Stanley A. Klein - US grants

The most frequent cause of visual loss in childhood is functional amblyopia. Recent studies show that a high spatial frequencies and under conditions of crowding, qualitatively different acuity losses are found in different amblyopes. The present proposal will capitalize on these differences as a means of clarifying the specific losses in peripheral vision and amblyopic vision compared to normal foveal vision. Just as color deficient individuals had contributed much to our knowledge of color vision, we expect amblyopes will contribute much to our knowledge of spatial vision. Several new psychophysics methodologies will be used to help unravel the comples story: 1. There is a golden opportunity to learn about the amblyopic distortion by comparing the amblyopic eye to the nonamblyopic eye. Under the proper conditions, amblyopes (even those with anomalous correspondence) can not determine the eye-of-origin of a stimulus. By using a double judgment methodology, eye-of-origin confusions will be used to convert subjective interocular comparisons into objective signal detection measurements. New methods for analyzing double judgment tasks with the effects of response bias and response correlation minimized will be used to extend signal detection methodology to the suprathreshold regime. The proposed experiments will focus upon the interactions between local features in peripheral as well as amblyopic vision: 1-Repetitive arrays of features will allow apatial frequency analysis to be used. Spatial frequencies within 3/2 octave of the cutoff spatial frequency will be exhaustively investigated. Effects of phase and background contrast will be explored. 2-A hierarchy of localized luminance distributions will be used to measure contrast sensitivity and phase sensitivity will be studied in amblyopes and peripheral vision. 3. After the first year, nonlocal phase effects, effects of crowding and two dimensional stimuli (ie vernier acuity) will be studies. 4. Quantitative modelling will be used to explain amblyopic and peripheral losses in terms of underlying mechanisms.

Normal visual acuity, that is, the minimum separation required to resolve two features (about 1 min of arc), is fairly well understood. Another class of tasks involves spatial localization of features, the so-called hyperacuities, with thresholds of 1 to 10 seconds of arc. Several models have recently shown how the visual system can achieve this astonishing performance. The demonstration that visual acuity and hyperacuity are relatively unaffected by retinal image motion at velocities ranging up to 3 deg/sec causes problems for many of these models. The importance of achieving good acuity in the presence of image motion should not be underestimated. The task of compensating for retinal motion is likely to be one of the main problems to be solved by the mechanisms of early vision. The eyes are constantly moving, so in order to preserve image quality there must have been strong evolutionary pressures to solve this problem. Dr. Stanley Klein is a leading investigator in the area of spatial vision. Dr. Klein's research has two goals. The first is to understand how high levels of acuity are maintained during image motion. The second goal is to use image motion to isolate distinct localization cues for a variety of hyperacuity tasks. The experiments will test predictions from several models of acuity and hyperacuity. The results of the proposed experiments will also provide additional constraints on a model of human spatial localization for stationary stimuli being developed by Dr. Klein.

DESCRIPTION ( applicant's abstract): Our broad goal is to develop a four stage computational model of spatial vision, based on plausible physiological mechanisms, that predict the performance of normal foveal and peripheral vision, as well as the visual deficits associated with amblyopia. There is a great diversity of masking effects not handled by present models. One limitation of most current models is that they overlook retinal front-end effects and second order effects, such as texture processing. Another major limitation of current models is their reliance on fixed cortical spatial filters, followed by a primitive decision stage. Rather than inventing exotic spatial filters to account for unexplainable data, we suggest that a more parsimonious explanation, based on decision stage limits, can account for much of the data. The six aims of the proposal are organized into three categories: Modelfest: To enhance cross-fertilization among vision modelers, we have organized the Modelfest group. Modelfest is a new approach to modeling that involves the sharing of resources, learning from each other's success and providing a method to cross validate proposed models. Modelfest also facilitates interactions between vision science and medical imaging researchers, two groups with very different approaches to modeling. This interaction has already benefited us on decision stage issues. Our goal is to continue organizing, administering and promoting the group so that progress on vision modeling accelerates (Aim 1). Four-stage model: Current general purpose vision models virtually ignore several important stages of visual processing. We have developed several innovative methods to characterize processing at four stages, from early retinal processes to late decision stage variables. The proposed experiments will be used to define the computational model structure and to set the parameters at each stage. (Aims 2-5) Amblyopia and peripheral vision: Many of the stimuli used to develop the four stage model will also be used to test amblyopes and peripheral vision to determine the stages at which the peripheral and amblyopic visual systems differ from normal foveal vision. Wile past work has focused on early stage differences, there are now indications that some of the losses are taking place at later stages where information is integrated. The experimental results will be used to extend our model of spatial vision so that it predicts the visual losses associated with amblyopic and peripheral vision. (Aim 6)

DESCRIPTION (provided by applicant): Our previous research and that of others show that present models of visual processing are not satisfactory in suprathreshold vision, when a background is present. The proposed research seeks to advance our understanding of suprathreshold vision in known and unknown (noise) backgrounds. We seek to understand contrast processing, visual plasticity and long-range interactions using physiologically plausible mechanisms in normal and amblyopic subjects. Aim 1 develops and refines new methods for experimentally partitioning several factors of noise masking. A highly efficient procedure is introduced for measuring templates and noise in early stages of pattern processing. Aim 2 makes use of these methods for testing a number of hypotheses regarding the detection of a broad class of stimuli in noise. These results have relevance to our understanding of human spatial vision and to many real world tasks such as detecting tumors in medical images and image compression algorithms. Aim 3 is devoted to answering fundamental questions about the factors that limit our ability to perceive changes in contrast. Two types of perturbation will be used to isolate the underlying mechanisms of contrast processing: adaptation and long-range influences of surrounding stimuli. Special attention will be paid to the connection between the perceived contrast and the ability to distinguish contrasts of a test pattern. These studies will be specifically directed to making connections with findings from neurophysiology. Aim 4 applies the basic research findings from the previous aims to perceptual learning. Over the past decade evidence has been gathering that nervous system plasticity enables learning of pattern detection and discrimination with a specificity that indicates early or pre-decision stage learning. The proposed experiments carefully test the hypothesis that the learning is early rather than at the decision stage. Aim 5 applies our methods to gaining a deeper understanding of the amblyopic deficit. Of particular interest is the nature of the perceptual distortion at suprathreshold vision as a manifestation of early cortical organization. The understanding of spatial vision has progressed rapidly such that we can now accurately predict detection thresholds for arbitrary patterns on blank backgrounds using standard filter models. The proposed research will extend our understanding of spatial vision to suprathreshold levels where current models have limited utility. The focus is on understanding contrast processing, visual plasticity and long-range interactions in normal and amblyopic subjects, within a guiding framework of physiologically plausible mechanisms.

[unreadable] DESCRIPTION (provided by applicant): The over-arching goal of this proposal is to develop powerful new methods for identifying multiple cortical sources associated with different early visual areas to reveal the cortical loci of abnormality in amblyopia and strabismic suppression. Technologies such as fMRI, EEG and MEG have enabled us to image the human brain during cognition, but have had limited application for the study of cortical dynamics of closely spaced visual areas. Modem fMRI methods can be used to identify some participating cortical loci but can easily overlook others since a disruption of temporal information within an area may not cause an overall change in cortical activation. Moreover, fMRI's limited temporal resolution obscures feedback processes between cortical areas. EEG and MEG have fine temporal resolution but to date have had limited success in identifying individual cortical sources. The project goals are to develop new methods that solve these problems and apply them to understanding the cortical deficit associated with amblyopia and strabismic suppression. Three Specific Aims are proposed: Aim 1: To develop new VEP/MEG methods for separating multiple cortical sources located in striate and extrastriate areas. Classical methods alone are unable to separate closely spaced early and late visual areas. A successful outcome of this aim will enable cortical dynamics of both early and late visual areas to be reliably distinguished. Aim 2: Amblyopia is a developmental disorder of spatial vision, which occurs in about 3% of the population. Many years of research confirm the deficit is cortical in origin with recent findings indicating involvement of visual areas beyond V1. To reveal the spatio-temporal participation of different cortical areas that contribute to amblyopia and strabismic suppression, we will apply the new evoked response (ER) methodology of Aim 1 that solves past limitations in identifying multiple cortical sources. The research will be able to determine not only where amblyopic/strabismic loss has occurred, but also the role of feedback between visual areas. Aim 3: Use fMRI and psychophysics to validate and extend the findings of ER experiments on the cortical site(s) of the deficit in amblyopia. In addition, novel stimuli such as ambiguous figures that are not readily employed using ER methods, will be employed to examine and distinguish areas that participate in the deficit associated with amblyopia. [unreadable] [unreadable]

DESCRIPTION (provided by applicant): The broad aim of our proposed research is to gain a deeper understanding of the mechanisms of perceptual learning (PL) in normal and compromised visual systems that could potentially benefit from effective perceptual learning paradigms. We will identify the mechanisms underlying the losses and to what extent perceptual learning can ameliorate these losses. In addition, by identifying the underlying mechanisms of PL we anticipate being able to explain the large individual differences commonly found in PL studies. Aim 1) The role of context in the generalization of perceptual learning. It is generally thought that PL does not transfer across tasks. However, we show examples using Double Training where learning does generalize. We will characterize the conditions for transfer and use advanced psychophysical methods to reveal the underlying mechanisms. In the test-pedestal method, the threshold for detecting the same test stimulus is measured in different contexts. Just by changing the context the task changes, yet the ideal observer would use the same template for each task. We hypothesize that perceptual learning in one context will not completely transfer to the related tasks. The methods will enable us to follow the development of the template during learning and what happens when the task changes even though the ideal template remains the same. Aim 2) The generalization of perceptual learning over space. The belief that PL involves modification of early visual areas is based on earlier findings that it does not transfer across retinal locations. However, our double training method does result in transfer of learning across locations, including locations that have not received any prior training. This new perspective on PL points to a crucial role for higher brain areas that engage attention and decision-making. Our preliminary studies have revealed a piggyback effect whereby promoter stimuli enable the transfer of learning of a second feature. We hypothesize the distinction of promoter is related to eye movements and global vs. local processing. Our methods enable measuring the perceptual template and its stability, provide estimates of internal noise and evaluate the role of response bias in PL. These methods are the most powerful to date to elucidate the processes associated with PL. Aim 3) Applying perceptual learning to improve visual function in people with compromised vision. Declining visual performance with age is well documented. Not surprising is the growing interest in new methods of enhancing visual function. It is reasonable to assume that training protocols can be enhanced by PL advances. By understanding the fundamental mechanisms of perceptual learning, future perceptual learning paradigms will be all the more effective in improving the quality of life of individuals with perceptual deficits, such as the elderly and persons with amblyopia. Individual differences on PL tasks are large and we have only speculative understanding of their source. Our new methods will enable us to identify the mechanisms that most account for the individual differences to provide guides for improving training methods. PUBLIC HEALTH RELEVANCE: Perceptual learning is an important way to enhance visual function in the elderly, amblyopia, and observers with normal vision. By understanding mechanisms of perceptual learning, future paradigms will be all the more effective in improving the quality of life for individuals with perceptual deficits.

In our daily lives our brains are constantly adapting to new visual experiences and learning optimal solutions to new tasks. Understanding the mechanisms of neural plasticity is crucial for developing effective training paradigms. With support from the National Science Foundation, Professor Stanley Klein of the University of California, Berkeley, is carrying out a research project to develop new methods for revealing brain dynamics and brain structures that support neural plasticity. In the past 20 years, functional magnetic resonance imaging (fMRI) has identified numerous brain regions that are involved in perception, cognition, attention, decision-making, action, and other functions. But fMRI has access only to slow brain changes, on the order of seconds. Much of the important neural processing by the brain takes place a thousand times faster. A goal of this project is to develop new methods that will enable human brain waves from electroencephalography (EEG) to be used for identifying brain areas and tracking their activity on the milli-second time scale. The proposed research makes use of recently developed computer algorithms that use the detailed folding patterns of the human brain to isolate the generators of EEG activity. The first application of the new methods is investigations of how cortical activity in closely spaced brain areas change as a product of learning perceptually challenging visual tasks. However, the importance of this research extends well beyond the domain of visual perceptual learning: Once scientists can reliably measure cortical activity in closely connected brain areas, they will be able to follow in temporal detail the flows of activity across distinct brain areas during cognitive activity.

The visual perception of patterns is so fundamental to our daily activities that when it is disrupted, the disruption often leads to an especially devastating impact on quality of life. This research on neural plasticity has the goal to develop methods that could inform not only these health issues but the visual learning process as applied to general education, vocational training, and many other aspects of daily living. During the project, the immediate educational impact is the involvement of graduate and undergraduate students in the research. The research forms the core dissertation research for graduate students. Professor Klein's laboratory is an active participant in the campus undergraduate research apprentice program (URAP) and admits several new undergraduate students each semester. The laboratory meets to discuss relevant research questions and findings. The students learn all aspects of EEG recording technology and get first hand research experience. After working in the laboratory, many URAP students have decided on and succeeded in higher education, be it basic research or clinical medicine. The techniques and software that are being developed will be posted on the internet for other researchers to use. The methods are sufficiently general to be useful to investigators who study sensory or higher cognitive functions.