For over half a century, vision scientists have been decomposing visual scenes into simple, more tractable components in an attempt to understand how the brain accomplishes vision. Although this endeavor has revealed much about the specialized subsystems of vision, surprisingly little is know about how, or even where in the brain, we process scenes as a whole. How is it, for instance, that the brain determines whether it is looking at a forest or a city skyline? One reason for the paucity of research on this topic may be that the neural representation of a scene is likely to be highly distributed, a coding scheme not easily identified by many traditional neuroscience methods. The objective of the proposed research is to use a new method of analyzing functional magnetic resonance imaging (fMRI) data that is designed to leverage activity patterns across the brain, in order to better understand how the brain categorizes natural scenes. In particular, the project combines expertise from computer vision and neuroimaging by applying statistical pattern recognition algorithms to fMRI data to understand how the brain distinguishes between different categories of natural scene (e.g., a beach versus a highway). The proposed project will use and develop a statistical pattern recognition approach to fMRI analysis to accomplish three more specific objectives: (i) to identify the neural representation of natural scene categories, (ii) to identify the computational principles for forming and using the neural representation of natural scene categories, and (iii) to explore the effects of attention and expectation on natural scene categorization. The insights gained from these experiments will be verified in a computational model of natural scene perception, which in turn will generate predictions for future experiments. Intellectual Merit of the Proposed Activity: Although previous research has shown that humans can quickly and effortless categorize natural scenes, there is very little understanding of how this is accomplished in the brain. The research proposed here will significantly advance our understanding of how natural scenes are represented in the brain and begin to uncover the computational strategies the brain employs in quickly and accurately extracting the gist of a scene. Broader Impacts of the Proposed Activity: The highly interdisciplinary nature of the proposed research requires intense interactions among psychologists, neuroscientists, and computer vision researchers. As such, the project not only promises to increase communication among very different disciplines but it will also to provide doctoral students with truly interdisciplinary training. The PIs are committed to providing a highly interactive research environment, mentoring students across disciplines, and fostering the interdisciplinary approach to science in general. Moreover, two of the three PIs are women working in fields in which women are traditionally underrepresented and are committed to improving the representation and visibility of women in science. Finally, the principles derived from this project are likely to have implications beyond the domain of natural scene perception. By refining the pattern recognition algorithms and their application to fMRI data, the project will expand the set of tools available to neuroscientists wishing to study a whole host of complex human behaviors that likely depend on subtle but distributed patterns of activity in the brain.

Agency
National Institute of Health (NIH)
Institute
National Eye Institute (NEI)
Type
Research Project (R01)
Project #
5R01EY019429-05
Application #
8142855
Study Section
Special Emphasis Panel (ZRG1-IFCN-B (50))
Program Officer
Steinmetz, Michael A
Project Start
2008-08-01
Project End
2013-07-31
Budget Start
2011-09-30
Budget End
2013-07-31
Support Year
5
Fiscal Year
2011
Total Cost
$311,026
Indirect Cost
Name
Stanford University
Department
Biostatistics & Other Math Sci
Type
Schools of Engineering
DUNS #
009214214
City
Stanford
State
CA
Country
United States
Zip Code
94305
Caddigan, Eamon; Choo, Heeyoung; Fei-Fei, Li et al. (2017) Categorization influences detection: A perceptual advantage for representative exemplars of natural scene categories. J Vis 17:21
Iordan, Marius C?t?lin; Greene, Michelle R; Beck, Diane M et al. (2016) Typicality sharpens category representations in object-selective cortex. Neuroimage 134:170-179
Greene, Michelle R; Botros, Abraham P; Beck, Diane M et al. (2015) What you see is what you expect: rapid scene understanding benefits from prior experience. Atten Percept Psychophys 77:1239-51
Walther, Dirk B; Shen, Dandan (2014) Nonaccidental properties underlie human categorization of complex natural scenes. Psychol Sci 25:851-60
Baldassano, Christopher; Beck, Diane M; Fei-Fei, Li (2013) Differential connectivity within the Parahippocampal Place Area. Neuroimage 75:228-37
Torralbo, Ana; Walther, Dirk B; Chai, Barry et al. (2013) Good exemplars of natural scene categories elicit clearer patterns than bad exemplars but not greater BOLD activity. PLoS One 8:e58594
Baldassano, Christopher; Iordan, Marius Catalin; Beck, Diane M et al. (2012) Voxel-level functional connectivity using spatial regularization. Neuroimage 63:1099-106
Walther, Dirk B; Caddigan, Eamon; Fei-Fei, Li et al. (2009) Natural scene categories revealed in distributed patterns of activity in the human brain. J Neurosci 29:10573-81