Anna Martelli Ravenscroft
Left vs Right processing of Face & Place in fovea & periphery
File:Left vs Right processing.pdf
Anna Martelli Ravenscroft
Background: Vision depends on multiple regions of the brain, from the specialized photoreceptors of the retina, through the lateral geniculate nucleus of the thalamus, back to V1 in the occipital and anterior again into specialized object recognition areas such as the fusiform face area in the mid-temporal lobe. [1] Vision begins in the retina, with the photoreceptors for light (rods) and color (cones) concentrated in the back of the eye in a small area called the fovea, where the light will be focused by the lens of the eye. The fovea is near the blindspot where the optic nerves leave the eye. As you move further from the fovea, the concentration cones drop off quickly, with rods taking up the work of peripheral vision. [1]
Vision processing occurs throughout the visual stream, with lower-level processing such as luminance, color, edge, and orientation occuring early (LGN and V1) and motion and form processing occuring in later regions (V2-V4 and MT). [1] Higher-order processing is considered to be processed in two "streams" – the dorsal stream handling motion and spatial information, while the ventral stream handles object recognition.[2] Several specific areas have been identified, including the FFA: fusiform face area and the PPA: parahippocampal place area.[3] The lateral occipital (LOC) also participates in object recognition, particularly face recognition in the occipital/fusiform area. [4]
Sensory representation is often topographic in the brain, with vision organized retinotopically – LGN and V1 have specific regions to process foveal input, while peripheral input is eccentric to the foveal processing. [1] Some of the higher-order processing areas are more activated by foveal input while others prefer peripheral input.[5] Many processes of the brain are lateralized, including speech and some visual processing.[1] This study focuses on the effect of lateralization on image processiong – to determine whether there were significant differences in activation between faces, places and object on the left vs right side, in addition to the foveal/peripheral presentation distinction.
Methods
Subjects: The subject is AL, a 22 year old female.
MR Acquisition: I used scan data acquired by Kalanit Grill-Spector's lab and provided as pre-processed par files. The experiment involved blocks of visual stimuli: images presented foveally or peripherally. Two localizer runs were performed. TR= 2s, 198 TRs, for a total of 396 seconds. The stimulus blocks were 12 seconds long, with 2 of each stimulus type (man face, child face, indoor scene, outdoor scene, abstract object, cars), and 16 total blank blocks between stimuli. In runs 3 and 4, alternating blocks of faces and places were presented either foveally or in the periphery. Foveally presented images were small, due to the smaller size of the fovea compared to the periphery. Again, there were TRs =2 seconds, 198 TRs, for a total of 396 seconds. Stimulus blocks were 12 seconds long, 4 of each stimulus (Face fovea [FF], face periphery [FP], place fovea [PF] and place periphery [PP]. There were 16 total blank blocks between stimulus presentations.Runs 6-9 were motion-related runs which I am not analysing.
MR Analysis: As noted, the scan data was provided in a pre-processed state as parfiles. Scans 1 and 2 were grouped, as were scans 3 and 4. I ran GLMs on the localizers runs and the trial runs. Then I ran the contrasts, using the first two rnus to localize the area, using T-values to run the contrast. Next, I created 6 ROIs: Left & Right side of FFA, PPA, and OFA. I ran timecourses and betas for each ROI on the trial runs scan group. Finally, I ran multi-voxel cross-correlation matrices for each ROI.
Results:
The following are the timecourses and betas for the left and right ROIs. (See PDF linked above for the images.)
The left OFA shows a preference for foveally presented images, and for faces more than places.
The right OFA shows a distinct preference for images presented peripherally and deactivation for foveal presentation. The pattern is almost direct opposite of the left OFA. However betas for both left and right OFA were very low (18.8% and 14.8% respectively) which leaves any "result" somewhat dubious.
The left PPA shows preference for places, with a strong preference for peripheral presentation.
The right PPA shows a very similar pattern with a stronger preference for places, and for peripheral presentation. Betas are stronger in the PPA ROIs (than OFAs), at 47.8% and 59.5% variance accounted for.
The left FFA shows high activation for foveally presented faces with lower activation for peripheral faces and all places.
The right FFA showed higher beta values for Faces on periphery (FP) and places on Periphery than for foveally presented faces (an odd finding for a "face" area). There is preference for faces but even more preference for images shown on the periphery!
Note that the beta values for the right FFA are the highest of the 6 ROIs at 68.4% while the left FFA shows 36.4% variance accounted for.
As noted above, I ran multi-voxel cross-correlations for several of the ROIs.
Left OFA shows some dissociation between foveal and peripheral presentation of faces and places.
Right OFA shows some dissociation within category between foveal and peripheral presentation of images. Across categories, there is a relatively strong dissociation between foveal and peripheral presentation of images. This matches well with the betas in the Right OFA.
Left PPA shows slight dissociation between foveal and peripheral presentation within category, and little differentiation between category.
Right PPA shows good dissociation within category between foveal and peripheral. Across categories, the dissociations aren't as strong, with some correlation by presentation location, especially for peripherally presented images.
Left FFA shows a strong dissociation between face foveal and face peripheral; but place does not show this distinction as well – showing positive correlation for both fovea and periphery, with less strong correlation for fovea. Across categories, there was low correlation in everything except peripherally presented images.
Right FFA is indistinguishable from Right PPA!
Conclusions:
There are clear differences between these left and right visual processing structures. A good picture of these differences can sometimes be supplemented by doing MVPA analysis, although it is less helpful in other cases. Some specific differences follow.
• Left OFA and right OFA are opposites in correlation for fovea vs periphery presentation, with some preference for faces over places in both. The MVPAs are relatively similar, although the left is more clear in distinguishing category and location effects.
• *Left PPA and right PPA beta values show very similar patterns with good betas, although the right showed stronger effects. In the MVPAs, there was a more muddy picture with little differentiation within or across categories or locations for left PPA but a fairly good distinction within and across categories for right PPA.
• Left FFA strongly distinguishes foveal and peripheral presentation of faces, but not places. There was a fair amount of correlation for peripheral faces and places, which is clearly shown in the beta values plot: FF was very high, with FP, PF, and PP nearly identical betas. Right FFA, on the other hand had a much less clear picture - preferring peripheral presentation of both faces AND places over foveal faces, with a 68.4% variance! The MVPA shows a difference pretty clearly between the two with much more clear dissociation by location across categories than in the left FFA.
• Both MFPA and beta correlations indicate that the right FFA and the right PPA have a preference for peripheral image presentation with a smaller preference for face vs place respectively. Interestly, their MFPAs are indistinguishable!
I was not expecting such clear lateralization in these structures. The results lead me to conjecture that lateralization gets stronger (shown by more heterogenous processing preferences) as we moved anterior into higher-order visual processing, at least in the ventral stream. It would be interesting to do a study to specifically examine whether laterality increases in the dorsal stream as well as the ventral stream of vision processing.
References
1. Principles of Neuroscience, 4th edition, Kandel et al.
2. What Has fMRI Taught Us About Object Recognition? Kalanit Grill-Spector
3. Brain Activation during Face Perception: Evidence of a Developmental Change E. H. Aylward1, J. E. Park1, K. M. Field1, A. C. Parsons1, T. L. Richards1, S. C. Cramer2, and A. N. Meltzoff
4. The fusiform face area subserves face perception, not generic within-category recognition, K Grill-Spector, N Knouf, N Kanwisher - NATURE NEUROSCIENCE, 2004
5 J Neurophysiol. 2008 Jul;100(1):249-67. Epub 2008 May 7. Relating retinotopic and object-selective responses in human lateral occipital cortex. Sayres R, Grill-Spector K.
Appendix: Thank you to Kevin Weiner for spending time walking me through mrVista and for his patience.