This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
柴田 和久 (名古屋大学・大学院環境学研究科) 13:10-14:00 “Pain: a precision signal for reinforcement learning and control”
Ben Seymour (情報通信研究機構(NICT) 脳情報通信融合研究センター 脳情報通信融合研究室, Computational and Biological Learning Lab, Department of Engineering, University of Cambridge, UK)
14:00-14:50 “Nonlinearity of neuroplastic adaptation of the sensorimotor system through musical training” 古屋 晋一 (Sony CSL)
“Functional analysis of the intercortical communication by multi-point optogenetical manipulation toward elucidation of the cortical internal model formed by a novel motor - visual associational maneuver task.”
ージングの開発」 11 Masako Isokawa (University of Texas Rio Grande Valley) “Ghrelin is the NMDA
receptor-stimulating peptide in the hippocampus” 12 武井 智彦 (京都大学医学研究科/白眉センター) 「巧みな運動制御を司る神経メカニズ
ム」 13 雨宮 薫 (NICT CiNet) “Local-to-distant development of cerebrocerebellar
sensorimotor network in typically developing brain”
10
講演要旨
11
Pain: a precision signal for reinforcement learning and control Ben Seymour (情報通信研究機構(NICT) 脳情報通信融合研究センター 脳情報通信融合研究室, Computational and
Biological Learning Lab, Department of Engineering, University of Cambridge, UK)
Since noxious stimulation usually leads to the perception of pain, pain has conventionally
been thought of as sensory nociception. But in this view, the variability of pain and its sensitivity to a broad array of cognitive and motivational factors have led to a common perception that it is inherently imprecise and intangibly subjective. Here, we argue that this view is misdirected, primarily since it fails to place primacy on the evolved function of pain - to direct behaviour away from harm. Computational models of how the brain achieves this - notably the reinforcement learning model - offer a mechanistic and quantitative understanding of the pain system. These models formalise the distinction between nociception and pain, and show how pain directs learning and decision-making with long-term minimization of harm as the ultimate objective. Importantly, this provides an explanation as to why pain is tuned by multiple factors, including attention, emotion and controllability, and why it is necessarily supported by a distributed network of brain regions. The result is a new view of pain as a precise and objectifiable control signal, offering a fresh approach to studying central contributors to chronic pain, and revealing new targets for therapeutics such as neurofeedback.
12
Nonlinearity of neuroplastic adaptation of the sensorimotor system through musical training 古屋 晋一 (Sony CSL)
Plasticity and meta-plasticity of the motor and sensory systems have gained attention by
psychologists and neuroscientists over decades. To probe impacts of long-term training on the sensorimotor neuroplasticity, musicians provide a unique opportunity. In this talk, I start with introducing neuroplastic changes in the integration mechanism bridging between the motor, proprioceptive, and auditory systems, which was investigated by transcranial magnetic stimulation and electroencephalogram. Through combining the neurophysiological observations with sensorimotor skills that were assessed behaviorally, we attempted to understand functional roles of the neuroplastic changes in the sensorimotor integration. Along with this line of research, I will also briefly show our recent finding of modulation of auditory perception through specific motor actions by musicians, which can provide insights into functional roles of the internal forward model. I will then move on to a negative side of neuroplasticity; maladaptive changes in the sensorimotor system through development of musicians' dystonia. This includes impairment of both the proprioceptive-motor integration and fine motor control as well as their underlying pathophysiology of the sensorimotor system that was identified through transcranial magnetic and direct-current stimulation and functional magnetic resonance imaging. Through my present talk, I'd like to shed light on a non-linear nature of neuroplasticity in terms of acquisition and loss of sensorimotor skills.
13
興奮と抑制の比率が知覚学習の安定性を制御する 柴田 和久 (名古屋大学・大学院環境学研究科)
記憶の形成や技能の獲得は、脳の可塑的な変化によって実現されています。一方で、既存の
記憶や技能を安定的に保持するために、脳は適度に安定的である必要もあります。可塑性と安定性のジレンマと呼ばれるこの古くて新しい問題を、脳はどのように解決しているのでしょうか。本トークでは、視覚知覚学習を例に、技能の安定性に視覚野の興奮と抑制の比率が関わることを示した一連の研究を紹介します。ひとつ目の研究(Shibata et al., Nat Neurosci, 2017)では、視覚訓練の直前、直後、数時間後に、磁気共鳴分光法(MRS)を用い、興奮性の神経修飾物質であるグルタミン酸と抑制性の神経修飾物質である GABAの濃度比率(興奮抑制比率)を測定しました。その結果、知覚学習が不安定な状態では視覚野の興奮抑制比率が高まり、安定的な状態ではその比率が低くなる、つまりより抑制的な状態になることがわかりました。ふたつ目の研究(Bang, Shibata et al., Nat Hum Behav, 2018)では、知覚学習の固定化が一度起こったあと、さらに再活性された状態の視覚野における興奮抑制比率を測定しました。一度固定された知覚学習が再活性によって再び不安定になると、視覚野の興奮性が上昇することがわかりました。このふたつの研究結果は、知覚学習の安定性は興奮抑制比率によって制御されていることを示唆しています。
14
The active inference in decision making by adult zebrafish revealed by in-vivo imaging of the telencephalic neural activities in the closed-loop virtual reality environment 岡本 仁 (理化学研究所 脳神経科学研究センター 意思決定回路動態研究チーム)
Selecting a logical behavioral choice from the available options, i.e. decision making is
essential for animals. Recently, adult zebrafish has drawn attention as a model animal for the study of decision making due to its capability of various adaptive behaviors1) and the conservation of the basic telencephalic structure2) which is contributing to the decision making. In the present study, we aimed at directly addressing this process by establishing the closed-loop virtual reality system for the head-tethered adult zebrafish with the 2-photon calcium imaging system. The adult zebrafish harboring G-CaMP7 in the excitatory neurons were trained to perform visual-based active and passive avoidance tasks and simultaneously the neural activities were imaged in the cellular level. Furthermore, after learning was once established in the closed-loop condition, we suddenly removed the visual feed-back to make the system open-loop. The Non-negative Matrix Factorization analysis revealed the one ensemble of neurons whose activities were suppressed by the recognized backward movement of the landscape, and the other ensemble suppressed by reaching the goal compartment. These ensembles recovered throughout the trials under the open-loop condition. These results suggest that these two ensembles encode the prediction errors between the status represented by the real sensory inputs and the favorable status to successfully escape from the danger, i.e. visual inputs of the backward moving landscape and the red wall color of the goal compartment, and the behaviors are taken so that these errors become minimum. Our result supports that the adult zebrafish behaves in decision making based on the active inference in the free energy principle3), where agents take actions to suppress the prediction errors by trying to make the internal representation of the bottom-up sensory states match those of the top-down predictions, and demonstrate the strong conservation of the basic principle of decision making throughout the evolution. 1: Aoki, T. et al., Imaging of neural ensemble for the retrieval of a learned behavioral program. Neuron 78: 881-894 (2013) 2: Mueller, T. et al., The dorsal pallium in zebrafish, Danio rerio (Cyprinidae, Teleostei). Brain
Res. 1381: 95-105 (2011) 3: Friston, K.J. et al., The graphical brain: Belief propagation and active inference. Network
Neuroscience 1: 381-414(2017)
15
Balancing brain plasticity/stability across the lifespan Takao Hensch (Harvard University)
Brain function is largely shaped by experience in early life. Our work has focused on the
biological basis for such “critical periods,” identifying both “triggers” and “brakes” on plasticity. Strikingly, the maturation of particular inhibitory circuits is pivotal for the onset of these windows. Manipulations of this excitatory-inhibitory balance can accelerate or delay timing regardless of chronological age. Closure of critical periods in turn reflects an active process, rather than a purely passive loss of plasticity factors. Lifting these brakes allows the reopening of plastic windows later in life. Both the shifted onset and brain rejuvenation provide novel insight into the etiology of mental illnesses and potential therapeutic strategies for recovery of function in adulthood.
16
Learning in the mouse cortex 牧野 浩史 (シンガポール南洋理工大学)
The brain is considered to optimize an internal model of a constantly changing environment
to minimize potential cost. To achieve this, the brain interacts with the environment through perception and action, where perception updates the brain’s internal states representing how sensation is caused, while action allows the animal to explore the environment. Learning is essentially an iterative process of perception and action through intricate interactions with the environment and involves changes in communication strategies across different brain areas, where signals carrying different information take distinct routes. Here I highlight recent findings of how different forms of learning can alter activity propagation in the mouse cortex. The experimental evidence with 2-photon and wide-field calcium imaging suggests that even distinct forms of learning such as sensory learning or motor learning converge onto the same principle of the brain operation, namely recruitment of top-down control. Such a universal mode of operation shaped by learning may facilitate the cost-minimization process by changing computations of the local circuitry dynamics.
17
The science of brain-machine interface in primates Junichi Ushiba (牛場 潤一) (Associate Professor, Department of Biosciences and Informatics, Faculty of Science and Technology,
Keio University, Japan; Keio Institute for Pure and Applied Sciences (KiPAS), Graduate School of Keio
University, Japan.)
Brain-Machine Interface (BMI) is a technology that decodes natural neural information from a
targeted sensorimotor-related brain area (NeuroImage 2018; Front Hum Neurosci 2017; J Neurophysiol 2013), and translates it into machine control signals. It has achieved telepathy-like machine control (BMC Neurosci 2010) or cyborg-like limb control, but also it achieves manipulation of the targeted neural activities via visual/somatosensory feedback (Front Neuroeng 2014; Clin Neurophysiol 2013). Retention of improved brain activity and motor behavior through BMI neural manipulation is feasible as medical application (Science 2017; Restor Neurol Neurosci 2016; Prog Brain Res 2016) for recovery from post-stroke hemiplegia (J Rehabil Med 2011), motor paralysis due to incomplete spinal cord injury, and dystonic writer's cramp (BMC Neurosci 2014).
BMI also have a potential impact on the science in motor control and learning, since BMI can characterize the spatio-temporal feature of the sensorimotor processing in a computational fashion (Brain Topogr 2015; J Rehabil Med 2014). Also, manipulation of a targeted brain activity though BMI can tell causal relationship between the brain activity and behaviors (J Rehabil Med 2015; J Rehabil Med 2014). In order to solidify neurobiological evidences of neural decoding and manipulation of BMI, we have recently been taking an optogenetic approach of large-scale neural recordings from primary motor cortex (M1) (Cell Reports 2018). So far, we established stable recording of activating ~100 cortical neurons in the M1 Layer 5 from naturally behaving non-human primates, Common Marmoset. The result of neural decoding tells us that M1 cortical neurons encoded movement direction information a few seconds after the offset of arm reaching movement. This suggests that M1 is not a simple final path of motor signal output to the muscles, but is an important node of post-processing in the sensorimotor system. Successful decoding of movement direction (~90%) promises to establish volitional control of machine devices in a BMI fashion, without presence of actual movement. Neurobiological natures of motor learning and motor recovery will be discussed deeply with this advanced BMI technology in near future, and clinical application of BMI in humans will be further promoted by such neurobiological reasonings.