Retinal Sampling Research Articles

Measuring Internal Representations from Behavioral and Brain Data

The study of internal knowledge representations is a cornerstone of the research agenda in the interdisciplinary study of cognition. An influential proposal assumes that the brain uses its internal knowledge of the external world to constrain, in a top-down manner, high-dimensional sensory data into a lower-dimensional representation that enables perceptual decisions and other higher-level cognitive functions [1-9]. This proposal relies on a precise formulation of the observer-specific internal knowledge (i.e., the internal representations, or models) that guides reduction of the high-dimensional retinal input onto a low-dimensional code. Here, we directly revealed the content of subjective internal representations by instructing five observers to detect a face in the presence of only white noise, to force a pure top-down, knowledge-based task. We used reverse correlation methods to visualize each observer's internal representation that supports detection of an illusory face. Using reverse correlation again, this time applied to observers' electroencephalogram activity, we established where and when in the brain specific internal knowledge conceptually interprets the input white noise as a face. We show that internal representations can be reconstructed experimentally from behavioral and brain data, and that their content drives neural activity first over frontal and then over occipitotemporal cortex.

Axotomy-induced retinal ganglion cell death in adult mice: Quantitative and topographic time course analyses

The fate of retinal ganglion cells after optic nerve injury has been thoroughly described in rat, but not in mice, despite the fact that this species is amply used as a model to study different experimental paradigms that affect retinal ganglion cell population. Here we have analyzed, quantitatively and topographically, the course of mice retinal ganglion cells loss induced by intraorbital nerve transection. To do this, we have doubly identified retinal ganglion cells in all retinas by tracing them from their main retinorecipient area, the superior colliculi, and by their expression of BRN3A (product of Pou4f1 gene). In rat, this transcription factor is expressed by a majority of retinal ganglion cells; however in mice it is not known how many out of the whole population of these neurons express it. Thus, in this work we have assessed, as well, the total population of BRN3A positive retinal ganglion cells. These were automatically quantified in all whole-mounted retinas using a newly developed routine. In control retinas, traced-retinal ganglion cells were automatically quantified, using the previously reported method (Salinas-Navarro et al., 2009b). After optic nerve injury, though, traced-retinal ganglion cells had to be manually quantified by retinal sampling and their total population was afterwards inferred. In naïve whole-mounts, the mean (±standard deviation) total number of traced-retinal ganglion cells was 40,437(±3196) and of BRN3A positive ones was 34,697(±1821). Retinal ganglion cell loss was first significant for both markers 5 days post-axotomy and by day 21, the last time point analyzed, only 15% or 12% of traced or BRN3A positive retinal ganglion cells respectively, survived. Isodensity maps showed that, in control retinas, BRN3A and traced-retinal ganglion cells were distributed similarly, being densest in the dorsal retina along the naso-temporal axis. After axotomy the progressive loss of BRN3A positive retinal ganglion cells was diffuse and affected the entire retina. In conclusion, this is the first study assessing the values, in terms of total number and density, of the retinal ganglion cells surviving axotomy from 2 till 21 days post-lesion. Besides, we have demonstrated that BRN3A is expressed by 85.6% of the total retinal ganglion cell population, and because BRN3A positive retinal ganglion cells show the same spatial distribution and temporal course of degeneration than traced ones, BRN3A is a reliable marker to identify, quantify and assess, ex-vivo, retinal ganglion cell loss in this species.

The Robustness of Various Forms of Perimetry to Different Levels of Induced Intraocular Stray Light

To compare directly the robustness of standard automated perimetry (SAP), short-wavelength automated perimetry (SWAP), frequency-doubling perimetry (FDP), and grating-resolution perimetry (GRP) stimuli to different degrees of intraocular stray light induced by commercially available opacity-containing filters. Five white opacity filters of increasing density were used to simulate the typical forward light scatter and stray light values associated with age-related lens opacification and significant cataract. The individually induced intraocular stray light value for each filter was quantified with a stray light meter and plotted against individual perimetric thresholds for the right eyes of three normally sighted trained observers for SAP, SWAP, FDP, and GRP. All tests were significantly but differently affected by increasing stray light. Overall average declines over a 1 log unit change in the stray light values were as follows: SAP, 4.85 dB; SWAP, 9.03 dB; FDP, 4.29 dB; and GRP, 1.36 dB. Standardized (z) scores were calculated after normalization to the spread of the normative data values for each instrument. These indicated that the standardized changes from baseline over the range of the five filters per log stray light unit were as follows: SAP, 2.177; SWAP, 1.96; FDP, 1.277; and GRP, 1.04. The increased stray light values induced by cataract-simulating filters has a significant effect on all tests. However, GRP, which is known to be limited by retinal sampling rather than contrast, remains the most robust of the tests to the effects of intraocular stray light. The degree to which the normative "sensitivity" range for different types of perimetry might incorporate a component caused by individual differences in intraocular stray light is discussed and requires further research.

Robust gender and age estimation under varying facial pose

AbstractThis paper presents a method for gender and age estimation which is robust to changing facial pose. We propose a feature point detection method, called the adapted retinal sampling method (ARSM), and a feature extraction method. A basic concept of the ARSM is to add knowledge about the facial structure to the retinal sampling method. In this method, feature points are detected on the basis of seven points corresponding to facial organs from a facial image. The reason why we used seven points as the basis of feature point detection is that facial organs are conspicuous in the facial region, and are comparatively easy to extract. As features robust to changing facial pose, skin texture, hue, and the Gabor jet are used for gender and age estimation. For classification of gender and estimation of age, we use a multilayered neural network. We also examine the left– right symmetry of faces in connection with gender and age estimation by the proposed method. © 2008 Wiley Periodicals, Inc. Electron Comm Jpn, 91(7): 32– 40, 2008; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/ecj.10125

On the calculation of optical performance factors from vertebrate spatial contrast sensitivity

A novel technique for calculating the visual optical modulation transfer function (OMTF) is described. The technique involves application of the Rovamo–Barten model of spatial vision to measured contrast sensitivity data. [For details of the basic model see; Rovamo, J., Mustonen, J., & Nasanen, R. (1994). Modelling contrast sensitivity as a function of retinal illuminance and grating area. Vision Research, 34, 1301–1314 and Barten, P. J. G. (1999). Contrast sensitivity of the human eye and its effects on image quality. Washington: SPIE Optical Engineering Press.] In order to obtain OMTF, the model was simplified for use in the high spatial frequency range and also modified to include a transfer function term relating to attenuation by the retinal receptor sampling process. Calculations of OMTF were initially obtained from published contrast sensitivity for the human, cat, rat and chicken. The results were found to correlate well with OMTF values directly obtained through a double-pass optical measuring technique applied to all four species. It was assumed, following this initial test, that the modified Rovamo–Barten model could be used to extract OMTF from vertebrate contrast sensitivity data in general. Using published behavioural contrast sensitivity, further OMTF values were calculated from the model for the pigeon, goldfish, owl monkey, and tree shrew. The results obtained were used to provide a direct inter-species comparison of optical performance for a matched stimulus luminance. This study also confirms that, in many cases, vertebrate optical and receptor sampling processes are well matched in their attenuation properties.

Investigating a space-variant weighted salience account of visual selection

Weighted salience models are a popular framework for image-driven visual attentional processes. These models operate by: sampling the visual environment; calculating feature maps; combining them in a weighted sum and using this to determine where the eye will fixate next. We examine these stages in turn. We find that a biologically plausible non-uniform retinal sampling causes feature coding unreliability. The linear weighted sum operation seems an adequate model if statistical feature dimension dependencies are considered. Using signal detection theory we find good discrimination between targets and non-targets in the weighted sum, but the fixation criterion of ‘peak salience’ is suboptimal.

Effect of optical correction and remaining aberrations on peripheral resolution acuity in the human eye

Retinal sampling poses a fundamental limit to resolution acuity in the periphery. However, reduced image quality from optical aberrations may also influence peripheral resolution. In this study, we investigate the impact of different degrees of optical correction on acuity in the periphery. We used an adaptive optics system to measure and modify the off-axis aberrations of the right eye of six normal subjects at 20 degrees eccentricity. The system consists of a Hartmann-Shack sensor, a deformable mirror, and a channel for visual testing. Four different optical corrections were tested, ranging from foveal sphero-cylindrical correction to full correction of eccentric low- and high-order monochromatic aberrations. High-contrast visual acuity was measured in green light using a forced choice procedure with Landolt C's, viewed via the deformable mirror through a 4.8-mm artificial pupil. The Zernike terms mainly induced by eccentricity were defocus and with- and against-the-rule astigmatism and each correction condition was successfully implemented. On average, resolution decimal visual acuity improved from 0.057 to 0.061 as the total root-mean-square wavefront error changed from 1.01 mum to 0.05 mum. However, this small tendency of improvement in visual acuity with correction was not significant. The results suggest that for our experimental conditions and subjects, the resolution acuity in the periphery cannot be improved with optical correction.

Effect of myopia on visual acuity measured with laser interference fringes

The aim was to determine how visual acuity is affected by myopia when optical factors of the eye are controlled. Grating acuity was measured with interference fringes to avoid the effects of aberrations, and ocular biometry was used to compensate for differences in retinal image size among subjects. Distance spectacle refractions ranged from +2.25 to −14.75 D. The retinal magnification factor (RMF) in mm/deg was computed for each eye from the distance refraction, central corneal power and ultrasound biometry. A forced-choice orientation discrimination method was used to measure acuity for high-contrast 543 nm laser interference fringes in three retinal locations: the fovea, and at 4 deg and 10 deg eccentricity in the temporal retina. Acuity, expressed in c/deg and adjusted for spectacle magnification, was not significantly correlated with refraction at any of the three retinal locations. When acuity was converted to retinal spatial frequency units (c/mm) via the RMF, acuity decreased with increasing myopia at all three retinal locations (significantly at the fovea and at 10 deg eccentricity). Retinal acuity values in highly myopic subjects (>6 D) are consistent with retinal sampling distances that are larger than published values of human cone or ganglion cell spacing. The results imply that a highly myopic eye has retinal neurons that are more widely spaced than normal, but the increased axial length enlarges the retinal image enough to compensate for the retinal stretching. The data are consistent with a retinal stretching model that primarily affects the posterior pole.

Glutamate excitotoxicity in glaucoma: throwing the baby out with the bathwater?

The recently published study by Kwon et al provides some interesting and valuable data concerning the possible role of glutamate in neuronal cell death in glaucoma.1 As pointed out in the accompanying editorial, this is a field of study that has been bedevilled by conflicting results and even scientific fraud.2 The editorial2 is right to draw attention to these matters and to state that ‘it is important to re-evaluate the evidence for and against glutamate being causal in the development of glaucoma.’ Additionally, the editorial goes on to imply that there is thus no role for glutamate and its receptors (particularly NMDA receptors) in glaucomatous neuronal death. However, it may be wise to urge caution in the interpretation of these new data, as the lack of vitreal glutamate elevation does not necessarily mean that there is no role for glutamate and its receptors in glaucoma. Indeed, Kwon et al​1 clearly discuss some of the relevant issues. Although an elevation of vitreal glutamate might be expected in glaucoma models, there is no a priori reason for assuming that this must be so, and indeed it may be that such an increase is not as central to the disease as implied in the editorial:2 there is a danger here of throwing the baby out with the bathwater. It is well known that glutamate is an amino acid that is abundant in all cells and that it is intimately involved in many metabolic processes, with only a relatively small proportion being involved in neurotransmission and the activation of the receptors involved in this.3 This compartmentalisation of glutamate, maintained by active transport processes, makes it difficult to measure changes in tissue levels of glutamate related to activation of glutamate receptors unless relatively noninvasive sampling techniques with good temporal resolution are applied close to the source of glutamate (eg microdialysis or push-pull perfusion). Thus, sampling of retinal tissue post-mortem or sampling of vitreal glutamate, as performed by Kwon et al,​1 might not yield information directly relevant to the activation of glutamate receptors on threatened neurones. A second issue is the nature of the NMDA receptor (the glutamate receptor that is thought to be involved in the neurodegenerative aspects of glutamate pathophysiology) itself. A feature of the NMDA receptor is that its ion channel is largely blocked by magnesium ions when cell membranes are maintained at a healthy resting membrane potential.4 If tissue is compromised (eg injury or anoxia), then the membrane potential will depolarise and this will then relieve the magnesium ion block of the NMDA receptors: the consequence of this is that even relatively ‘normal’ levels of glutamate will have a much greater effect and there will be a substantial inward current into cells carried by sodium and calcium ions. Thus, large increases in glutamate levels may not be necessary in order to have sustained enhanced activation of NMDA receptors leading to the start of neurodegenerative processes. Therefore, looking for increases in vitreal or retinal glutamate levels in glaucoma or glaucoma models may not be the correct question. What will be more telling and significant is to see whether antagonists of NMDA receptors are effective treatments in such conditions. Experimental work shows some promise in this regard,5 and the forthcoming clinical trials data with memantine will be of great significance here.

Wavefront-customized intraocular lenses.

Recent advances in ocular wavefront measurement and in intraocular lens materials and manufacturing methods have brought cataract surgery to the brink of a period in which customized correction of higher-order aberrations with intraocular lenses may become standard practice. Retinal image quality in pseudophakic eyes is limited by the wavefront aberrations of the cornea and the intraocular lens. The Tecnis Z9000 is the first commercially available intraocular lens designed to account for the wavefront aberrations of the cornea, specifically spherical aberration. Clinical findings with the Tecnis Z9000 intraocular lens show improved contrast sensitivity at low and mid spatial frequencies. However, if the lens decenters or tilts modestly, higher-order aberrations are created, and the lens may underperform relative to standard intraocular lenses. At present, one firm is developing an intraocular lens that may be modified in vivo with near ultraviolet energy. Such a technology offers tremendous potential for a fully customized intraocular lens. Wavefront-customized intraocular lenses offer the promise of near perfect retinal image quality, such that only diffraction, chromatic aberration, retinal sampling and neural factors will limit vision in pseudophakic eyes.

Is neural filling–in necessary to explain the perceptual completion of motion and depth information?

Retinal activity is the first stage of visual perception. Retinal sampling is non-uniform and not continuous, yet visual experience is not characterized by holes and discontinuities in the world. How does the brain achieve this perceptual completion? Fifty years ago, it was suggested that visual perception involves a two-stage process of (i) edge detection followed by (ii) neural filling-in of surface properties. We examine whether this general hypothesis can account for the specific example of perceptual completion of a small target surrounded by dynamic dots (an 'artificial scotoma'), a phenomenon argued to provide insight into the mechanisms responsible for perception. We degrade the target's borders using first blur and then depth continuity, and find that border degradation does not influence time to target disappearance. This indicates that important information for the continuity of target perception is conveyed at a coarse spatial scale. We suggest that target disappearance could result from adaptation that is not specific to borders, and question the need to hypothesize an active filling-in process to explain this phenomenon.

The Linkage of Visual Motion Signals

Recovering a reliable 3D percept from the retinal sampling of dynamic images requires the linkage of motion signals across space and time. In this paper, we review recent experimental results that enhance our understanding of the perceptual processes of motion integration, segmentation, and selection that are necessary to solve this inverse optics problem. Simple paradigms involving the presentation of moving contours are used to assess our ability to link sparse motion information. Our results indicate that human motion perception strongly depends on both primitive stimulus characteristics, such as contrast, eccentricity, and duration, as well as higher level characteristics such as feature classification and spatial configurations. Further, the perceived direction of a moving object depends little upon its familiarity. Finally, pursuit eye movements of compositional stimuli are highly correlated with perceived motion coherence. This ensemble of results is analysed within the context of current theories of motion perception.

Fast image classification using a sequence of visual fixations

Based on human retinal sampling distributions and eye movements, a sequential resolution image preprocessor is developed. Combined with a nearest neighbor classifier, this preprocessor provides an efficient image classification method, the sequential resolution nearest neighbor (SRNN) classifier. The human eye has a typical fixation sequence that exploits the nonuniform sampling distribution of its retina. If the retinal resolution is not sufficient to identify an object, the eye moves in such a way that the projection of the object falls onto a retinal region with a higher sampling density. Similarly, the SRNN classifier uses a sequence of increasing resolutions until a final class decision is made. Experimental results on texture segmentation show that the preprocessor used in the SRNN classifier is considerably faster than traditional multiresolution algorithms which use all the available resolution levels to analyze the input data.

Retinally reconstructed images: digital images having a resolution match with the human eye

Current digital image/video storage, transmission and display technologies use uniformly sampled images. On the other hand, the human retina has a nonuniform sampling density that decreases dramatically as the solid angle from the visual fixation axis increases. Therefore, there is sampling mismatch. This paper introduces retinally reconstructed images (RRI), a representation of digital images that enables a resolution match with the retina. To create an RRI, the size of the input image, the viewing distance and the fixation point should be known. In the coding phase, we compute the codes, which consist of the retinal sampling locations onto which the image projects, together with the retinal outputs at these locations. In the decoding phase, we use the backprojection of the retinal onto the input image grid as B-spline control coefficients, in order to construct a 3D B-spline surface with nonuniform resolution properties. An RRI is then created by mapping the B-spline surface onto a uniform grid, using triangulation. Transmitting or storing the codes instead of the full resolution images enables up to two orders of magnitude data compression, depending on the resolution of the input image, the size of the input image and the viewing distance. The data reduction capability of retinal and RRI is promising for digital video storage and transmission applications. However, the computational burden can be substantial in the decoding phase.

On image quality of microlens arrays in diurnal superposition eyes

High-resolution compound eyes of the superposition type have been found in some diurnal insects, suggesting high-quality optics. Here we study whether a superposition eye could potentially achieve a higher resolution than that of a single, diffraction-limited facet, like in modern optical instruments made of a number of facets. We have developed a schematic diffractional model of the particularly acute superposition eye of the diurnal lepidopteran agaristid moth Phalaenoides Tristifica, and estimated the amount of aberrations required to fit the experimental data. Our results show that if the ommatidia were perfect (aberration-free) and perfectly aligned towards a common centre O, the resolution of the superposition image would be about 10 times higher than that provided by a single facet, and this in spite of the large phase lags between the pencils of light stemming from neighbouring ommatidia, due to the spherical geometry of the eye. In the insect species taken as a model system here, however, retinal sampling is too coarse to take full advantage of the high spatial frequency content of the superposition image. We show that the experimental data can be fitted if aberrations such as misalignment are introduced. A corollary of these results is that a superposition compound eye which was better constructed and avoided misalignment would provide its host with a better resolution than its `apposition eye counterpart'. This might explain why some diurnal compound eyes such as that of the fast-flying lepidopteran Macroglossum can afford sampling the focal image with a number of rhabdom exceeding the number of ommatidia (Warrant et al 1998 J. Exp. Biol. at press).

Recognition of band-pass filtered hand-written numerals in foveal and peripheral vision.

The purpose of the present study was to find out what differences between foveal and peripheral pattern recognition remain unexplained by the inhomogeneities of retinal sampling and the optics of the eye. We measured contrast thresholds for pattern recognition at different eccentricities. The effects of retinal sampling were homogenised by using M-scaling of the stimuli, and the effects of the optics of the eye were by-passed either by using strong external noise (signal-to-noise ratio is not affected by optical attenuation) or by computing retinal image contrast by means of the optical modulation transfer function. The stimuli were hand-written numerals filtered to two-octave bands of various centre object spatial frequencies (c/object). The results were described as contrast thresholds and recognition efficiency. At all eccentricities, lowest contrast thresholds and highest recognition efficiencies were found at medium object spatial frequencies. At high object spatial frequencies the peripheral retinal contrast thresholds and recognition efficiencies were nearly as good as at the fovea, but at low object spatial frequencies most of the data showed superiority of the fovea to the periphery. Therefore, at high object spatial frequencies peripheral recognition performance could be explained relatively well by the retinal sampling gradient, or equivalently by the cortical magnification factor, together with the effects of the optics of the eye. Some eccentricity dependent deterioration of recognition at low object spatial frequencies remained unexplained.

Detection of faces and facial landmarks using iconic filter banks

We describe a novel and general approach for the detection of objects, specifically faces and facial landmarks. The approach is based on biologically motivated image representation and classification schemes. Early processing related to feature extraction involves retinal sampling along non-linear lattices and micro saccades, while classification is done using retinal filter banks implemented via self-organizing feature maps (SOFM). The optimal set of face, eye pair, eye, nose, and mouth feature models, respectively, is found by an enhanced SOFM approach using cross-validation and corrective training. Experimental results on a data set of over 400 images prove the feasibility of our approach.

Off-axis optical quality and retinal sampling in the human eye

Using the double pass procedure, Navarro et al. (1993; Journal of the Optical Society of America A, 10, 201–212) measured the monochromatic modulation transfer function (MTF) of the human eye as a function of retinal eccentricity. They chose conditions as similar as possible to those encountered in natural viewing. We report new measurements obtained with conditions chosen instead to optimize retinal image quality: we paralyzed accommodation, used a 3 mm pupil, and corrected defocus and oblique astigmatism at each retinal location. MTFs were estimated at the tangential focus, circle of least confusion, and sagittal focus produced by oblique astigmatism. Though optical blur is well-known to have little effect on peripheral visual acuity, it can nonetheless substantially reduce aliasing by receptoral and post-receptoral spatial sampling.

A neural theory of retino-cortical dynamics

As a result of a variety of factors—the movements of the eyes, those of external objects, integration time, vergence, accommodation, and nonuniform retinal sampling—the retinal encoding is highly transient, blurred, and distorted. Yet this problem received very little attention, for most of the models proposed in the literature are built around the analysis of static (or steady-state) and uniformly focused images. Thus, a fundamental problem in visual perception consists of the understanding of the processes underlying the synthesis of phenomenally static, sharp percepts from transient, blurred activities. We present a continuous-time neural theory that proposes two major roles for the retinal transient activity: First, we propose that the nonmonotonic behavior of retinal neurons serves as a simple form of memory that adaptively filters visual signals to guide attentional mechanisms. Second, we propose that the transient activity is essential in achieving sharp dynamic percepts while preserving a good sensitivity to light. Theoretical analysis shows that an extraretinal on-center off-surround feedback anatomy is required to sharpen the “blurred output” from the retinal level. Mathematical properties of such feedback loops indicate that a transient reset mechanism is necessary to avoid smearing. It is proposed that transient retinal cells realize the reset by sending inhibitory signals to sustained activity distributions at higher levels (extraretinal areas: e.g., lateral geniculate nucleus (LGN) and/or visual cortical areas). In this theoretical framework, the continuous time behavior of the visual system can be analyzed in three major phases. In the first phase, sustained retinal signals are sharpened by feedback dominant extra-retinal loops. When the input moves, a second phase is engaged. In this phase, transient retinal cells reset rapidly and briefly extra-retinal activities. In the third phase, extra-retinal loops enter a feedforward mode thereby transferring a faithful copy of retinal activity into their own cells. The feedforward mode is maintained by the transient components of the retinal sustained units. When retinal units enter their steady-state mode, the overall system returns to the first phase where sharpening occurs through feedback dominant extra-retinal loops. The predictions of the theory are compared with various experimental data with emphasis on masking and motion deblurring phenomena.

Direct observation of receptors and images in simple and compound eyes

The relation between the quality of the optical image and the fineness of the retinal mosaic has been studied in eyes of three different optical types: the simple eyes of spiders, the superposition compound eyes of moths and the apposition compound eyes of butterflies. In all three it is possible to observe both the receptor mosaic and the image in the living eye, using appropriate ophthalmoscopic techniques. Whereas in humans the retinal sampling frequency approaches the optical cut-off frequency quite closely, in diurnal insects of both types the image is undersampled by a factor of 2–3, and in crepuscular spiders this factor may be greater than 100. Reasons for these differences are discussed.