Shape-selective stereo processing in human object-related visual areas

Abstract

Object related areas in the human ventral stream were previously shown to be activated, in a shape-selective manner, by luminance, motion, and texture cues. We report on the preferential activation of these areas by stereo cues defining shape. To assess the relationship of this activation to object recognition, we employed a perceptual stereo effect, which profoundly affects object recognition. The stimuli consisted of stereo-defined line drawings of objects that either protruded in front of a flat background ("front"), or were sunk into the background ("back"). Despite the similarity in the local feature structure of the two conditions, object recognition was superior in the "front" compared to the "back" configuration. We measured both recognition rates and fMRI signal from the human visual cortex while subjects viewed these stimuli. The results reveal shape selective activation from images of objects defined purely by stereoscopic cues in the human ventral stream. Furthermore, they show a significant correlation between recognition and fMRI signal in the object-related occipito-temporal cortex (lateral occipital complex).

Figure 1

Stereo‐experiment: stimuli and time course. (a) Examples of the types of stimuli used in the experiment: Objects defined by random dot stereograms (RDS) that appeared either to float in front of the background surface, “front” objects (FRO), or to be sunk behind the surface, “back” objects (BKO). Black and white, luminance‐defined object line drawings (luminance objects (LMO)). Gratings defined by RDS that appeared either to float in front of the background surface, “front” gratings (FRG), or to be sunk behind the surface, “back” gratings (BKG). Black and white, luminance‐defined gratings (LMG). In addition, zero‐disparity RDS were also used as controls (fRDS). Number of epoch repetitions for each stimulus type on its lower right. (b) A sample of the epochs in the course of the experiment. Short block design. Epochs of 8 sec with 6 sec of interleaving blanks.

Figure 2

Illustration of the “Stereo Front” effect. Should be viewed with red‐green or red‐magenta glasses. Green on left eye, red on right eye creates a stereo “front” object (FRO). Red on left eye, green on right eye creates stereo “back” object (BKO). Small picture at lower right represents what should be seen (LMO). Note that the “front” object (green on left eye) is perceived much easier than the “back” object (green on right eye).

Figure 3

Comparison of object‐selective activation by stereo and luminance cues. Activation foci produced by stereo and luminance cues superimposed on folded, inflated, and unfolded left hemisphere of same subject. Sulci (dark gray), gyri (light gray). LO, lateral occipital; pFs, posterior fusiform; DF, dorsal foci; yellow, “luminance” objects‐selective voxels; blue, “stereo” objects‐selective voxels; red, overlap between them. Borders of retinotopic areas are indicated by white‐dotted lines. LO activation focus also shown on conventional anatomical images (lower left). Note the substantial overlap between stereo‐ and luminance‐defined object selective regions (red) in LOC (LO and pFs). STS, superior‐temporal sulcus; ITS, inferior‐temporal sulcus; OTS, occipito‐temporal sulcus; Cos, collateral sulcus. Note that only the posterior part of the brain was scanned.

Figure 4

Object selective activation foci in four hemispheres. Activation foci shown on four unfolded hemispheres in two subjects to illustrate typical inter‐subject variability. Coloring of activation foci and notations similar to Figure 3. Note the overall similarity of the activation pattern across the different hemispheres and the substantial overlap (red) between stereo object‐selective voxels and luminance object‐selective voxels in LO and pFs. Note also that DF showed selectivity to stereo‐defined objects but without significant overlap.

Figure 5

Recognition rates and activation levels of the different conditions. (a) Mean recognition performance (% correct). (b–d) Regions defined by their preferential activation to luminance‐defined objects vs. luminance‐defined gratings and noise (LMO > LMG and fRDS). (b) Posterior fusiform (pFs). (c) Lateral occipital (LO). (d) Dorsal foci (DF). (e,f) Regions in retinotopic areas were identified by their preferential visual activation to random dots compared to blank (FRO, BKO, FRG, BKG, and fRDS > blank), and by meridian mapping done separately. (e) V4/V8. (f) V1/V2. Circle denotes an unbiased, significantly higher value for stereo objects over stereo gratings, asterisk denotes an unbiased, significantly higher, value for “front”‐objects over “back”‐objects (for P‐values see Results). Values in (b–f) represent mean activation level across subjects over all the condition's epochs. Note the trend for preferential “front” over “back” activation (asterisk), particularly in the pFs.

Figure 6

Activation levels of the different conditions in stereo‐defined, object‐selective regions. Same histograms as described in Figure 5, except the activations were obtained from regions that were highlighted by their preference to stereo‐defined objects compared to stereo‐defined gratings (FRO and BKO > FRG and BKG). Circle denotes an unbiased, significantly higher value for luminance objects over luminance gratings and noise, asterisk denotes a significantly higher value for “front”‐objects over “back”‐objects, triangle denotes the same for gratings (for P‐values see Results). Note that although the voxels were chosen by their stereo‐object selectivity, they manifest a similar selectivity for luminance‐defined objects (circle) indicating a clear convergence of object‐related stereo and luminance signals in the LOC.

Figure 7

Correlation between recognition rates and fMRI activation for the different anatomical areas. Scatter plots showing normalized fMRI activation vs. normalized recognition rate for the individual subjects and specific object categories (LMO, FRO, BKO). Anatomical object‐selective ROI (LO, pFs, DF) are identical to the ones in Figure 6, V4/V8, V1/V2 identical to ones in Figure 5. Results from each region are shown separately, ordered according to correlation value. (a) Posterior fusiform (pFs), n = 7; (b) lateral occipital (LO), n = 7; (c) V4/V8, n = 6; (d) dorsal foci (DF), n = 8; (e) V1/V2, n = 6. Each subject is denoted by a unique symbol colored in black for LMO, dark gray for FRO and light gray for BKO. Each symbol specifies normalized recognition vs. normalized fMRI activation as calculated by equation image and equation image Regression line (solid), its equation and R (correlation) value are given for each plot. Note the high correlation in the anterior part of the LOC, the pFs (a), and the low correlation in retinotopic visual areas and DF (c–e).