Subthreshold facilitation and suppression in primary visual cortex revealed by intrinsic signal imaging.

Neurons in primary visual cortex (area 17) respond vigorously to oriented stimuli within their receptive fields; however, stimuli presented outside the suprathreshold receptive field can also influence their responses. Here we describe a fundamental feature of the spatial interaction between suprathreshold center and subthreshold surround. By optical imaging of intrinsic signals in area 17 in response to a stimulus border, we show that a given stimulus generates activity primarily in iso-orientation domains, which extend for several millimeters across the cortical surface in a manner consistent with the architecture of long-range horizontal connections in area 17. By mapping the receptive fields of single neurons and imaging responses from the same cortex to stimuli that include or exclude the aggregate suprathreshold receptive field, we show that intrinsic signals strongly reveal the subthreshold surround contribution. Optical imaging and single-unit recording both demonstrate that the relative contrast of center and surround stimuli regulates whether surround interactions are facilitative or suppressive: the same surround stimulus facilitates responses when center contrast is low, but suppresses responses when center contrast is high. Such spatial interactions in area 17 are ideally suited to contribute to phenomena commonly regarded as part of "higher-level" visual processing, such as perceptual "popout" and "filling-in."

Le développement visuel et cognitif chez les enfants nés à terme ou prématurément

Au cours des 25 dernières années, les recherches sur le développement visuel chez l’humain à l’aide de l’électrophysiologie cérébrale et des potentiels évoqués visuels (PEV) ont permis d’explorer plusieurs fonctions associées au cortex visuel. Néanmoins, le développement de certaines d’entre elles (p. ex. segmentation des textures), tout comme les effets de la prématurité sur celles-ci, sont des aspects qui nécessitent d’être davantage étudiés. Par ailleurs, compte tenu de l’importance de la vision dans le développement de certaines fonctions cognitives (p. ex. lecture, visuomotricité), de plus en plus de recherches s’intéressent aux relations entre la vision et la cognition. Les objectifs généraux de la présente thèse étaient d’étudier le développement visuel chez les enfants nés à terme et nés prématurément à l’aide de l’électrophysiologie, puis de documenter les impacts de la prématurité sur le développement visuel et cognitif. Deux études ont été réalisées. La première visait à examiner, chez des enfants nés prématurément, le développement des voies visuelles primaires durant la première année de vie et en début de scolarisation, ainsi qu’à documenter leur profil cognitif et comportemental. À l’aide d’un devis semi-longitudinal, dix enfants nés prématurément ont été évalués à l’âge de six mois (âge corrigé) et à 7-8 ans en utilisant des PEV, et des épreuves cognitives et comportementales à l’âge scolaire. Leurs résultats ont été comparés à ceux de 10 enfants nés à terme appariés pour l’âge. À six mois, aucune différence de latence ou d’amplitude des ondes N1 et P1 n’a été trouvée entre les groupes. À l’âge scolaire, les enfants nés prématurément montraient, comparativement aux enfants nés à terme, une plus grande amplitude de N1 dans la condition P-préférentielle et dans celle co-stimulant les voies M et P, et de P1 (tendance) dans la condition M-préférentielle. Aucune différence n’a été trouvée entre les groupes aux mesures cognitives et comportementales. Ces résultats suggèrent qu’une naissance prématurée exerce un impact sur le développement des voies visuelles centrales. L’objectif de la seconde étude était de documenter le développement des processus de segmentation visuelle des textures durant la petite enfance chez des enfants nés à terme et nés prématurément à l’aide des PEV et d’un devis transversal. Quarante-cinq enfants nés à terme et 43 enfants nés prématurément ont été évalués à 12, 24 ou 36 mois (âge corrigé pour les prématurés à 12 et 24 mois). Les résultats indiquaient une diminution significative de la latence de la composante N2 entre 12 et 36 mois en réponse à l’orientation, à la texture et à la segmentation des textures, ainsi qu’une diminution significative d’amplitude pour l’orientation entre 12 et 24 mois, et pour la texture entre 12 et 24 mois, et 12 et 36 mois. Les comparaisons entre les enfants nés à terme et ceux nés prématurément démontraient une amplitude de N2 réduite chez ces derniers à 12 mois pour l’orientation et la texture. Bien que ces différences ne fussent plus apparentes à 24 mois, nos résultats semblent refléter un délai de maturation des processus visuel de bas et de plus haut niveau chez les enfants nés prématurément, du moins, pendant la petite enfance. En conclusion, nos résultats indiquent que la prématurité, même sans atteinte neurologique importante, altère le développement des fonctions visuelles à certaines périodes du développement et mettent en évidence l’importance d’en investiguer davantage les impacts (p. ex. cognitifs, comportementaux, scolaires) à moyen et long-terme.

Confounds in pictorial sets: The role of complexity and familiarity in basic-level picture processing

Complexity is conventionally defined as the level of detail or intricacy contained within a picture. The study of complexity has received relatively little attention-in part, because of the absence of an acceptable metric. Traditionally, normative ratings of complexity have been based on human judgments. However, this study demonstrates that published norms for visual complexity are biased. Familiarity and learning influence the subjective complexity scores for nonsense shapes, with a significant training x familiarity interaction [F(1,52) = 17.53, p <.05]. Several image-processing techniques were explored as alternative measures of picture and image complexity. A perimeter detection measure correlates strongly with human judgments of the complexity of line drawings of real-world objects and nonsense shapes and captures some of the processes important in judgments of subjective complexity, while removing the bias due to familiarity effects.

Local object gist: meaningful shapes and spatial layout at a very early stage of visual processing

In his introduction, Pinna (2010) quoted one of Wertheimer’s observations: “I stand at the window and see a house, trees, sky. Theoretically I might say there were 327 brightnesses and nuances of color. Do I have ‘327’? No. I have sky, house, and trees.” This seems quite remarkable, for Max Wertheimer, together with Kurt Koffka and Wolfgang Koehler, was a pioneer of Gestalt Theory: perceptual organisation was tackled considering grouping rules of line and edge elements in relation to figure-ground segregation, i.e., a meaningful object (the figure) as perceived against a complex background (the ground). At the lowest level – line and edge elements – Wertheimer (1923) himself formulated grouping principles on the basis of proximity, good continuation, convexity, symmetry and, often forgotten, past experience of the observer. Rubin (1921) formulated rules for figure-ground segregation using surroundedness, size and orientation, but also convexity and symmetry. Almost a century of research into Gestalt later, Pinna and Reeves (2006) introduced the notion of figurality, meant to represent the integrated set of properties of visual objects, from the principles of grouping and figure-ground to the colour and volume of objects with shading. Pinna, in 2010, went one important step further and studied perceptual meaning, i.e., the interpretation of complex figures on the basis of past experience of the observer. Re-establishing a link to Wertheimer’s rule about past experience, he formulated five propositions, three definitions and seven properties on the basis of observations made on graphically manipulated patterns. For example, he introduced the illusion of meaning by comics-like elements suggesting wind, therefore inducing a learned interpretation. His last figure shows a regular array of squares but with irregular positions on the right side. This pile of (ir)regular squares can be interpreted as the result of an earthquake which destroyed part of an apartment block. This is much more intuitive, direct and economic than describing the complexity of the array of squares.

Semantic-level Visual Content Descriptor combining visual and textual cues as collaterals

In this paper, we introduce a novel high-level visual content descriptor which is devised for performing semantic-based image classification and retrieval. The work can be treated as an attempt to bridge the so called “semantic gap”. The proposed image feature vector model is fundamentally underpinned by the image labelling framework, called Collaterally Confirmed Labelling (CCL), which incorporates the collateral knowledge extracted from the collateral texts of the images with the state-of-the-art low-level image processing and visual feature extraction techniques for automatically assigning linguistic keywords to image regions. Two different high-level image feature vector models are developed based on the CCL labelling of results for the purposes of image data clustering and retrieval respectively. A subset of the Corel image collection has been used for evaluating our proposed method. The experimental results to-date already indicates that our proposed semantic-based visual content descriptors outperform both traditional visual and textual image feature models.

Motion detection and motion verbs: language affects low-level visual perception

Recent theories propose that semantic representation and sensorimotor processing have a common substrate via simulation. We tested the prediction that comprehension interacts with perception, using a standard psychophysics methodology.While passively listening to verbs that referred to upward or downward motion, and to control verbs that did not refer to motion, 20 subjects performed a motion-detection task, indicating whether or not they saw motion in visual stimuli containing threshold levels of coherent vertical motion. A signal detection analysis revealed that when verbs were directionally incongruent with the motion signal, perceptual sensitivity was impaired. Word comprehension also affected decision criteria and reaction times, but in different ways. The results are discussed with reference to existing explanations of embodied processing and the potential of psychophysical methods for assessing interactions between language and perception.

An fMRI-study of locally oriented perception in autism: altered early visual processing of the block design test

Autism has been associated with enhanced local processing on visual tasks. Originally, this was based on findings that individuals with autism exhibited peak performance on the block design test (BDT) from the Wechsler Intelligence Scales. In autism, the neurofunctional correlates of local bias on this test have not yet been established, although there is evidence of alterations in the early visual cortex. Functional MRI was used to analyze hemodynamic responses in the striate and extrastriate visual cortex during BDT performance and a color counting control task in subjects with autism compared to healthy controls. In autism, BDT processing was accompanied by low blood oxygenation level-dependent signal changes in the right ventral quadrant of V2. Findings indicate that, in autism, locally oriented processing of the BDT is associated with altered responses of angle and grating-selective neurons, that contribute to shape representation, figure-ground, and gestalt organization. The findings favor a low-level explanation of BDT performance in autism.

The interphotoreceptor retinoid binding protein gene in therian mammals: Implications for higher level relationships and evidence for loss of function in the marsupial mole

The subclass Theria of Mammalia includes marsupials (infraclass Metatheria) and placentals (infraclass Eutheria). Within each group, interordinal relationships remain unclear. One limitation of many studies is incomplete ordinal representation. Here, we analyze DNA sequences for part of exon 1 of the interphotoreceptor retinoid binding protein gene, including 10 that are newly reported, for representatives of all therian orders. Among placentals, the most robust clades are Cetartiodactyla, Paenungulata, and an expanded African clade that includes paenungulates, tubulidentates, and macroscelideans. Anagalida, Archonta, Altungulata, Hyracoidea + Perissodactyla, Ungulata, and the “flying primate” hypothesis are rejected by statistical tests. Among marsupials, the most robust clade includes all orders except Didelphimorphia. The phylogenetic placement of the monito del monte and the marsupial mole remains unclear. However, the marsupial mole sequence contains three frameshift indels and numerous stop codons in all three reading frames. Given that the interphotoreceptor retinoid binding protein gene is a single-copy gene that functions in the visual cycle and that the marsupial mole is blind with degenerate eyes, this finding suggests that phenotypic degeneration of the eyes is accompanied by parallel changes at the molecular level as a result of relaxed selective constraints.

A magnetoencephalographic study of low-level auditory processing in developmental dyslexia

The possibility that developmental dyslexia results from low-level sensory processing deficits has received renewed interest in recent years. Opponents of such sensory-based explanations argue that dyslexia arises primarily from phonological impairments. However, many behavioural correlates of dyslexia cannot be explained sufficiently by cognitive-level accounts and there is anatomical, psychometric and physiological evidence of sensory deficits in the dyslexic population. This thesis aims to determine whether the low-level (pre-attentive) processing of simple auditory stimuli is disrupted in compensated adult dyslexics. Using psychometric and neurophysiological measures, the nature of auditory processing abnormalities is investigated. Group comparisons are supported by analysis of individual data in order to address the issue of heterogeneity in dyslexia. The participant pool consisted of seven compensated dyslexic adults and seven age and IQ matched controls. The dyslexic group were impaired, relative to the control group, on measures of literacy, phonological awareness, working memory and processing speed. Magnetoencephalographic recordings were conducted during processing of simple, non-speech, auditory stimuli. Results confirm that low-level auditory processing deficits are present in compensated dyslexic adults. The amplitude of N1m responses to tone pair stimuli were reduced in the dyslexic group. However, there was no evidence that manipulating either the silent interval or the frequency separation between the tones had a greater detrimental effect on dyslexic participants specifically. Abnormal MMNm responses were recorded in response to frequency deviant stimuli in the dyslexic group. In addition, complete stimulus omissions, which evoked MMNm responses in all control participants, failed to elicit significant MMNm responses in all but one of the dyslexic individuals. The data indicate both a deficit of frequency resolution at a local level of auditory processing and a higher-level deficit relating to the grouping of auditory stimuli, relevant for auditory scene analysis. Implications and directions for future research are outlined.

Learning higher-level features with convolutional restricted Boltzmann machines for sentiment analysis

In recent years, learning word vector representations has attracted much interest in Natural Language Processing. Word representations or embeddings learned using unsupervised methods help addressing the problem of traditional bag-of-word approaches which fail to capture contextual semantics. In this paper we go beyond the vector representations at the word level and propose a novel framework that learns higher-level feature representations of n-grams, phrases and sentences using a deep neural network built from stacked Convolutional Restricted Boltzmann Machines (CRBMs). These representations have been shown to map syntactically and semantically related n-grams to closeby locations in the hidden feature space. We have experimented to additionally incorporate these higher-level features into supervised classifier training for two sentiment analysis tasks: subjectivity classification and sentiment classification. Our results have demonstrated the success of our proposed framework with 4% improvement in accuracy observed for subjectivity classification and improved the results achieved for sentiment classification over models trained without our higher level features.

Featureless visual processing for SLAM in changing outdoor environments

Vision-based SLAM is mostly a solved problem providing clear, sharp images can be obtained. However, in outdoor environments a number of factors such as rough terrain, high speeds and hardware limitations can result in these conditions not being met. High speed transit on rough terrain can lead to image blur and under/over exposure, problems that cannot easily be dealt with using low cost hardware. Furthermore, recently there has been a growth in interest in lifelong autonomy for robots, which brings with it the challenge in outdoor environments of dealing with a moving sun and lack of constant artificial lighting. In this paper, we present a lightweight approach to visual localization and visual odometry that addresses the challenges posed by perceptual change and low cost cameras. The approach combines low resolution imagery with the SLAM algorithm, RatSLAM. We test the system using a cheap consumer camera mounted on a small vehicle in a mixed urban and vegetated environment, at times ranging from dawn to dusk and in conditions ranging from sunny weather to rain. We first show that the system is able to provide reliable mapping and recall over the course of the day and incrementally incorporate new visual scenes from different times into an existing map. We then restrict the system to only learning visual scenes at one time of day, and show that the system is still able to localize and map at other times of day. The results demonstrate the viability of the approach in situations where image quality is poor and environmental or hardware factors preclude the use of visual features.

Learning Nonlinear Visual Processing from Natural Images

The paradigm of computational vision hypothesizes that any visual function -- such as the recognition of your grandparent -- can be replicated by computational processing of the visual input. What are these computations that the brain performs? What should or could they be? Working on the latter question, this dissertation takes the statistical approach, where the suitable computations are attempted to be learned from the natural visual data itself. In particular, we empirically study the computational processing that emerges from the statistical properties of the visual world and the constraints and objectives specified for the learning process. This thesis consists of an introduction and 7 peer-reviewed publications, where the purpose of the introduction is to illustrate the area of study to a reader who is not familiar with computational vision research. In the scope of the introduction, we will briefly overview the primary challenges to visual processing, as well as recall some of the current opinions on visual processing in the early visual systems of animals. Next, we describe the methodology we have used in our research, and discuss the presented results. We have included some additional remarks, speculations and conclusions to this discussion that were not featured in the original publications. We present the following results in the publications of this thesis. First, we empirically demonstrate that luminance and contrast are strongly dependent in natural images, contradicting previous theories suggesting that luminance and contrast were processed separately in natural systems due to their independence in the visual data. Second, we show that simple cell -like receptive fields of the primary visual cortex can be learned in the nonlinear contrast domain by maximization of independence. Further, we provide first-time reports of the emergence of conjunctive (corner-detecting) and subtractive (opponent orientation) processing due to nonlinear projection pursuit with simple objective functions related to sparseness and response energy optimization. Then, we show that attempting to extract independent components of nonlinear histogram statistics of a biologically plausible representation leads to projection directions that appear to differentiate between visual contexts. Such processing might be applicable for priming, \ie the selection and tuning of later visual processing. We continue by showing that a different kind of thresholded low-frequency priming can be learned and used to make object detection faster with little loss in accuracy. Finally, we show that in a computational object detection setting, nonlinearly gain-controlled visual features of medium complexity can be acquired sequentially as images are encountered and discarded. We present two online algorithms to perform this feature selection, and propose the idea that for artificial systems, some processing mechanisms could be selectable from the environment without optimizing the mechanisms themselves. In summary, this thesis explores learning visual processing on several levels. The learning can be understood as interplay of input data, model structures, learning objectives, and estimation algorithms. The presented work adds to the growing body of evidence showing that statistical methods can be used to acquire intuitively meaningful visual processing mechanisms. The work also presents some predictions and ideas regarding biological visual processing.