視覚 Visual System
P2-2-100 面に基づく方位情報の統合により生起される曲率選択性 -V4におけるsparse codingの神経メカニズム Surface based integration of local orientations for the construction of curvature selectivity -Neural mechanism of sparse coding in V4- ○羽鳥康裕1,2, 益田達郎1, 酒井宏1 ○Yasuhiro Hatori1,2, Tatsuroh Mashita1, Ko Sakai1 筑波大院シス情コンピュータサイエンス1, 学振特別研究員2 Dept. of Comp. Sci., Grad. Sch. of SIE, Univ. of Tsukuba, Ibaraki, Japan1, JSPS Research Fellow2 Recent physiological studies have reported that V4 neurons code curvature as mid-level representation of object shape (Pasupathy & Connor, JNP, 2002; Yau et al., Cer. Cor., 2013), and that the representation is sparse (0.89; Carlson et al., Cur. Biol., 2011). Such neurons in V4 are often selective to the direction of curvature, indicating their joint-coding of surface or direction of figure. Our computational study (Hatori & Sakai, LNCS, 2012) has showed indeed that surface plays a crucial functional role in the integration of the responses of V1 neurons for the construction of curvature selectivity. The integration of orientation-selective responses of V1 cells in the direction of object surface reproduced crucial characteristics of V4 cells in vivo that are (1) single cells are selective to a particular curvature and its direction, and (2) population activity is biased toward acute curvature. However, such conditional integration that takes into account surface direction is not straightforward in neural wiring. In the present study, we investigated essential norm governing the integration. Since accumulative evidence suggests that sparseness is a neural principle for the integration of ascending information, we hypothesize that sparseness contributes to the control of the integration. To test the hypothesis, we applied component analysis (ICA) with sparseness constraint to activities of model V2 neurons responding to natural contours, and analyzed the basis functions obtained from ICA. During the learning, weights of sparseness were varied systematically. The appropriate basis functions were produced only when sparseness was medium (0.6-0.9): the bases were unstructured (<0.6) or did not reproduce the physiological characteristics (>0.9) for the other sparseness. These results suggest that sparseness is a crucial factor for the integration of ascending responses that controls the construction of curvature selectivity in V4.	P2-2-101 Myopia induced by visual deprivation in mice with optic nerve crush ○Xiao-Hua Wu1,2,3, Ping-Ping Zhang1,2,3, Yun-Yun Li1,2,3, Shi-Jun Weng1,2,3, Yong-Mei Zhong1,2,3, Xiong-Li Yang1,2,3 Institute of Neurobiology,Fudan University, Shanghai, China1, Institutes of Brain Science,Fudan University, Shanghai, China2, State Key Laboratory of Medical Neurobiology,Fudan University, Shanghai, China3 Myopia and axial elongation can be generated in several species by depriving the eyes of form vision. In this study, we investigated whether the eye-brain link is necessary for form deprivation induced myopia. Eye refraction was measured by an automated eccentric infrared photorefractor. When mice (C57 BL/6) were monocularly deprived of form vision with white translucent occluders for 4 weeks, a significant myopic shift (-5.55±1.83D), relative to the contralateral fellow eyes, was induced. The effects of optic nerve crush (ONC) on eye growth and on development of form deprivation myopia were further examined. Mice with ONC but without occluders became myopic or hyperopic 4 weeks after the surgery. In the myopic-shift group, the eyes with ONC showed significantly relative myopic shifts (-5.73±4.81D), associated with an increase in corneal thickness and a decrease in retinal thickness (P<0.05). In the hyperopic-shift group, the eyes with ONC turned out to be more hyperopic (4.59±3.03D) than the fellow eyes, with a significant decrease in both lens and retinal thickness (P<0.05). These results suggest that an intact optic nerve is indispensable for normal eye growth. Moreover, following ONC mice with occluders also developed a more severe myopia (-9.69±4.17D), as compared to those with occluders but without ONC (-5.55±1.83D). Meanwhile, significant increase in axial length and vitreous chamber (P<0.05), and decrease in retinal thickness (P<0.001), lens thickness and corneal radius of curvature (P<0.05) were also observed in form deprivation myopia with ONC. These findings strongly suggest the involvement of higher centers in the formation of myopia.
P2-2-103 Preferential neuronal responses to snakes in the monkey pulvinar ○Van Quan Le1, Etsuro Hori1, Minh Nui Nguyen1, Jumpei Matsumoto1, Taketoshi Ono1, Hisao Nishijo1 System Emotional Science, Graduate School of Medicine and Pharmaceutical Sciences, University of Toyama1 Snakes are nature enemies (predators) of monkeys, and snake bite is one of the major causes of morbidity and mortality of primates in many rural tropical areas. Furthermore, recent anthropological studies suggest that snakes areresponsible for evolutionary changes of the primate brain. On the other hand, neuropsychological and behavioral studies using humans and monkeys suggest that the subcortical visual pathway including the amygdala, pulvinar and superior colliculus is a part of the fear module circuitry. However, neuronal responses in the pulvinar to snakes have been unknown. In the present study, responses of neurons in the monkey pulvinar were recorded during a delayed non-matching-to-sample (DNMS) task in which monkeys were required to discriminate among 4 categories of visual stimuli [photos of snakes (snakes facing monkeys and attacking toward the sides), faces (angry and neutral human faces, angry and neutral monkey faces), hands (humanand monkey right and left prone or supine hands) and simple geometrical patterns (circle, cross, square, and star)]. Of 650 pulvinar neurons recorded, 97 (14.92%) responded to the visual stimuli. Of these, 76 neurons were tested with all of the stimuli, and responded differentially to these stimuli. Of the 76 neurons, 45 (59.21%), 35 (46.05%) 25 (32.89%) and 12(15.79%) responded stronger to snakes, faces, hands and simple patterns, respectively. These results suggest that the pulvinar might play an important role in detection of snakes.	P2-2-104 Past failures can bias human decisions ○Arman Abrahamyan1, Justin L Gardner1 Gardner Research Unit, RIKEN Brain Science Institute1 Past choice can bias present decision even if it may not be the optimal approach. For example, when deciding whether a visual stimulus was presented on the left or on the right, mice can exhibit a tendency to switch sides (left to right or vice versa) not based on the stimulus presentation side but based on success or failure on the previous trial (Busse et al., 2011, J Neurosci). Here we demonstrate that some people exhibit similar suboptimal strategies. Investigating these strategies is important both because it gives insight into how humans make sensory decisions and because these strategies bias psychophysical measurements and thus suggest that many reports of psychophysical sensitivity may have been widely underestimated.Similar to the task performed by mice, in our experiment humans reported if a weak or strong visual stimulus was presented on the left or right side of fixation. Each response was followed by auditory feedback to indicate correct and incorrect decisions. Using a probabilistic choice model (Busse et al., 2011, J Neurosci), we found that four out of seven subjects demonstrated a tendency to switch sides (i. e., from left to right or vice versa) when they failed on the previous trial. Successful responses on the previous trial had little or no effect on subjects' decisions. We conclude that even experienced subjects employed a suboptimal strategy in which past failures influence current choices. While classical psychophysics has developed methodologies that were thought to be immune to subjects' biases (Green & Swets, 1966), our results suggest that accurate measurements of psychophysical thresholds require new approaches to modeling subjects' decisions. These suboptimal strategies may show systematic differences between species. For mice, this suboptimal strategy of switching could reflect a more global strategy of systematic foraging for randomly located food, while humans may prefer to change answers when the previous answer was wrong.
P2-2-105 Humans exploit the uncertainty in priors to improve direction perception ○Steeve Laquitaine1, Justin Gardner1 Gardner Research Unit, Riken Brain Science Institute1 What prior representation do humans use when estimating the direction of motion? We used a motion direction estimation task in which three subjects were presented random dots moving in ten directions and were asked to report the motions directions. By changing independently the set of displayed directions and the coherence of the motion, we could tease apart the role of the uncertainty in subjects' priors and the level of sensory evidence in the estimation of the direction of motion.By examining the reported and the displayed directions we distinguished three hypotheses. 1) Priors may be ignored. Thus, the reported directions should depend on sensory evidence and match the displayed directions. 2) Subjects may use the mean but ignore the prior's uncertainty. Then, weak sensory evidence, at low coherence, should bias the reported directions toward the most likely direction, regardless of the prior's strength. 3) Subjects may use both the mean and the uncertainty of the prior, consistent with the Bayesian framework. Then the directions reported when sensory evidence is weak should be biased more toward the most likely direction as it becomes more certain. We found that weakening sensory evidence by lowering motion coherence biased direction estimation toward the most likely direction, the mean of the prior, ruling out hypothesis 1 (F(2,72)=304.90, p<0.001, ANCOVA). More importantly, the bias increased as the prior became more certain and when sensory evidence was weak, at lower coherences (F(1,22)=92.43, p<0.001 at 12 % and F(1,22)=6.78, p=0.02 at 35% coherence, ANCOVA), but the prior's strength had no effect at higher coherence (F(1,22)=2.13, p=0.16 at 100% coherence, ANCOVA) precluding hypothesis 2 and validating hypothesis 3. Altogether those results show that humans exploit the mean of the prior and its uncertainty to improve direction perception as prescribed by Bayesian inference.	P2-2-106 行動試験を用いた視覚反応の可塑性の検討 Plasticity of behavioral visual response ○松本信圭1, 南澤玄樹1, 松木則夫1, 池谷裕二1 ○Nobuyoshi Matsumoto1, Genki Minamisawa1, Norio Matsuki1, Yuji Ikegaya1 東京大院・薬・薬品作用1 Lab. of Chem. Pharmacol., Grad. Sch. of Pharm. Sci., Univ. of Tokyo, Tokyo, Japan1 Neurons in the primary visual cortex (V1) are known to preferentially respond to a specific direction of grating stimulus motion. We previously found that a flash stimulus induces a sustained depolarization (1.5-4.0 s) and that when the depolarization is paired with a drifting grating stimulus, the direction selectivity in V1 neurons will be rapidly altered. In this study, we tried to investigate a behavioral consequence of the neuronal plasticity induced by the pairing protocol. We assessed the visual perceptibility of mice with a virtual optomotor system in which a virtual cylinder comprising a vertical sine wave gratings were projected in the three-dimensional (3-D) coordinate space on four displays arranged in a quadrangle around a testing arena. The virtual cylinder appears to rotate while the gratings are drifted leftward or rightward. Using the system, we found that pairing horizontally-directed drifting grating with a flash reduced animal's tracking responses to the virtual cylinder, suggesting a decline in its visual function. This effect was prevented by local application of the NMDA receptor blocker AP5 on V1. The reduction was not observed when vertically- or obliquely-directed drifting grating was paired with a flash. These findings imply that mammals have a possibility to change their visual perception depending on the experience history. This would challenge the concept that visual function cannot easily be altered by visual experiences after the developing period.
P2-2-107 ヨーロッパモノアラガイにける皮膚光受容器の構造特性 Structural characteristic of dermal photoreceptor in Lymnaea ○滝上慧1, 堀越哲郎2, 榊原学1 ○Satoshi Takigami1, Tetsuro Horikoshi2, Manabu Sakakibara1 東海大院・開発工・生物1, 東海大・工・医用生体2 Graduate School of High-Technology for Human Welfare, Univ of Tokai, Shizuoka1, Dept Bio-Medical Engineering, Univ of Tokai, Kanagawa2 Dermal photoreceptors in Lymnaea stagnalis located on the mantle were histologically characterized. Our previous study demonstrated that the shadow response originated from dermal photoreceptors induces withdrawal response which is the only escape behavior available to this animal through an interneuron, right pedal dorsal 11 from which we can detect the light induced off response (Sunada et al., 2010). Observations based on behavioral pharmacology revealed that cGMP is involved as a second messenger in the dermal photoreceptor (Pankey et al., 2010). Another gastropod, Onchidium, has known to have dermal photoreceptors containing rhodopsin as a photopigment (Katagiri et al., 2001); Scallop has arestin, one of a photo-sensitive protein, in photoreceptors (Gomez et al., 2011). Thus we chose three antibodies; anti-cGMP, anti-rhodopsin, and anti-arestin in order to identify the candidate for substances in dermal photoreceptors whether these antibody are distributed on the Lymnaea mantle. Our electrophysiological recording detected the light off-response from the right parietal nerve with a suction electrode on the mantle. Structures positive against both of each antibody were found in the serial sections of several micron meters thickness fixed with paraformaldehyde from the mantle. The structures containing numerous micro particles were round shape cell of 10 - 30 micron meter in diameter and distributed around pneumostome on the mantle. Histological examination by back-filled method with fluorescent dye and immunohistochemistry with the three antibodies mentioned above confirmed these structures were the most plausible candidate for the dermal photoreceptors in Lymnaea.	P2-2-108 視覚的夢内容の神経デコーディング Neural decoding of visual dream contents ○堀川友慈1,2, 玉置應子1, 宮脇陽一1,3, 神谷之康1,2 ○Tomoyasu Horikawa1,2, Masako Tamaki1, Yoichi Miyawaki1,3, Yukiyasu Kamitani1,2 ATR・CNS・DNI, Kyoto, Japan1, NAIST, Nara, Japan2, NICT, Kyoto, Japan3 Dreaming is a subjective experience during sleep often accompanied by vivid visual contents. Previous research has attempted to link physiological states with dreaming but has not demonstrated how specific visual dream contents are represented in brain activity. The recent advent of machine learning-based analysis has allowed for the decoding of stimulus- and task-induced brain activity patterns to reveal visual contents. Here, we extend this approach to decode spontaneous brain activity associated with dreaming with the assistance by lexical and image databases. We measured the brain activity of sleeping human subjects using fMRI while monitoring sleep stages by EEG. Subjects were awakened when a specific EEG pattern was observed during the sleep-onset period. They gave a verbal report on the visual experiences just before awakening and then returned to sleep. This procedure was repeated to collect over 200 reports in each subject. The words describing visual contents were extracted and grouped into 16-26 synsets, synonym sets defined in the lexical database WordNet, for systematically labeling dream contents. Decoders were trained on fMRI responses to natural images describing each synset, and then tested on dream data. Pairwise and multilabel decoding revealed that accurate classification, detection, and identification regarding dream contents could be achieved with the higher visual cortex, with semantic preferences of individual areas mirroring known stimulus representation. Decoder outputs, on average, represented reported contents toward the time of awakening, while the individual time courses showed complex evolution. Our results demonstrate that specific dream contents are represented in activity patterns of visual cortical areas, which are shared by stimulus perception. Our method uncovers contents represented by brain activity not induced by stimulus or task, which could provide insights into the functions of dreaming and spontaneous neural events.
P2-2-109 サル腹側前部側頭皮質における知覚的および意味的アイデンティティのニューロン表現 Neural representations of perceptual and semantic identities of individuals in the ventral anterior inferior temporal cortex of monkeys ○永福智志1, 中田龍三郎1, 田村了以1 ○Satoshi Eifuku1, Ryuzaburo Nakata1, Ryoi Tamura1 富山大学大学院医学薬学研究部（医学）1 Dept Intgr Neurosci, Univ of Toyama, Toyama1 Recognition of individual identity implicates two kinds of generalization. One is the generalization across different visual appearances. For example, we can easily perceive a particular individual identity from faces which were given in various views. We designate the identity which is generalized across the visual appearances as the perceptual identity. Another is the generalization beyond the visual appearances. Usually each individual is semantically associated with various attributes other than his physical face, such as names, avatars, etc. We designate the identity which is generalized beyond the visual appearance, as the semantic identity. To investigate the neural representations of the perceptual and semantic identities, we recorded neuronal activity from the ventral, anterior inferior temporal cortex (AITv) of macaque monkeys during the performance of an asymmetrical paired-association (APA) task which required associative pairing between an abstract pattern and five different facial views of a single person. To analyze the representations composed by the population of AITv neurons, we computed multi-dimensional stimulus space, based on the activities of AITv neurons in response to the stimulus presentation, and applied multivariate analyses on the stimulus space. The results of the analyses demonstrated that the population of the AITv neurons represent not only the view-invariant, perceptual identity of faces used in the APA task, but also the associative pairing between faces and abstract patterns learned for the APA task, which constitute the semantic identity of the individual. Importantly, the arrangement of the four abstract patterns in the stimulus space was topologically consistent with that of the gravity centers of four face clusters composing of four different identities. Taken together, the results suggest that the AITv represents both perceptual and semantic identities.

上部に戻る

前に戻る