Bay Area Vision Research Day 2018

The matrix metalloproteinases (MMPs) are the most prominent family of proteinases involved in corneal wound repair. Recently the substrate repertoire of MMPs has expanded to include cell membrane-bound precursors of many growth factors, growth-factor receptors, cell adhesion molecules and chemokines. Their role in various biological processes, including tissue repair, has similarly expanded and evolved. The important regulatory roles of MMPs in various physiological and pathological processes including inflammation, neovascularization, and fibrosis will be discussed.

The visual system has evolved to function within the natural environment. Our understanding of the organizing principles of vision has been substantially advanced by examining statistical regularities in natural visual imagery. Much less is known, however, about how the natural environment shapes 3D vision. I will describe work aimed towards advancing our understanding of the geometric and optical properties of natural 3D environments, and how this work can deepen our understanding of visual perception within 3D space.

We have recently shown that sequences of random pixel arrays can produce illusory percepts of coherent apparent motion, and further that the nature of the illusory motion can be primed by preceding veridical motion stimuli. In this talk I will describe how bottom-up factors (e.g. the velocity, frame rate, and duration of priming) and top-down factors (e.g. range of response options and instructions to participants) interact to modulate the experienced motion during the random pixel arrays. When the priming motion is fast, smooth, and long-lasting, subsequent random frames are usually perceived to have motion in the opposite direction of the prime (similar to motion aftereffects). But when the priming motion is slow, temporally discrete, and short-lasting, subsequent random frames are perceived to continue the motion of the prime. These positive priming effects manifest differently across participants, are subject to attentional manipulation, and can carry more complex motion information from the motion prime than is typically attributed to the motion priming system. For example, participants can be primed to experience not only drifting (i.e. unidirectional) motion sequences, but also rebounding (bi-directional) and even staircase (bi-axial) motion sequences. The ability to prime 2-step motion sequences suggests the intriguing possibility that working memory processes are involved in the perception of apparent motion.

When a contact lens is inserted into the ocular environment, it necessarily comes into intimate contact with the cornea epithelium. This contact provides an opportunity for the corneal epithelial cells to physically attach through adhesive bonds onto the lens. This lens-to-cell adhesion will affect lens comfort and can cause cell damage upon removal of the lens. Consequently, there is an interest in formulating lens surfaces to minimize such adherence. We report the use of a rheometer, a device to measure the relationship between stress and strain, to directly measure evidence of lens-to-cell adhesion. This quantitative measure of the strength of adhesion is reported as a function of the crosslink density of silicon hydrogel contact lenses and it is found that adhesion is diminished as the crosslink density is increased.

In glaucoma and other optic neuropathies, retinal ganglion cells (RGCs) die and are not replaced from endogenous sources, leading to a permanent loss of vision. In recent years, we and others have made significant progress in understanding the molecular regulation of RGC differentiation during normal retinal development, and in so doing have uncovered fundamental biological principles of ligand opponency and transcription factor compensation. Furthermore, we have explored the relevance of such developmental pathways in the regulation of human stem cell-to-RGC differentiation. Finally, we have begun to make progress in RGC transplant and integration, with attention to local and distal wiring and synapses. Implications for the future include application to disease models and translation to human testing.

Retinal degenerative diseases lead to blindness due to loss of the “image capturing” photoreceptors, while neurons in the image-processing inner retinal layers are relatively well preserved. Information can be reintroduced into the visual system using electrical stimulation of the surviving inner retinal neurons. Some electronic retinal prosthetic systems have been already approved for clinical use, but they provide low resolution and involve very difficult implantation procedures.

We developed a photovoltaic subretinal prosthesis which converts light into pulsed electric current, stimulating the nearby inner retinal neurons. Visual information is projected onto the retina by video goggles using pulsed near-infrared (~880nm) light. This design avoids the use of bulky electronics and wiring, thereby greatly reducing the surgical complexity. Optical activation of the photovoltaic pixels allows scaling the implants to thousands of electrodes, and multiple modules can be tiled under the retina to expand the visual field.

We found that similarly to normal vision, retinal response to prosthetic stimulation exhibits flicker fusion at high frequencies (>20 Hz), adaptation to static images, antagonistic center-surround organization and non-linear summation of subunits in the receptive fields, providing high spatial resolution. In rats blinded by retinal degeneration, photovoltaic arrays with 55μm pixels enabled grating visual acuity up to a pixel pitch. If these results translate to human retina, such implants could restore visual acuity up to 20/200. 3-dimensional electro-neural interfaces enable even smaller pixels, which may provide visual acuity better than 20/100. Ease of implantation of these wireless arrays, combined with high resolution opens the door to highly functional restoration of sight.

Observers rewarded for finding color targets in a training task can later have their attention captured by those rewarded colors in an oddball shape search, a phenomenon known as value-driven attentional capture (VDAC). In the current work, observers were rewarded for finding simple shapes (circles and diamonds) in the training phase. In the test phase searching for an oddball color target, previously rewarded shapes slowed response times (RTs) when presented asdistractors and speeded RTs when presented as targets. We will present evidence that VDAC to shapes is not as strong as VDAC to colors, but still present. In further experiments, we establish that the color and shape feature dimensions are independent and additive, even when trained in conjunction. It seems that VDAC potentiates features dimensions independently, even when they are trained in conjunction.

The rich perceptual experience of daytime vision results from neural processing of signals originating in individual cone photoreceptors. Two fundamental operations applied to cone signals by post-receptoral circuitry are summation and inhibition. I will discuss experiments in which these elaborations were probed psychophysically at the single-cone scale using an adaptive optics ophthalmoscope equipped with retinal tracking and dynamic stimulus delivery.

First, I will describe work wherein spatial summation of luminance increments was measured in the central fovea both with and without high-order aberrations and retinal stimulus motion. We found the region over which foveal cone signals were pooled completely—spanning about two dozen photoreceptors, on average—was essentially invariant to these pre-retinal factors, suggesting a post-receptoral origin for psychophysical measurements of spatial summation.

Second, I will present results from a study in which single-cone thresholds were measured against chromatic backgrounds that preferentially altered the steady-state activity of L, M, and S cones. Comparing the resultant sensitivity maps to the underlying spectral topography of the cone mosaic (obtained via densitometry) revealed evidence of spatial antagonism between adjacent cones. These spatiochromatic interactions occurred at spatial scales consistent with the anatomical dimensions of H1 and H2 horizontal cells in the central retina.

Humans are sensitive to information about collections of objects via ensemble or summary statistical representations. Ensemble perception is efficient and provides rich information ranging from low-level features to visuo-social cues. It also provides emergent information that is not available at the level of any individual object. One frequently encountered but puzzling emergent summary statistical percept is an amplification effect: the perceived average in a crowd of objects is sometimes exaggerated. This occurs for average object size, average object and face attractiveness, average temporal frequency, and even judgments of average value or price of a crowd of objects. Ensemble amplification is not specific to spatial information or low-level features, but seems to be a persistent characteristic of visual ensemble coding. Here we consider the possible reasons for the amplification of ensemble percepts, which we will argue help constrain the nature of ensemble coding models.

Advances in computer vision (and more broadly in machine learning) are gradually transforming the policies and practices of governments, including on issues like poverty reduction, disaster response, and economic development. This transition will require multi-disciplinary teams of researchers that span computer science, public health, economics, and other fields of social science.It also raises issues related to the
reproducibility, fairness, and transparency of machine learning algorithms, which can profoundly affect millions of people's lives.Examples of ongoing work between bay area academics and policy-makers will be discussed, including projects in Uganda, Rwanda, Afghanistan, and India.

When we look at the world we experience it as a whole. Space appears as a continuum and objects are located in different places as they should be. However, there is a great deal of neuropsychological and neurobiological evidence that cortical visual areas involved in visual perception do not immediately process a scene as a whole but rather in fragments and features. There is no more dramatic evidence for this than the different ways visual perception breaks down after stroke or brain injury (e.g., Balint’s syndrome, unilateral neglect). I will review results of experiments with neurological patients that I have studied over the years. The results provide insights into fundamental visual processing in the human brain and strongly support a role for spatial attention in feature integration (.i.e., the binding problem).