Kalanit Grill-Spector - US grants

The ability of the human brain to continuously change and update with experience is one of the fundamental properties that distinguishes it from artificially constructed devices. One manifestation of experience-dependent changes in the human brain is visual priming, in which performance on repeated objects is improved compared to performance on new stimuli (responses are both faster and more accurate). Despite the rich knowledge on the behavioral aspects of visual priming, much less is known about the underlying neural mechanisms. Recent advances in brain imaging have allowed Dr. Kalanit Grill-Spector to examine the brain changes that occur with stimulus repetition. For example, she and her colleagues have found that repeating object images reduces the brain activity in object-selective areas in the human visual cortex. However, it is counterintuitive why reduced cortical activity would be associated with improved behavioral performance? With NSF funding, the goal of Dr. Grill-Spector's project is to understand the neural mechanisms underlying visual priming and to understand how these mechanisms produce changes in cortical activity in object-selective areas in the human brain. These experiments are designed to distinguish between alternative theories explaining visual priming and to rule out non-specific processes, such as changes in the overall attention level. To do so, Dr. Grill-Spector will combine brain imaging experiments and behavioral measurements. By manipulating repetition parameters, top-down attention and stimulus strength, and measuring the effect of these factors on both behavior and brain activation, she will determine which mechanism is more relevant. The results of this project will be critical for understanding the neural basis of experience-dependent changes in the visual system and will have a large impact on the fields of perceptual learning, implicit memory and visual cognition.

These experiments will have an important contribution in understanding the physiological basis of a fundamental and intensively studied cognitive process: visual priming. Elucidating the cognitive and neural bases of visual priming will impact understanding of perceptual learning and plasticity of object representations in high-level visual areas, which in turn will affect a range of fields from clinical aspects of implicit memory to AI models of object recognition. For example, better understanding of the neural basis of perceptual learning or skill learning will facilitate better rehabilitation programs for patients who experienced strokes or other localized brain damage to sensory cortical areas. Furthermore, the proposed work will provide substantial novel insights into the neural processes associated with visual memory, which has important implications for the acquisition of novel object recognition skills. This type of skill is an important aspect of many forms of scientific and medical training (e.g., understanding radiological photos or learning to visually distinguish between chromosomes). In sum, by providing a greater understanding of neural plasticity, this work will inform our understanding of experience-dependent changes in the visual system as well as a wide range of learning and memory phenomena.

DESCRIPTION (provided by applicant): Object recognition is a central topic of research in visual perception and cognitive neuroscience. However, little is known about the nature of object representations in the human brain. What features do we use to recognize objects? How does the brain represent these features? Novel high-resolution fMRI techniques being developed in our lab allow us to examine object representations at a much finer scale than standard fMRI, with voxels as small as 0.5x0.5xlmm. High resolution fMRI provides activations that are more localized and have higher contrast to noise than standard fMRI. Using high resolution fMRI, we propose to develop a new approach to find the set of object images or features that optimally activate regions within object-selective cortex. We propose to examine the fine-scale functional organization of object representation in humans using an event-related design in which we will examine the responses to individual images of objects without pregrouping images into sets that share common features or belong to the same category. In Aim 1 we will examine the reliability of responses to individual object images across trials, and find whether the group of optimal stimuli for activating a given region corresponds to grouping by features, objects or categories. In Aim 2 we will test the hypothesis that certain features, locations and spatial frequency bands within an image account disproportionately for object classification. If we find diagnostic features for object classification, we will then examine using high resolution fMRI whether responses in object-selective cortex correspond to the presence of diagnostic features and whether these activations predicts success at object categorization. Our proposed research will provide: (i) new methods for high resolution fMRI in clinical scanners, (ii) novel approaches for examining object representations in humans and (iii) a comprehensive multidisciplinary study to understand the neural basis of object recognition. Results of these experiments will provide critical constraints on any theory of object recognition.

[unreadable] DESCRIPTION (provided by applicant): Face recognition is crucial for social interaction and development, and undergoes a prolonged maturation until the teens. However, little is known about the maturation of the normal psychological or neural processes that support the development of face perception in children. The goal of this proposal is to use cutting-edge psychophysical and neuroscience methods to elucidate the development of psychological processes involved in face recognition, the underlying neural systems, and their link. Such a rigorous and convergent approach in elucidating normal development is crucial for understanding the many developmental disorders that involve deficits in visual or face processing such as Williams Syndrome, autism and congenital prosopagnosia. Recent psychophysical studies found that face recognition in adults involves multiple stages of processing, including visual categorization (face versus non face) and individual identification (e.g., George versus Bill). Imaging studies revealed that in adults, specific face selective regions in the visual cortex, such as the fusiform "face area" (FFA) are also involved in the categorization and identification of faces. Our preliminary data indicate that the FFA is substantially smaller in children (7-11) compared to adults, correlating with their lower levels of face recognition memory. These data suggest a striking and prolonged maturation of face processing and its neural substrates. However, it is not known how this maturation relates to specific stages of face processing, and whether this maturation is specific to the FFA, or involves the visual cortex more generally. We hypothesize that development of face processing specifically involves maturation of the FFA. Therefore, we propose to examine the links between maturation of the FFA and face perception. We will determine: (1) which behavioral aspects of face processing mature after age 7, using psychophysics; (2) whether the maturation of the FFA occurs in tandem with (or lags) maturation of early and/or higher order visual cortex, using standard fMRI; (3) whether the (smaller) FFA in children is involved in face (or object) categorization and identification, using a combination of psychophysics and standard fMRI and (4) whether maturation of FFA involves increases in face selectivity and/or fine-scale structures within the FFA, using high resolution fMRI. We expect our results to address wide gaps in our understanding of normal visual development, add significant knowledge to theories of face perception and object representation, and provide an essential base for future research on developmental disorders and pediatric imaging in general. [unreadable] [unreadable] [unreadable]

Face perception is critical for social interaction and development, and it undergoes a prolonged maturation that last until the teens. However, little is known about the maturation of the neural or psychological processes that support face perception in children. With support from the National Science Foundation, Dr. Kalanit Grill-spector and her colleagues at Stanford University will conduct a program of research that will use cutting-edge neuroimaging tools (high resolution functional magnetic resonance imaging or fMRI) and behavioral methods to elucidate the maturation of face processing in school age children in terms of the psychological processes involved, the development of the underlying neural systems, and their relation.

Imaging studies conducted with adult participants have revealed face-selective regions in the human visual cortex whose activity is correlated with subjects' performance in face perception tasks. These findings support a key role for face-selective cortex in face perception in adults. The current project will examine whether the maturation of face perception involves maturation of face-selective cortex specifically, or visual cortex generally. Behavioral measures will be combined with fMRI to investigate how the maturation of visual and face-selective cortex affects specific stages of face processing. For example, the studies will examine whether children's performance is different than adults on basic visual tasks such as classifying images (face, car, etc.), whether children's performance is lower for identifying specific individuals and whether performance differs for faces of children vs. faces of adults. This research is important because it will provide a much needed foundation for understanding the basic neural mechanisms of the development of visual perception in children. Understanding normal maturation is also crucial for understanding many learning disabilities that involve deficits in visual perception and developmental disorders that include altered visual perception (such as Williams Syndrome and autism).

The ability to recognize faces is essential for navigating our social world. The human visual system can effortlessly categorize, identify, and remember thousands of faces over a lifetime. Research using functional magnetic resonance imaging (fMRI) has identified several regions in the human visual cortex that specifically and selectively respond to faces, but key questions remain with respect to the neural mechanisms underlying face recognition. For example, it is unclear whether face-selective regions are equally responsive to all human faces, or how perceptual measures of similarity correlate with neural measures of similarity between faces. With support from the National Science Foundation, the investigator will use recent methodological innovations that enable high-resolution fMRI, combined with innovative psychophysical methods and computational models to study the neural basis of within-category representation of faces. The project will identify the fundamental properties that drive responses in face-selective regions, determine whether these responses are tuned to the distribution of faces experienced by individuals over their lifetime, and determine whether measures of similarity of neural responses are tightly related to measures of perceptual or physical similarity among faces. Overall, this research will provide significant advancement in the understanding of how neural responses support our ability to identify individual faces.

The research will have significant implications beyond providing support for a particular computational theory of face representation. It would provide a useful tool for comparing the representations of any other visual category, for example, comparing between neural responses to faces and to objects, which is an issue of central debate. The understanding of neural correlates of normal face identification also provides an important baseline for understanding impairments in face identification as manifested in cogenital prosopagnosia, Asperger's Syndrome and Autism, and as such has broad health and societal implications. This project also aims to advance neuroimaging methods by further developing high-resolution fMRI techniques and by examining whether different experimental designs and analyses for high-resolution imaging show convergent results. Finally, this research will provide training opportunities for students at the undergraduate, graduate, and post-doctoral levels.

Description (provided by applicant): Experience is thought to play a critical role in shaping the cortical representations that support object recognition by creating neural responses are selective for some dimensions of change and invariant to others. Although many previous studies have examined the effects of supervised training on object selective regions of the brain, much less is known about the degree to which statistical regularities in the retinal input can directly shape the neural substrates involved in object recognition. Unsupervised learning is important because it allows the brain to employ simple self organizing mechanisms that turn the continuous flux of visual input into the stable objects of our experience. While behavioral and computational work strongly suggests that unsupervised learning plays a key role in object recognition, most related neuroscience work examining the role of input statistics has focused on its effects in early visual areas. Here we propose experiments that combine cutting edge techniques in fMRI, psychophysics, and computational modeling to examine two hypotheses concerning unsupervised learning in object recognition. First, we propose that neural responses may become tuned to match the range and frequency of shape and object exemplars experienced during unsupervised training. That is, neural responses will increase and become more selective for items seen more frequently during unsupervised training relative to infrequently seen or untrained items. This may provide a mechanism which improves discrimination performance for stimuli seen most frequently. Second, behavioral and computational evidence suggests the intriguing hypothesis that the brain uses spatio-temporal correlations as a means for binding different images as belonging to the same object, allowing for recognition of the same object across dramatic transformations, such as changes in its appearance due to rotation. We will determine if spatio- temporal correlations in the visual input during unsupervised training increases the invariance of both brain responses and perceptual performance relative to similar items trained in an uncorrelated manner and pre- training responses (and performance). Third, we will examine if mechanisms of unsupervised learning generalize to supervised learning. In all of our experiments we will examine neural responses and performance both before and after unsupervised training, and use computational modeling to link fMRI data to the possible underlying neural mechanisms such as sharpening of neural tuning and increased firing rates. The proposed work will fill important gaps in knowledge by providing the first account of the neural mechanisms that generate effective representations for object recognition from the statistics of visual experience. PUBLIC HEALTH RELEVANCE The results of these studies will be important for understanding the role of visual experience in shaping normal visual representations. As these mechanisms do not require explicit instruction, they are especially important for unraveling the means by which pre-verbal children and animals learn to recognize objects. Understanding these mechanisms will form a much needed foundation for studying development disorders such as congenital prosopagnosia, autism and Williams Syndrome. Further, if we find significant behavioral improvements due to the statistics of the visual inputs, these training paradigms may be used as an intervention to offset developmental visual disabilities.

The ability to recognize faces, which is critical for everyday social interactions, improves from childhood to adulthood. In adults, face recognition is mediated by a series of regions in the ventral aspect of human occipital and temporal cortex, constituting the ventral face network. The selectivity and spatial extent of the ventral face network develop from childhood to adulthood, in correlation with age-related improvements in face recognition. However, the neural mechanisms of this development, the rate of development, and the relation among factors that govern this development are not well understood. The goals of the proposed research are to fill these substantial gaps in knowledge by determining the relationship between anatomical and functional brain development and understanding how these developments ultimately lead to improved behavior. To achieve these goals, the research will combine cross-sectional and longitudinal measurements in children (5-10 years old) and adults (23-28 years old) obtaining innovative multimodal measurements of functional magnetic resonance imaging (fMRI), quantitative MRI (qMRI), diffusion weighted imaging (DWI), and behavior in each participant. Aim 1 will employ longitudinal measurements of fMRI, qMRI, and face recognition behavior to determine what is the rate of the development of gray matter and functional selectivity in the face network, if development of gray matter and function occur together or one precedes the other, and if neural developments correlate with behavioral improvements in face recognition. Aim 2 will employ cross-sectional and longitudinal measurements using fMRI, DWI, qMRI, and behavior to determine if and how white matter properties of the face network develop, if white matter developments are linked with either functional or behavioral development, and if development of white matter and function occur together or in sequence. Aim 3 will use fMRI and population receptive field (pRF) modeling to determine if and how pRFs in the ventral face network change from childhood to adulthood, and if development of pRF properties is related to fixation patterns on faces. Aim 4 will test if neural responses in face-selective become less sensitive to face transformations from childhood to adulthood, consequently improving generalization across different instances of the face under different sizes or views. Critically, in each aim we will examine not only the development of the ventral face network but the development of the human ventral visual stream more broadly to elucidate the specificity of developmental effects. Overall, the proposed research will advance understanding of neural mechanisms underlying the development of face recognition, it will elucidate completing developmental theories regarding the relation between anatomical and functional brain development, and will provide the first measurements of the developmental rate of multiple facets of human ventral temporal cortex anatomy and function. This research will provide an essential basis for future research on typical and atypical conditions including developmental prosopagnosia, Williams Syndrome, and autism.

DESCRIPTION (provided by applicant): Humans recognize and categorize the visual input in about a tenth of a second. However, it is still a mystery how the brain achieves this remarkable ability. The cortical visual recognition system consists of a processing stream starting in V1 and ascending into high-level visual areas associated with recognition in ventral temporal cortex (VTC). The goal of the proposed research is to make important theoretical and empirical progress in our understanding of the neural basis of recognition by examining the interplay between neural implementation, representations, and computations in human VTC. Prior research from our lab used high-resolution functional magnetic resonance imaging (HR-fMRI) to advance understanding of the functional organization of VTC by generating an organizational framework detailing its neuroanatomical and topological characteristics. Leveraging these findings, this proposal uses an innovative approach with cutting edge techniques combining HR-fMRI, macro-anatomical, cytoarchitechtonic and myeloarchitectonic measurements, high spatial and angular resolution diffusion imaging (HARDI), advanced tracking algorithms, and computational modeling to address the following three key questions: (1) Is neural microarchitecture an implementational constraint underlying the topological organization of functional representations in VTC? (2) Does structural and functional connectivity regulate functional representations of VTC? (3) How does the neural implementation relate to computations in VTC? Aim 1 will inform if/how the topology of functional representations in VTC is determined by the underlying cytoarchictecture and myeloarchitecture, which may have evolved to optimize particular computations. Aim 2 will investigate the fine-scale functional and structural connectivity of high-level visual cortex determining how information is segregated and integrated within and across adjacent specialized cortical networks. Aim 3 will develop the first generative and quantitative model of VTC computations with the ability to predict responses to stimuli varying in shape, position, and size, while also determining if there is a perceptually-relevant hierarchical processing of information across VTC. This research has important clinical applications for identifying abnormalities in the functional neuroanatomy of VTC within individual brains, and thus, is relevant for patient populations with anatomical or functional VTC deficits, and for individuals with atypical perception or recognition. Overall, the research will break new ground in understanding the neural bases of visual recognition in humans by elucidating the interplay between neural implementation, representations, and computations in human VTC.

Project Summary/Abstract Perception of ecologically relevant visual stimuli such as faces and bodies is achieved through two processing streams extending from early visual cortex (EVC) to lateral occipito-temporal cortex (LOTC) and ventral temporal cortex (VTC), respectively. However, if and how the underlying microstructure and white matter connections constrain the functional organization and support neural computations in these visual streams remains poorly understood. Leveraging advancements achieved in the prior funding period, we propose a unique multimodal approach, combining functional magnetic resonance imaging (fMRI), quantitative MRI (qMRI), diffusion MRI (dMRI), anatomical quantification, and innovative computational modeling to elucidate how structural factors constrain the functional organization of LOTC and VTC. The research has three main aims. Aim 1 will test a quantitative model of functional-anatomical correspondence in high-level visual cortex. Using fMRI, analysis of micro- and macro-structure, the research will quantify the correspondence between macroanatomical landmarks, cytoarchitecture, and functional regions in LOTC and VTC. Aim 2 will determine how white matter connections regulate the functional organization of high-level visual cortex. Using dMRI and fMRI this aim will test (i) if different white matter connections from EVC to downstream regions contribute to the segregation of functional regions within and across visual streams, and (ii) if the eccentricity of the origin of these white matter connections impacts the visual field coverage of downstream regions. Aim 3 will develop and test a spatiotemporal population receptive field model of responses in visual cortex. This aim will provide not only an innovative approach using fMRI and computational modeling to predict responses to a large range of stimuli that vary in size, position, timing, and duration, but will also provide a quantitative framework to test the impact of top-down attention on basic visual computations. Overall, the proposed research will significantly advance understanding of high-level vision by filling in longstanding gaps in knowledge. The research will (1) provide a parsimonious model of how the microstructure and connections scaffold the function and computations of both ventral and lateral streams, (2) break new ground in computational models of visual cortex, and (3) generate innovative multimodal in vivo methods to quantify microstructural properties of visual cortex. Together, the research has important implications for clinical conditions that are associated with malfunction of high-level vision including developmental prosopagnosia, autism, and dyslexia.

PROJECT SUMMARY Extensive research has elucidated the function and structure of adult human visual cortex. However, the developmental mechanisms of human visual cortex are largely unknown for two main reasons. First, there is a paucity of macro-and micro-anatomical data on human brain development outside primary visual cortex (area V1). Second, prior microstructural research on brain development has been done mostly in animal models, but these models are inadequate for elucidating the development of human visual cortex, which has structures that do not exist in other mammals and shows a more protracted development than other species. To address this glaring gap in knowledge, we propose a groundbreaking research project that will generate an exciting, new collaboration between the Grill-Spector lab at Stanford University, who is expert in pediatric in vivo neuroimaging and the Paredes lab at UCSF, who is expert in human pediatric histology and stereology in postmortem brain tissue. Here, we propose to measure the structural development of human visual cortex during infancy using neuroimaging and immunohistochemistry (IHC) methods. The former will use noninvasive neuroimaging to determine macrostructural development of visual cortex longitudinally over 3 timepoints during the first year of life. The latter will use IHC and stereology in postmortem infant brains to elucidate how cellular populations and their microstructures develop. We propose to focus on primary visual cortex (V1), as well as face- and place-selective regions as they (i) can be identified within individual brains from macroanatomical landmarks alone, (ii) are located in different cytoarchitectonic regions, and (iii) show differential development: V1 matures first and face-selective regions last. In Aim 1, we will measure in vivo structural development of visual cortex in infants. Using innovations in quantitative magnetic resonance imaging (qMRI) and diffusion magnetic resonance imaging (dMRI) we will measure for the first time in vivo structural development of primary visual cortex (V1) and high-level visual cortex (face- and place-selective regions) during 3 timepoints in the first year of life (2, 7, and 12 months). In Aim 2, we will quantitatively measure cellular and microstructural development of human visual cortex. Using IHC, we will examine the development of cell types (neurons, astrocytes, oligodendrocytes) and cellular structures (arborization, synapses, and myelin) of infant brain samples that include the calcarine sulcus (where V1 resides), FG (where face-selective regions reside), and CoS (where place-selective regions reside). We will test if the same microstructural mechanisms occur in V1 and high-level visual cortex and produce models from IHC data to relate to neuroimaging data in Aim 1. The proposed research will provide key data that will fill significant gaps in knowledge on visual cortex development, and will pave a new, cutting-edge methodology for quantitative, ground-truth measurements of cortical microstructure in infants. The outcome of our research has important implications for developing non- invasive biomarkers of typical and atypical brain development.