シンポジウム
38 知覚の内的表象から探る脳の予測機能
38 Internal representation of perception and predictive coding
座長：久保 郁（国立遺伝学研究所）・渡部 喬光（IRCN） |
|---|
| 2022年7月3日 14:00～14:24 沖縄コンベンションセンター 会議場B5～7 第4会場
4S04a-01
錯視反応を用いた動きの視覚情報処理と予測を担う神経回路の解析
Motion processing and prediction using optical Illusion
*久保 郁(1)
1. 国立遺伝学研究所
*Fumi Kubo(1)
1. National Institute of Genetics, Mishima, Japan
Keyword: visual system, optical illusion, zebrafish, prediction
Direction-selective (DS) neurons compute the direction of motion in a visual scene. Previously, brain-wide functional imaging in larval zebrafish revealed hundreds of DS neurons scattered throughout the brain. However, the exact population that causally drives motion-dependent behaviors, e.g. compensatory eye and body movements, remains largely unknown. To identify the behaviorally relevant population of DS neurons, we employed a classical optical illusion, a motion aftereffect (MAE), which causes the well-known “waterfall illusion”. Together with region-specific optogenetic manipulations and cellular-resolution volumetric calcium imaging, we found that MAE-responsive neurons represented merely a fraction of the entire population of DS cells in larval zebrafish. These MAE-responsive neurons were spatially clustered in a nucleus in the ventral lateral pretectal area. Laser ablation of the MAE-responsive cells abolished the optokinetic eye movements, whereas optogenetic activation of them drove the entire cycle of optokinetic eye movements in the absence of the visual stimulus. Thus, our illusion-based behavioral paradigm, combined with the cellular resolution calcium imaging and optogenetics, identified key circuit elements of global motion processing in the vertebrate brain. In addition, since the MAE paradigm allows us to investigate how visual perceptions are generated based on past experiences, the identified circuit elements provide an entry point to elucidate neural circuits that underlie predictive coding. | | 2022年7月3日 14:24～14:48 沖縄コンベンションセンター 会議場B5～7 第4会場
4S04a-02
予測誤差最小化による行動の最適化のための神経回路の探索
Quest for the neural circuits to optimize behavior by minimization of prediction error
*岡本 仁(1)
1. 理化学研究所 脳神経科学研究センター
*Hitoshi Okamoto(1)
1. RIKEN Center for Brain Science
Keyword: virtual reality, prediction error, decision making, neural imaging
Based on the recent finding that the brain of adult zebrafish, a small tropical fish, is small and relatively simple, but shares the most basic structure with that of the mammalian brain, we have devised a new virtual reality setup specifically for adult zebrafish and studied the time-lapse change of neuronal activity in the brain of zebrafish swimming in this device as they learn to avoid danger. As a result, we found in the part of the zebrafish brain that corresponds to the cerebral cortex of mammals that repeated avoidance training produces a cluster of neurons whose activity represent the prediction error between the training-informed prediction of the optimal future situation and the actual view, and that the fish behaves in a way that minimizes the activity of this cluster of neurons, resulting in the most efficient escape behavior. To elucidate how such prediction and prediction error is generated by training, and how it is implemented for the improvement of behaviors, we are now observing neural activities of the deeper parts of the telencephalon evolutionarily equivalent to the direct and indirect pathway neurons of the striatum and the neurons in the globus pallidum during escape behavior in virtual reality in parallel with the trans-synaptic neural tracing with various viral vectors to elucidate how these neurons are connected. These studies have lead us to the working hypothesis that the dorsal pallium (equivalent to cortex)-basal ganglia pathways have two mutually interacting components, one for the generation of prediction of the optimum status for fish under given circumstances and the other for the selection of proper behavior for minimization of the prediction error to realize the predicted optimum status. | | 2022年7月3日 14:48～15:12 沖縄コンベンションセンター 会議場B5～7 第4会場
4S04a-03
Thalamocortical networks underlying perceptual inference during sensory decision-making
*Lukas Ian Schmitt(1)
1. RIKEN Center for Brain Science
Keyword: Thalamus, Inference, Decision-Making, Cortex
Information that reaches the brain from the senses is often ambiguous and disconnected but our perception of the world (internal model) is stable and continuous. The relative insensitivity of perceptual experience to incomplete and uncertain information suggests that the brain can maintain a representation of recent sensory experiences (sensory history) and use it to interpret subsequent ambiguous inputs (sensory inference). While recent work has suggested that some “higher-order” thalamic nuclei might stabilize representations of information in cortical networks, it remains unclear whether such interactions can support dynamic maintenance of sensory history over the variable timescales needed to allow sensory inference. Here I present evidence that network activity the posterior parietal cortex (PPC) reflects patterns of previously encountered auditory stimuli and that these representations are dynamically controlled by interactions between this cortical area and its thalamic counterpart, the pulvinar (PUL). In addition, we find that history representations created through PPC/PUL interactions enable inferences to be made based on ambiguous stimuli during a sensory decision-making task. Together, these findings indicate that PPC/PUL interactions play an important role in establishing and updating models of the sensory environment and suggest a critical role for thalamocortical interactions in maintaining stable perceptual experience. | | 2022年7月3日 15:12～15:36 沖縄コンベンションセンター 会議場B5～7 第4会場
4S04a-04
SalienceとConfidenceによる妄想の理解
Understanding delusion from salience and confidence.
*宮田 淳(1)
1. 京都大学
*Jun MIYATA(1)
1. Kyoto University
Keyword: Delusion, salience, confidence
Delusion is defined as three factors: 1) a falsely formed belief with 2) conviction and 3) incorrigibility (Jaspers, 1913). This definition can be interpreted from neuroscientific concepts of salience, confidence, and belief-updating. The midbrain-striatum dopamine neurons are known to code the salience of stimuli. In psychosis, the hyper-dopaminergic state of the striatum is considered to cause aberrantly heightened salience attribution to daily-life stimuli, leading to the formation of delusion and hallucination (aberrant salience hypothesis: Kapur, 2003). Meanwhile, people with delusions are known to need less amount of evidence for decision-making compared with healthy people, called the jumping to conclusions (JTC) bias (Garety et al, 1991). On the other hand, healthy people’s decision-making is known to be conservative, needing more amount of evidence than Bayes theorem expects. This tendency is called the conservatism bias (Phillips and Edwards, 1966). The neural correlates of JTC and conservatism bias, and their relationship with aberrant salience were unclear. Using a probabilistic reasoning task called beads task and resting-state functional magnetic resonance imaging (rsfMRI), we found an association between JTC/conservatism and functional connectivity of the striatum. Salience processing is not limited to the midbrain-striatum system, and visual and auditory salience has been extensively studied in separate backgrounds. It was though unclear there was domain-specificity between subjective aberrant salience and specific salience systems. Using a subjective salience experience inventory, we showed that subjective sensory-salience was associated with the functional connectivity between the visual and sensorimotor networks, while subjective cognitive salience was associated with the connectivity between the midbrain and striatum. Confidence is one of the measures of metacognition, and recent studies indicate the association between the prefrontal cortex and confidence. We found that the strength of confidence in delusional belief was associated with the functional connectivity between the anterior. Our series of studies provide an understanding of delusion, from the viewpoints of salience and confidence. | | 2022年7月3日 15:36～16:00 沖縄コンベンションセンター 会議場B5～7 第4会場
4S04a-05
神経遷移ダイナミクスと内部表象の揺らぎとその制御
Brain state dynamics underpinning perceptual internal model and controlling cognitive modes.
*渡部 喬光(1)
1. 東京大学国際高等研究所ニューロインテリジェンス国際研究機構
*Takamitsu Watanabe(1)
1. UTokyo WPI-IRCN
Keyword: Brain dynamics, TMS, energy landscape analysis, flexibility
Our internal model of perception is not always stable, and such dynamics of the internal model are thought to be underpinned by whole-brain neural dynamics. In this talk, I will first show that specific brain state dynamics are responsible for such fluctuation of the human perceptual internal model. Then, I will introduce a brain-state-dependent neural stimulation system that is composed of an offline/online EEG analysis and a transcranial magnetic stimulation (TMS) device. Using the neural intervention method, I will corroborate the brain-behaviour association and demonstrate that human perceptual internal models could be externally controlled through modifying the brain state dynamics in more physiological manners than conventional neural intervention methods. Moreover, I will present preliminary results of the generalisation of this approach. That is, I will talk about several distinct types of brain-state dynamics, such as one that is highly associated with atypically unstable cognitive mode and another one that is strongly correlated with overly stable cognition. Also, I will elucidate how the brain-state-dependent intervention can stabilise/de-stabilise such neural dynamics and induce cognitive and behavioural changes. Furthermore, I will present results on long-term neural and behavioural effects of such non-invasive neural intervention, which may become a basis for future clinical application. Finally, I will interpret these neural and behavioural observations from psychological and neurobiological perspectives. | | | |
|
|
