Jonathan D. Victor - US grants

I plan related studies of neural interactions at two levels of the visual system: the retina of the cat and the visual cortex of man. Retinal signal-processing will be investigated by studying responses of single cat retinal ganglion cells. Stimuli will be generated by a computer-controlled electronic display; neural activity will be monitored with microelectrodes inserted in the optic tract of anaesthetized adult cats. Ganglion cell responses will be interpreted in terms of models of retinal processing that include not only the classical center and surround mechanisms, but also the nonlinear subunit pathway of the Y cell and the contrast gain control. A variety of mathematical techniques, including linear systems analysis and the sum-of-sinusoids method of nolinear systems analysis, will be used to construct and analyze physiological models. A concise and accurate description of retinal processing will serve as a model for understanding of information processing by the local circuits of the mammalian central nervous system. Visual processing in the human cortex will be investigated by studying evoked responses to novel classes of visual stimuli. These stimuli are designed to separate signals due to cortical processing from signals due to precortical mechanisms, and consist of discriminable texture pairs with identical power spectra. Alternation between such textures evoke asymmetric responses which occur with greater latency than the traditional pattern-reversal evoked responses. This technique will be used to examine population properties of cortical visual neurons, such as the role of cooperative interactions and the spatial selectivity of cortical feature-detectors. The possibility that this technique may provide a sensitive probe of abnormalities of cortical function will be explored by studying evoked responses in patients with metabolic and structural neurologic disease.

The retina is a unique model for studying neuronal processing in general: the input is easily accessible to patterned physiological stimuli and the anatomy is organized and well-characterized. Retinal signal-processing will be investigated by studying the responses of cat retinal ganglion cells to patterns of light in space and time. These responses will be interpreted in terms of models of retinal processing that include not only the classical center and surround mechanisms, but also the nolinear subunit pathway of the Y cell, adaptation, and the contrast gain control. The models will be tested by their ability to predict responses to a wide range of visual stimuli. Patterns of light will be generated by a computer-controlled electronic display with high spatial and temporal resolution. Gangion cell activity will be monitored with microelectrodes inserted in the optic tracts of anaesthetized adult cats. A variety of mathematical techniques, including linear systems analysis and the sum-of-sinusoids method of nonlinear systems analysis, will be used to construct and analze physiological models. The immediate benefits of a full concise model of retinal processing are a clarification of the importance of the many retinal mechanisms that have been proposed, and an insight into the purpose of separation of visual information into the parallel X and Y channels. The long-term benefit of this project is a better understanding of information processing by the local circuits of the mammalian central nervous system.

This project consists of a set of inter-related experimental, computational, and theoretical studies, whose goal is to advance the understanding of the design principles of visual processing, and how these design principles are reduced to computations that can be carried out by neurons and neural circuitry. The many successes of ?normative theories? constitute our starting point: for example, how a sensory system's limited capacity should be deployed to effectively represent and transmit task-relevant information about its inputs. However, here we recognize that along with these successes ? in our lab and many others -- there are many divergences between normative predictions and what the visual system actually does. These discrepancies indicate that there are important constraints not recognized by current normative theories, such as limits to the detail with which natural-image priors are used. We focus on the extraction of figure from ground: this is a computationally-challenging process that is centrally important to visual function, and it also has a number of characteristics that we can use to advantage, building on recent advances in our lab. Distinguishing figure from ground is a fundamentally statistical process, so understanding how the visual system processes local image statistics is critical. In previous years, we developed a theoretical and experimental framework for this: we showed how luminance, contrast, orientation, and shape could be dissociated via the construction of a space of synthetic textures, and we then used this space to measure human visual sensitivity to these components individually and in combination and to analyze its relationship with natural-image statistics. Aim 1 consists of psychophysical experiments to characterize three key aspects of figure-ground processing: Aim 1A, the influence of the statistics of figure, ground, and figure-ground differences, Aim 1B, the influence of figure shape, and Aim 1C, the influence of task-specific knowledge. Aim 2 is motivated by models that formalize the hypothesis that visual computations make use of simplified Gaussian approximations to natural image statistics. To test these models, Aim 2A consists of computational studies to determine the statistics of image patches in figure and ground. Aim 2B makes further psychophysical measures that will determine a phenomenological model. Comparison of the phenomenological model and normative models built from natural-image statistics will proceed in stages: does the phenomenological model have the form predicted by normative theories (i.e., do measured threshold surfaces have the predicted shape)?I f so, what is the level of detail of natural-image priors that are needed to account, quantitatively, for perceptual thresholds? Successful completion of this research is expected to provide both specific and generalizable insights into principles of sensory processing, which in turn will provide the groundwork for advanced neural prosthetics and assistive devices.

The proposed research consists of two complementary approaches to further the understanding of how visual information is represented and processed in visual cortex, particularly primary visual cortex. Aims I will focus on the temporal structure of the neural discharge. These aims will determine (a) the extent to which the temporal pattern of a neural discharge contributes to the representation of particular kinds of visual information, including elementary spatial features and color; (b) whether the temporal representation of visual information can facilitate subsequent visual processing; and (c) the relationship of temporal coding (representation of visual information within the spike train of a single neuron) to spatial coding (representation of visual information across two or more neurons). Standard visual stimuli (bars, gratings, and compound gratings) will be used. The usual analytical approaches of histogram, Fourier, and cross-correlation analysis will be augmented by a family of new methods recently developed in this laboratory, based on the embedding of neural responses into a metric space. Aim II will focus on the generation of response dynamics by subregions of the receptive field. The main strategy will be an adaptation of the m-sequence technique to allow analysis of the temporal structure of the response. Through comparison of receptive field maps generated by all spikes, by bursts, and by "reliable" spikes, we will determine to what extent these temporal structures participate in the signaling and extraction of features. The m-sequence analysis will show the extent to which these temporal structures arise from an overall modulation of response properties, specific receptive field subregions, or specific interactions over an extended region of space. The goals of these quantitative experiments are new qualitative insights into how the visual cortex represents and processes information, on phenomenological (Aims I) and mechanistic (Aim II) levels. This will serve as a basis for a better understanding of brain function in health and disease.

This project was developed at an NSF Ideas Lab on "Cracking the Olfactory Code" and is jointly funded by the Physics of Living Systems program in the Physics Division, the Mathematical Biology program in the Division of Mathematical Sciences, the Chemistry of Life Processes program in the Chemistry Division, and the Neural Systems Cluster in the Division of Integrative Organismal Systems. The project is a synergistic combination of laboratory experiments and computer modeling that will lead to better understanding of how animals use the sense of smell to navigate in the real world. Almost universally, from flies to mice to dogs, animals use odors to find critical resources, such as food, shelter, and mates. To date, no engineered device can replicate this function and understanding the code used by the brain will lead to many novel applications. Cracking codes, from neural codes to the Enigma code of WWII, is aided by a deep understanding of the content of messages that are being transmitted and how they will be used by their intended receivers. To crack the olfactory code, the team will focus on how odors move in landscapes, how animals extract spatial and temporal cues from odor landscapes, and how they use movement for enhancing these cues while progressing towards their targets. The proposed work encompasses physical measurement of odor plumes, behavioral measurement of animals' paths through olfactory environments, electrophysiological and optical measurement of neural activity during olfactory navigation, perturbations of the environment via virtual reality and of neuronal hardware via genetics, and multilevel mathematical modeling. The PIs will teach and work with undergraduate, graduate and postdoctoral students and especially recruit students from underrepresented groups in science. The project's results may lead to improved methods for the detection of explosives, new olfactory robots to replace trained animals, and new theoretically-grounded advances in robotic control. The project will inform the development of technologies that interfere with the ability of flying insects (including disease vectors and crop pests) to locate their odor target, thus opening a new door for developing 'green' technologies to solve problems that are of global economic and humanitarian importance.

This proposal is a synergistic combination of laboratory experiments and computational modeling that will probe how animals use olfaction to navigate in their environment. Specifically, this effort seeks to solve the difficult problem of olfactory navigation through the following aims: (i) Generate and quantify standardized, naturalistic odor environments that can be used to perform empirical and theoretical tests of navigation strategies; (ii) Determine phenomenological algorithms for odor-guided navigation through behavioral experiments in diverse animal species; (iii) Determine how odor cues for navigation are encoded and used in the nervous system by recording neuronal data and simulating putative neural circuits that implement these processes; (iv) Manipulate olfactory environments and neural circuitry, to evaluate model robustness. In contrast to previous attempts to understand olfactory navigation, the present strategy emphasizes mechanisms that are biologically feasible and explores the wide range of temporal and spatial scales in which animals successfully navigate. The project will generate datasets of immediate use and importance to scientists in theoretical biology and mathematics, engineering (fluid mechanics, electronic olfaction, and robotics) and biology (neuroscience, ecology and evolution).

Project Summary/Abstract It is well known that normal anatomy in a medical image can mask the presence of disease. However this process is not well understood. Part of the problem is that we lack knowledge of the relevant statistical descriptors that characterize perceptual effects of image statistics. While image acquisition noise is largely characterized by its second moments (power-spectrum or covariance matrix), background anatomy has a complex structure that requires higher-order statistics ? and an understanding of their perceptual relevance ?to characterize fully. This is an important limitation because reading ?through? this background is a critical component of many clinical tasks. In a statistical sense, reading through the background means exploiting redundancies in the presentation of normal anatomical structures for the purpose of isolating disease processes. The need for background characterization is well recognized in screening mammography, our focus, as screening mammography typically includes an assessment of the background via the BIRADS density score. However, this score has limited utility as a statistical descriptor. The basis for this project is to translate a successful approach from basic vision science to medical imaging, in order to identify the relevant high-order statistical properties of medical images and their perceptual impact. In this approach, a set of local image statistics (co-occurrence probabilities) are used to build an ?alphabet? for the statistical structure of synthetic visual textures and their local features (such as edges). Perceptual sensitivities to local features can be concisely characterized and modeled via this alphabet, and it has been shown that sensitivity to these elements is matched to their informativeness in natural scenes. This motivates our general approach, and many specifics of our research plan. Our plan is to develop algorithms in Aim 1 that selectively alter (either increase or decrease) the co-occurrence statistics of mammograms, while retaining their general background appearance. The sub-aims explore four strategies, building on a Fourier domain approach for which a proof-of-principle is in hand. Then, Aim 2 will use these images to assess perceptual sensitivity. Aim 2A will develop the psychophysical paradigm. Aims 2B-D will determine whether the principles identified in previous studies of synthetic visual textures (sign-invariance, approximate scale-invariance, and quadratic combination) extend to medical images, as this will enable a comprehensive yet concise description of perceptual sensitivity. We will pursue these aims using a database of full-field digital mammograms. The project is expected to yield a validated approach for modulating high-order statistical properties of mammograms and baseline data of perceptual sensitivity to these modulations. These findings will improve our understanding how normal anatomy impacts the statistical properties of screening mammograms, and give us valuable baseline data on how the statistics of normal anatomy affect perception.