Information About Color and Orientation in the Primate Visual Cortex - High-Throughput Image Reconstruction and Analysis - page 233

stimulus highly unreliable in terms of determining what stimulus was presented to
the investigated brain. This is true for both imaging and electrophysiology. Below
we show that, with our imaging data, the application of the SVM classifier in a
focused ''searchlight'' mode makes it possible to identify the stimulus that elicited
a single response with high accuracy.
10.2 Methods and Results
A detailed description of experimental methods and the SVM algorithm can be
found elsewhere [21, 26]. In each experiment, we first identified the V1/V2 border
by imaging the ocular dominance (OD) columns [Figure 10.1(b)]. OD columns are
present only in V1, so their termination marks the border between these two visual
areas of the cortex. The OD columns were visualized by subtracting the responses
elicited by stimulating the right eye from those driven by stimulation of the left
eye. We then identified the CO thin stripes by imaging the V2 areas that preferred
color gratings to black/white ones [regions indicated by arrows in Figure 10.1(c)]
[20, 28].
To determine how the spatially distributed activity across the cortical sur-
face represents information about stimulus color and orientation, 16 different
Figure 10.2 Mean responses to four different gratings. Each mean response was calculated by
averaging the 32 single-trial responses to the given grating.