Purchase this article with an account.
Matthew L. Patten, Andrew E. Welchman; Decoding fMRI responses to disparity-defined depth configurations. Journal of Vision 2012;12(9):40. doi: https://doi.org/10.1167/12.9.40.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Human brain imaging evidence suggests that disparity-defined structures—from simple planes to complex 3D shapes—evoke discriminable responses in multiple brain areas. However, the nature of the representations in different areas remains unknown. We measured fMRI responses to center-surround configurations of two depth planes, examining how (1) changes in the relationship between center and surround and (2) changes in the position in depth of the entire stimulus, affect cortical activity. Observers (N=8) viewed random dot stereograms depicting a central target plane (±3, ±9 or ±15 arcmin) surrounded by an annulus (±6 arcmin). We used multivoxel pattern classification analysis (MVPA) to quantify information about depth configurations in visual cortical areas. First, we contrasted MVPA accuracies when target depth was defined relative to (i) fixation or (ii) the surround. We found that early visual areas (V1, V2) are most sensitive to depth differences expressed relative to fixation. In contrast, higher ventral area LO is more sensitive to the center relative to its surround, suggesting responses related to depth configuration. Dorsal (V3A, V3B/KO, V7) and intermediate ventral (V4) areas appeared equally sensitive for depth differences expressed relative to fixation or the surround. Second, we evaluated how near vs. far decoding changed as a function of disparity difference, finding that performance in most areas increased as the difference between near and far stimuli increased, while performance was constant in LO. Finally, we assessed transfer between conditions (training the classifier on one stimulus configuration and testing on another). We find that responses in LO do not support transfer across position in depth. These results highlight the role of higher ventral areas in computing information about depth configurations, but suggest that depth representations in LO are not fully abstracted and are modulated by the position of objects relative to the observer.
Meeting abstract presented at VSS 2012
This PDF is available to Subscribers Only