortical orientation maps give rise to biased responses even in large voxels.

We tested this hypothesis by examining the spatial scale of the bias that gives rise to classifier performance. We conducted high spatial resolution imaging (0.75 x 0.75 mm inplane) and explored the consequences for classifier (linear support vector machines and Fisher linear discriminant) performance of resampling at progressively lower resolution. Classifiers could still correctly identify the orientation of a stimulus at above chance levels even with voxels resampled to an inplane resolution of approximately 1 cm x 1cm. Examination of the weights of the classifier analysis at high spatial resolution shows that the orientation specificity is due to draining veins. Long elongated areas along the cortical surface were weighted heavily in the analysis because they had consistent and repeatable orientation biases. These areas aligned precisely with large draining veins visualized in T2* weighted venograms.

We hypothesize that the orientation specificity of these draining veins is either due to sampling a large-scale orientation bias in the part of the visual field represented by the vein (e.g. a bias for cardinal or radial orientations), or that draining veins are not neutral to the underlying columnar organization; i.e. veins may drain specifically from cortical columns that are functionally related. The specificity of veins may distort maps of functional organization, but may nevertheless provide a basis for using columnarly organized responses using conventional resolution fMRI.