No / Purpose: Does a physiologically plausible model of the retinal ganglion cell (RGC) receptive field (RF) predict the spatial tuning properties of the Hermann Grid Illusion (HGI)? Methods: The spatial tuning of a single intersection HGI was measured psychophysically in normal observers using a nulling technique at different vertical grid line luminances. We used a model based upon a standard RGC RF, balanced to produce zero response under uniform illumination, to predict the response of the model cell to the equivalent range of stimulus conditions when placed in either the 'street' or the 'intersection' of a single element of a Hermann grid. We determined the equivalent of the nulling luminance required to balance these responses and minimise the HGI. Results: The model and the psychophysical data demonstrated broad spatial tuning with similarly shaped tuning profiles and similar strengths of illusion. The line width at the peak of the model tuning function was around twice the model RGC RF centre size. Modelling the psychophysical functions gave RF centre sizes smaller than expected from human anatomical evidence but similar to that suggested by primate physiological evidence. In the model and psychophysically the strength of the illusion varied with the luminance of the vertical grid line when HGI strength was expressed as a Michelson nulling contrast, but this effect was smaller when HGI strength was expressed as a nulling luminance. Conclusions: The shape, width, height and position of the spatial tuning function of the HGI can be well modelled by a RGC RF based model. The broad tuning of these functions does not appear to require a broad range of different cell sizes either in the retina or later in the visual pathway.

Yes / Purpose.: To establish whether a clinically exploitable relationship exists between surrogate measures of retinal ganglion cell number and functional sampling density and visual contrast sensitivity in healthy young eyes. Methods.: Psychometric functions for contrast detection were measured at 9° eccentricity in superior and inferior visual field from 20 healthy adults (age 23–43, median 26 years). Functions were compared with corresponding localized regions of retinal nerve fiber layer (RNFL) thickness measured by optical coherence tomography, a surrogate of retinal ganglion cell number, and to grating resolution acuity, a psychophysical surrogate of retinal ganglion cell sampling density. Correlations between psychometric function parameters and retinal ganglion cell surrogates were measured by Spearman's rank correlation. Results.: All measures exhibited a 2- to 4-fold variation in our sample. Despite this, correlations between measures were weak. Correlations between psychometric function parameters (threshold, spread) and RNFL thickness ranged in magnitude from 0.05 to 0.19 (P = 0.43–0.85). Grating resolution was sampling limited for 16 of 20 participants in superior visual field, and for 12 of 20 participants in inferior visual field. Correlations between psychometric function parameters and grating resolution acuities ranged in magnitude from 0.05 to 0.36 (P = 0.12–0.85) when all data were considered, and from 0.06 to 0.36 (P = 0.26–0.87) when only sampling-limited data were considered. Conclusions.: Despite considerable variation in both psychometric functions for contrast detection and surrogate measures of retinal ganglion cell number and sampling density among healthy eyes, relationships between these measures are weak. These relationships are unlikely to be exploitable for improving clinical tests in healthy populations.

In the developing visual system, the axon of a pre-synaptic cell must be guided to a post-synaptic partner. Retinal ganglion cells (RGCs) in the eye are an excellent model to study this process. Multiple classes exist that respond to specific types of light input, and these project to different destinations in the brain that process distinct types of information. The RGC axons that navigate to the lateral geniculate nucleus (LGN) do so in a class-specific manner. Axons from RGCs that mediate non-image forming functions innervate the ventral LGN (vLGN) and the intergeniculate leaflet (IGL). Axons from RGCs that process image-forming information bypass these regions to innervate the dorsal LGN (dLGN). The extracellular protein reelin was identified as a potential factor in RGC axonal targeting of the vLGN and IGL, and the reeler mutant mouse used to study the effects of its functional absence. Anterograde labeling of RGCs and their axons with Cholera toxin B (CTB) revealed reduced patterns of retinal innervation to the vLGN and IGL in mutant mice. Moreover, the absence of functional reelin resulted in axons incorrectly growing into inappropriate regions of the thalamus. We identified these misrouted axons as those of the intrinsically photosensitive RGCs (ipRGCS), a class of RGCs known to project to the affected subnuclei. In contrast to defects in ipRGC targeting, no deficits were seen in retinogeniculate or corticothalamic projections in classes of axons that normally target the dLGN. Immunohistochemistry did not reveal any effects of the absence of the functional reelin on the LGN cytoarchitecture, which is unlike many other brain regions altered in the reeler. In summary, results suggest that intact reelin is required for class-specific retinogeniculate targeting to the vLGN and IGL. The defects are likely to be in targeting and not in neuronal positioning.

Teoricamente, conexões por sinapses elétricas entre neurônios poderiam levar ao aumento da faixa de resposta dinâmica da rede neural. A faixa de resposta dinâmica de uma rede de neurônios pode ser definida como a faixa de valores de intensidade dos estímulos de entrada para a qual o conjunto de neurônios produz resposta antes de atingir a saturação. Em um cenário biológico, propôs-se que junções gap entre células ganglionares da retina aumentariam a faixa dinâmica da retina. O teste experimental dessa proposta apresenta várias dificuldades, o que torna a modelagem computacional uma alternativa metodológica para o estudo do papel das sinapses elétricas na faixa dinâmica da camada ganglionar da retina. O objetivo deste trabalho foi a construção de um modelo biologicamente plausível da camada ganglionar da retina da salamandra, submetida a sinais de entrada realísticos conforme evidências experimentais e com a inclusão de sinapses elétricas conectando suas células, para estudar in silico os possíveis efeitos dessas sinapses elétricas sobre a faixa dinâmica da camada ganglionar. A camada ganglionar foi modelada como uma rede bidimensional cujos neurônios foram modelados pelo formalismo de Hodgkin-Huxley. Cada neurônio recebeu um de dois tipos de entrada sináptica, transiente ou sustentada. Avaliou-se o efeito da inibição pré-sináptica das células ganglionares e o efeito de diferentes padrões de conectividade mediados pelas sinapses elétricas. Os resultados sugerem que o acoplamento elétrico aumenta a sensibilidade do sistema e altera o ponto de saturação, mas não necessariamente aumenta a faixa dinâmica. / Theoretically, connections by electrical synapses between neurons could lead to an increase in their dynamic range. The dynamic range of a network of neurons can be defined as the range of input stimuli values for which the network responds before saturation. In a biological scenario, it is hypothesized that gap junctions between retinal ganglion cells may increase the dynamic range of the retina. However, the experimental testing of this hypothesis presents several difficulties, which makes computational modeling a methodological alternative to study the role of electrical synapses on the dynamic range of the ganglion cell layer of the retina. In this work we constructed a biologically plausible computational model of the ganglion cell layer of the salamander retina. A bidimensional network was built with cells modeled by the Hodgkin-Huxley formalism connected via gap junctions and subject to realistic inputs constrained by experimental evidence, to study in silico the effects of gap junctions on the dynamic range of the model. We studied the effect of different gap junction-mediated connectivity patterns, input type combinations (transient, sustained and mixed between the two) and presynaptic inhibition on the dynamic range. Our results suggest that gap junction coupling increases the network\'s sensitivity and alters the saturation point but not necessarily increases the dynamic range.

Experimental and clinical data have documented the adverse effects of perinatal ethanol intoxication on peripheral organs and the central nervous system. There is little known, however, about potential damaging effects of perinatal ethanol on the developing visual system. The purpose of this study was to examine the effects of neonatal ethanol intoxication on RGC morphology, estimate the total number of neurons in RGC layer and dorsolateral geniculate nucleus (dLGN), and on the eye-specific fiber segregation in the dLGN), in YFP and C57BL/6 mice pups. Ethanol (3 g/kg/day) was administered by intragastric intubation throughout postnatal days (PD) 3-20 or 3-10. Intubation control (IC) and untreated control (C) groups were included. Blood alcohol concentration (BAC) was measured in separate groups of pups on PD3, PD10, and PD20 at 4 different time points, 1, 1.5, 2 and 3 h after the second intubation. Numbers neurons in the RGCs and dLGN were quantified on PD10, PD20 using unbiased stereological procedures. The RGC images were taken using a confocal microscope and images were traced using Neurolucida software. On PD9, intraocular injections of cholera toxin-

[Truncated abstract] Microenvironmental changes associated with apoptotic neural degeneration may instruct a proportion of newly transplanted donor cells to differentiate towards the fate of the deteriorating host cellular phenotype. In the work described in this thesis, this hypothesis was tested by inducing apoptotic retinal ganglion cell (RGC) death in neonatal and adult rats and mice, and then examining whether intravitreally grafted cells from a range of sources of donor neural tissue became incorporated into these selectively depleted retinae. Donor tissues were: a postnatal murine cerebellar-derived immortalised neural precursor cell line (C17.2); an adult rat hippocampal-derived clonal stem-like line (HCN/GFP); mouse embryonic day 14 (E14) primary dissociated retinal cells (Gt[ROSA]26); and adult mouse ciliary pigmented margin-derived primary neurospheres (Gt[ROSA]26). In neonates, rapid RGC death was induced by removal of the contralateral superior colliculus (SC), and in adults, delayed RGC death was induced by unilateral optic nerve (ON) transection. Some adult hosts received ON transection coupled with an autologous peripheral nerve (PN) graft. Donor cells were injected intravitreally 6-48 h after SC ablation (neonates) or 0, 5, 7 or 14 days after ON injury (adults). Cells were also injected into non-RGC depleted neonatal and adult retinae. At 4 or 8 weeks, transplanted cells were identified, quantified and their differentiation fate within host retinae was assessed. Transplanted male C17.2 cells were identified in host retinae using a Y-chromosome marker and in situ hybridisation, or by their expression of the lacZ reporter gene product Escherichia coli beta-galactosidase (beta-gal) using Xgal histochemistry or a beta-gal antibody. No C17.2 cells were identified in axotomised adult-injected eyes undergoing delayed RGC apoptosis (n = 16). Donor cells were, however, stably integrated within the retina in 29% (15/55) of mice that received C17.2 cell injections 24 h after neonatal SC ablation; 6-31% of surviving cells were found in the RGC layer (GCL). These NSC-like cells were also present in intact retinae, but on average there were fewer cells in GCL. In SC-ablated mice, most grafted cells did not express retinal-specific markers, although occasional donor cells in the GCL were immunopositive for beta-III tubulin (TUJ1), a protein highly iii expressed by, but not specific to, developing RGCs. Targeted rapid RGC depletion thus increased C17.2 cell incorporation into the GCL, but grafted C17.2 cells did not appear to differentiate into an RGC phenotype.

Everyday an enormous amount of information is stored, processed and transmitted digitally around the world. Neurally-inspired compression models have been rapidly developed and researched as a solution to image processing tasks and channel error correction control. This dissertation presents a deep neural network (DNN) for gray high-resolution image compression and a fault-tolerant transmission system with channel error-correction capabilities. A feed-forward DNN implemented with the Levenberg-Marguardt learning algorithm is proposed and implemented for image compression. I demonstrate experimentally that the DNN not only provides better quality reconstructed images but also requires less computational capacity as compared to DCT Zonal coding, DCT Threshold coding, Set Partitioning in Hierarchical Trees (SPIHT) and Gaussian Pyramid. An artificial neural network (ANN) with improved channel error-correction rate is also proposed. The experimental results indicate that the implemented artificial neural network provides a superior error-correction ability by transmitting binary images over the noisy channel using Hamming and Repeat-Accumulate coding. Meanwhile, the network’s storage requirement is 64 times less than the Hamming coding and 62 times less than the Repeat-Accumulate coding. Thumbnail images contain higher frequencies and much less redundancy, which makes them more difficult to compress compared to high-resolution images. Bottleneck autoencoders have been actively researched as a solution to image compression tasks. However, I observed that thumbnail images compressed at a 2:1 ratio through bottleneck autoencoders often exhibit subjectively low visual quality. In this dissertation, I compared bottleneck autoencoders with two sparse coding approaches. Either 50\% of the pixels are randomly removed or every other pixel is removed, each achieving a 2:1 compression ratio. In the subsequent decompression step, a sparse inference algorithm is used to in-paint the missing the pixel values. Compared to bottleneck autoencoders, I observed that sparse coding with a random dropout mask yields decompressed images that are superior based on subjective human perception yet inferior according to pixel-wise metrics of reconstruction quality, such as PSNR and SSIM. With a regular checkerboard mask, decompressed images were superior as assessed by both subjective and pixel-wise measures. I hypothesized that alternative feature-based measures of reconstruction quality would better support my subjective observations. To test this hypothesis, I fed thumbnail images processed using either bottleneck autoencoder or sparse coding using either checkerboard or random masks into a Deep Convolutional Neural Network (DCNN) classifier. Consistent, with my subjective observations, I discovered that sparse coding with checkerboard and random masks support on average 2.7\% and 1.6\% higher classification accuracy and 18.06\% and 3.74\% lower feature perceptual loss compared to bottleneck autoencoders, implying that sparse coding preserves more feature-based information. The optic nerve transmits visual information to the brain as trains of discrete events, a low-power, low-bandwidth communication channel also exploited by silicon retina cameras. Extracting high-fidelity visual input from retinal event trains is thus a key challenge for both computational neuroscience and neuromorphic engineering. % Here, we investigate whether sparse coding can enable the reconstruction of high-fidelity images and video from retinal event trains. Our approach is analogous to compressive sensing, in which only a random subset of pixels are transmitted and the missing information is estimated via inference. We employed a variant of the Locally Competitive Algorithm to infer sparse representations from retinal event trains, using a dictionary of convolutional features optimized via stochastic gradient descent and trained in an unsupervised manner using a local Hebbian learning rule with momentum. Static images, drawn from the CIFAR10 dataset, were passed to the input layer of an anatomically realistic retinal model and encoded as arrays of output spike trains arising from separate layers of integrate-and-fire neurons representing ON and OFF retinal ganglion cells. The spikes from each model ganglion cell were summed over a 32 msec time window, yielding a noisy rate-coded image. Analogous to how the primary visual cortex is postulated to infer features from noisy spike trains in the optic nerve, we inferred a higher-fidelity sparse reconstruction from the noisy rate-coded image using a convolutional dictionary trained on the original CIFAR10 database. Using a similar approach, we analyzed the asynchronous event trains from a silicon retina camera produced by self-motion through a laboratory environment. By training a dictionary of convolutional spatiotemporal features for simultaneously reconstructing differences of video frames (recorded at 22HZ and 5.56Hz) as well as discrete events generated by the silicon retina (binned at 484Hz and 278Hz), we were able to estimate high frame rate video from a low-power, low-bandwidth silicon retina camera.

Teoricamente, conexões por sinapses elétricas entre neurônios poderiam levar ao aumento da faixa de resposta dinâmica da rede neural. A faixa de resposta dinâmica de uma rede de neurônios pode ser definida como a faixa de valores de intensidade dos estímulos de entrada para a qual o conjunto de neurônios produz resposta antes de atingir a saturação. Em um cenário biológico, propôs-se que junções gap entre células ganglionares da retina aumentariam a faixa dinâmica da retina. O teste experimental dessa proposta apresenta várias dificuldades, o que torna a modelagem computacional uma alternativa metodológica para o estudo do papel das sinapses elétricas na faixa dinâmica da camada ganglionar da retina. O objetivo deste trabalho foi a construção de um modelo biologicamente plausível da camada ganglionar da retina da salamandra, submetida a sinais de entrada realísticos conforme evidências experimentais e com a inclusão de sinapses elétricas conectando suas células, para estudar in silico os possíveis efeitos dessas sinapses elétricas sobre a faixa dinâmica da camada ganglionar. A camada ganglionar foi modelada como uma rede bidimensional cujos neurônios foram modelados pelo formalismo de Hodgkin-Huxley. Cada neurônio recebeu um de dois tipos de entrada sináptica, transiente ou sustentada. Avaliou-se o efeito da inibição pré-sináptica das células ganglionares e o efeito de diferentes padrões de conectividade mediados pelas sinapses elétricas. Os resultados sugerem que o acoplamento elétrico aumenta a sensibilidade do sistema e altera o ponto de saturação, mas não necessariamente aumenta a faixa dinâmica. / Theoretically, connections by electrical synapses between neurons could lead to an increase in their dynamic range. The dynamic range of a network of neurons can be defined as the range of input stimuli values for which the network responds before saturation. In a biological scenario, it is hypothesized that gap junctions between retinal ganglion cells may increase the dynamic range of the retina. However, the experimental testing of this hypothesis presents several difficulties, which makes computational modeling a methodological alternative to study the role of electrical synapses on the dynamic range of the ganglion cell layer of the retina. In this work we constructed a biologically plausible computational model of the ganglion cell layer of the salamander retina. A bidimensional network was built with cells modeled by the Hodgkin-Huxley formalism connected via gap junctions and subject to realistic inputs constrained by experimental evidence, to study in silico the effects of gap junctions on the dynamic range of the model. We studied the effect of different gap junction-mediated connectivity patterns, input type combinations (transient, sustained and mixed between the two) and presynaptic inhibition on the dynamic range. Our results suggest that gap junction coupling increases the network\'s sensitivity and alters the saturation point but not necessarily increases the dynamic range.

Light entering the eye is transformed by the retina into electrical signals. Extensive processing takes place in the retina before these signals are transmitted to the brain. Beginning in the outer retina, light-evoked electrical signals are distributed into parallel pathways specialized for different visual tasks, such as the detection of dark vs. bright ambient light, the onset or offset of light, and the direction of stimulus motion. Pathway diversity is a consequence of cell type diversity, differential cell connectivity, synapse organization, receptor expression, or any combination thereof. Cell connectivity itself can be accomplished through excitatory or inhibitory chemical synapses, or electrical coupling via gap junctions. Gap junctions are further specialized based on the expression of different connexin subunit isoforms. In aggregate, this diversity gives rise to ganglion cells with highly specialized functions, including ON and/or OFF responses, contrast-tuning and direction-selectivity (DS). The directionally-selective circuit, a circuit specialized for the encoding of stimulus motion, makes use of many of these circuit specializations. Bipolar cells, in response to glutamate release from cone photoreceptors, provide highly-sensitive glutamatergic input to amacrine cells and DS ganglion cells (DSGCs) in this circuit, while amacrine cells provide cholinergic and directionally-tuned GABAergic input to DSGCs. One population of DSGCs also transmit signals laterally to one another via gap junctions. Thus numerous specializations in bipolar cells, amacrine cells and ganglion cells endow DSGCs with their unique encoding abilities. In Chapters 2 and 3 of this dissertation I focus on synchronized firing between gap junction-coupled DSGCs. sDSGCs exhibit fine-scale correlations, with action potentials in an sDSGC more likely within ~2ms of action potential firing in a coupled neighbour. I first characterize electrical coupling of DSGCs through the identification of the molecular composition of DSGC gap junctions (Chapter 2). Physiological and immunohistochemical methods allowed me to demonstrate an important role for connexin 36 subunits in DSGC electrical coupling. Next (Chapter 3) I investigate the sub-cellular mechanisms underlying neuronal correlations between electrically coupled DSGCs. Using paired recordings, I show that chemical input (from bipolar cells and amacrine cells), electrical input (from gap junctions), and Na+ channel activity in DSGC dendrites underlie the generation of correlated spiking activity. While a common feature of electrically coupled networks, the mechanisms underlying correlations were previously unclear. In Chapter 4 I focus on the mechanisms within the DS circuit that endow these neurons with impressive sensitivity to stimulus contrast. Using physiological and pharmacological methods I first assess the relative contrast sensitivity of ganglion cells and starburst amacrine cells (SACs) in the DS circuit. The sensitivity of DSGC and SAC excitatory currents to antagonists of Na+ channels suggests an important role for these channels in amplifying low contrast responses and other weak inputs to the circuit. This role is later attributed to the differential expression of voltage-gated Na+ channels in specific bipolar cell populations. In aggregate, this dissertation describes several novel circuit mechanisms within the well-studied DS circuit. I also provide specific roles for such specializations in visual coding. / Graduate

Ce travail se situe dans le cadre d'un projet pluridisciplinaire visant à développer une prothèse visuelle. La technique la plus utilisée actuellement dans de nombreux types de neuroprothèses est basée sur l'excitation par voie électrique via des électrdes. Les inconvénients d'une telle technique (très invasive, de faible résolution spatiale et par contact) pourraient être surmontés en utilisant une stimulation par laser infra-rouge. Nous présentons dans un premier temps les caractéristiques des trois diodes lasers fibrés émettant à 1875 nm, 1535 nm et 1470 nm pour des gammes de puissances optiques de quelques centaines de mW qui ont été utilisés et intégrés à deux dispositifs de mesures permettant l'observations de variations d'échanges ioniques transmembranaires (imagerie de fluorescence des ions calciums et mesure électrophysiologique par la technique de patch clamp). Nous montrons ensuite que des réponses biologiques ont été obtenues par les trois lasers, non seulement sur des cellules ganglionnaires de la rétine et du vestibule de culture mais aussi sur des tranches de rétine. L'influence des paramètres clés comme la longueur d'onde, la durée de stimulation, les seuils d'énergie a été étudié, et a permis d'établir que les seuils d'énergie de stimulation dépendent de la valeur du coefficient d'absorption de l'eau qui varie suivant la longueur d'onde utilisée. Enfin, une étude est consacrée pour expliquer les mécanismes physiques et biologiques apparaissant au cours de l'interaction du laser avec le neurone au niveau cellulaire. Des simulations numériques quantifiant l'élévation de température associées à des tests pharmacologiques cherchant à déterminer la nature des canaux ioniques spécifiques mis en jeu suggèrent la prédominance d'un effet thermique. / This work is part of a pluridisciplinary project, aiming at developing a visual prosthesis. The most used technique for this kind of neuroprosthesis is based on the electrical stimulation of nerves by electrodes. Drawbacks of such a technique (very intrusive, low spatial resolution and physical contact) could be overcome by the use of an infra red laser based stimulation. We present first the three fibre pigtailed laser diode characteristics emitting few hundred of mW at 1875 nm, 1535 nm and 1470 nm. These lasers have been integrated on two measurement devices (a fluorescence microscope and a microscope using patch clamp recording), for the observation of ionic membrane exchanges. Our results show that action potentials have been obtained by laser stimulation from the three lasers, both on retinal or vestibular ganglion cells from mass cultures and on retinal slices. The effect of key parameters as the wavelength, the stimulation time, the energy thresholds has been studied and show that the energy thresholds clearly depend on the absorption coefficient of water which varies with the wavelength. Finally, we present the results of a preliminary study aiming at determining the biophysical interaction mechanisms at cell level. Numerical simulations giving the local increase of temperature and tests of specific blocking molecules in order to know the exact nature of the ionic channels involved suggest a predominant thermal mechanism.