視覚3 Vision 3
O2-7-3-1 V4野細胞の自然動画刺激に対する視覚応答のIn vivo 2光子カルシウムイメージング In vivo 2-photon calcium imaging of visual responses to natural movies in monkey V4 ○池添貢司1,2, 西本伸志3, 竹内遼介1, 深澤宇紀4, 齋藤優介4, 藤田一郎1,2 ○Koji Ikezoe1,2, Shinji Nishimoto3, Ryosuke Takeuchi1, Takanori Fukazawa4, Yusuke Saito4, Ichiro Fujita1,2 大阪大院・生命・認知脳1, 脳情報通信融合研究センター2, カリフォルニア大バークレー校3, 大阪大・基礎工4 Grad Sch of Front Biosci, Osaka Univ, Toyonaka, Japan1, CiNet, Osaka Univ and NICT, Suita, Japan2, Univ of California, Berkley, Berkley, USA3, Dept of Eng Sci, Osaka Univ, Toyonaka, Japan4 Cortical area V4 of monkeys is located in the middle tier of the ventral visual pathway. V4 neurons respond selectively to basic visual attributes such as orientation and spatial frequency, and to complex features such as curved contours, 2D shapes, and patterns/textures. However, it is unclear how V4 neurons are spatially arranged in the cortex according to their stimulus preferences. Quantitative and comprehensive descriptions of the response properties as well as the identification of the locations of neurons are necessary to reveal the functional architecture at the single-cellular resolution. To this end, we recorded responses of V4 neurons to natural movie sequences by applying in vivo 2-photon calcium imaging techniques in immobilized monkeys (Macaca fascicularis) under opiate-analgesia. All the recording regions were in layer 2. They spanned 150 µm x 150 µm, and contained 46.4 ± 9.7 (mean ± s.d.) neurons loaded with a fluorescent calcium indicator, Oregon Green 488 BAPTA-1 AM. V4 neurons exhibited a series of calcium transient responses to natural movie. We fit a motion-energy encoding model to the neural responses evoked by a 30-minute natural movie. The models were then evaluated by testing how well they predicted the neural responses to another 30-minute natural movie (evaluation movie). The models predicted the responses to the evaluation movie with a significant degree of accuracy. The receptive fields of the neurons estimated from the models were spatially localized in the visual field. The neurons recorded within a single region often showed similar receptive field positions. The models also revealed orientation and spatial frequency preferences of the recorded neurons. Thus, the combination of the model-based analysis with in vivo 2-photon calcium imaging techniques allows a quantitative description of response properties of individual neurons and their precise spatial arrangement in V4.	O2-7-3-2 適応的サンプリングにより同定したマカクV4ニューロンのテクスチャ選択性を決める画像特徴 Image features determining texture selectivity of macaque V4 neurons revealed by adaptive sampling ○岡澤剛起1,2, 田嶋達裕3, 小松英彦1,2 ○Gouki Okazawa1,2, Satohiro Tajima3, Hidehiko Komatsu1,2 生理研・感覚認知情報1, 総研大・生理科学2, NHK放送技術研究所3 National Institute of Physiological Sciences, Okazaki, Japan1, SOKENDAI, Okazaki, Japan2, STRL, NHK, Tokyo, Japan3 Categorization of materials such as bark, fabric, water is a very important function in object vision. Natural materials can be characterized by specific textures that consist of a complex combination of low-level image features such as local contrast, spatial frequency and orientation, but how these complex combinations are represented in the visual system is largely unknown. In this study, we attempted to examine this problem by systematically mapping responses of macaque V4 neurons to a large number of texture images defined in high dimensional space consisting of complex combinations of visual features derived from a parametric texture synthesis algorithm developed by Portilla and Simoncelli (2000). The algorithm uses image statistics derived from wavelet subbands and the textures were defined in a 7 dimensional parameter space generated by reducing the dimension of texture synthesis parameters by using Fisher linear discriminant analysis. To efficiently sample textures that evoke stronger neuronal responses, we introduced an adaptive sampling procedure (Yamane et al. 2008). In the experiment, we first recorded neuronal responses to 50 textures randomly selected from a pre-generated large image set. Subsequently, we selected other 50 textures such that textures neighboring those that evoked stronger activities are more frequently chosen. By repeating this procedure, we obtained neuronal responses to 250~500 textures. When we fitted the responses in the 7 dimensional parameter space used to define textures, we found that in many neurons responses can be explained by these parameters at least to some extent. A control experiment showed that neuronal activities tended to change for the texture images whose phases were randomized but did not change for different texture images generated by the same synthesis parameters. These results suggest that texture selectivity of V4 neurons can be explained by selectivity to higher order image statistics of textures.
O2-7-3-3 V4ニューロンの色選択性応答に対する輝度コントラストの影響 Effect of luminance contrast on color selective activity in area V4 ○波間智行1,2, 小松英彦1,2 ○Tomoyuki Namima1,2, Hidehiko Komatsu1,2 生理学研究所1, 総研大・生命科学2 National Institute for Physiological Sciences, Okazaki, Japan1, Dept Life Sci, SOKENDAI, Okazaki, Japan2 Luminance contrast of a color stimulus may cause significant influence on the perceived color, and it is important to know how neural representation of color signal is mixed with luminance information to understand the effect of luminance contrast on color perception. We have previously reported that color selective responses in the anterior inferior temporal cortex (AITC) is not affected by the luminance contrast whereas those in the posterior inferior temporal cortex (PITC) tend to be affected in a color dependent manner. In the present study, we examined the effect of luminance contrast on the color selectivity of V4 neurons as an attempt to understand how the luminance and color signals are combined along the ventral pathway. We employed the same method as used in the study of AITC and PITC: Two color stimulus sets were used. Both color stimulus sets consisted of 15 colors including one white point. Color stimulus sets were displayed with higher luminance (20cd/m2) or lower luminance (5cd/m2) than the gray background (10cd/m2). Single neuron activities were recorded from the prelunate gyrus in visual area V4 of monkeys performing a visual fixation task. We found that the effect of luminance contrast varies across V4 neurons, but in general responses of V4 neurons tended to be more affected by the luminance contrast than IT neurons. When we examined the correlation between the responses of the population of color selective V4 neurons to the bright stimuli and those to the dark stimuli for each color, we found that the effect of luminance contrast was large for sharply color selective neurons but small for broadly color selective neurons. In the former, a large effect was observed for colors with low saturation. These results suggest that color signal mixed with luminance signal is selectively conveyed from sharp color selective neuron in V4 to PITC and that the separation between color signal and luminance signal takes place in higher stage than PITC.	O2-7-3-4 Representation of Stereoscopic Depth in Visual Area V4 at the Single Neuron and the Population Levels ○Mohammad Abdolrahmani1, Takahiro Doi1, Hiroshi M. Shiozaki1, Ichiro Fujita1 Lab for Cognitive Neuroscience, School of Frontier Biosciences,Osaka University1 The visual system calculates depth by solving the binocular correspondence problem. Humans successfully perceive depth in random-dot stereograms (RDSs) despite the abundant false combinations of visual features. However, the stereoscopic mechanisms are so confounded by contrast reversal of all dots in one of the images that no binocular depth is perceived in anti-correlated RDSs (aRDSs). The mechanisms highly tolerate anti-correlation: depth perception is perfectly preserved even when half the dots are anti-correlated, and it gradually degrades for stronger anti-correlation. Here, we studied the neural mechanisms underlying the correspondence process by examining neural responses to the graded anti-correlation of RDSs in macaque visual area V4. Center-surround RDSs were presented on receptive fields of recorded cells. Binocular disparity and the level of anti-correlation were randomly changed trial to trial for the center patch of the RDSs, while the surrounding annulus was always 100% correlated and fixed at zero disparity. Disparity selectivity dropped sharply when only 35% of the dots were anti-correlated and remained weak for stronger levels of anti-correlation. Thus, disparity selectivity of a single V4 cell is inconsistent with depth perception. Rather, human performance is consistent with pooled responses of V4 cells. The phase of attenuated disparity tuning curves did not change when anti-correlation was applied to less than half the dots. For stronger anti-correlation levels, the phase shifted with varying degrees in different directions across cells. Pooling responses of cells with similar disparity preferences improves the disparity selectivity for anti-correlation levels of 0-50%. Pooling responses to anti-correlation levels of > 50% does not yield a reliable signal due to inconsistent tuning shapes. We suggest that pooling activities across V4 cells allows depth perception to be robust to degradation in the correlation between left and right eye images.

上部に戻る

前に戻る