神経データ解析 Neuronal Data Analysis
P1-2-223 脳リズムからの観測時系列データに基づく位相モデルの統計的推定 Statistical estimation of phase models for interacting brain rhythms using measured time series data ○太田絵一郎1,2, 青柳富誌生1,2 ○Kaiichiro Ota1,2, Toshio Aoyagi1,2 京都大院・情報1, 科学技術振興機構CREST2 Grad Sch Informatics, Kyoto Univ1, JST CREST2 Rhythmic brain activity are observed in EEG, local field potential, and unit recordings of regularly firing neurons. Moreover, simultaneous recording of multiple rhythms has been made easier recently. Synchrony among multiple brain rhythms is an interesting research topic today, as it is likely to be involved in cortical functions such as attention. Synchrony is caused by some inherent interactions, but their detailed mechanism is generally complicated, which hampers the analysis of synchronization phenomena as well as understanding their possible contribution to cortical functions. It is thus desirable to analyze rhythms and their interactions using a simple, comprehensible model. In this work, we propose a methodology useful for constructing a mathematical model of interacting rhythms that is simple and easily analyzable. Our statistical method estimates the parameters in the phase model such that the model will explain the observed time series data such as EEG waveforms. In the phase model, each rhythmic unit is described by a single variable, the phase. That is, if we have N interacting rhythmic elements, the corresponding phase model consists of just N phase variables, much less than the dimensionality of underlying detailed dynamics. Despite its simplicity, the phase model can reproduce a variety of dynamical states including, e.g., bistability and switching behavior. This means that estimated models can give information about the dynamics beyond simple correlations. The major advantage of our method is that it requires only oscillatory time series data. We carried out numerical experiments using nonlinear oscillators and spiking neuron models and confirmed that sufficient length of data could lead to reliable estimation of possible pairwise interactions. Our method can readily be applied to real data, and provides an efficient tool for understanding the link between brain rhythms and functions.	P1-2-224 心的な視標運動追跡課題遂行中のMEGデータにおける脳活動とアーチファクトの分離 Source separation of cortical and extra-brain source activities in real MEG data during covert pursuit eye movements ○森重健一1,2, 吉岡琢2, 石井信2,3, 佐藤雅昭2, 川人光男4 ○Ken-ichi Morishige1,2, Taku Yoshioka2, Shin Ishii2,3, Masa-aki Sato2, Mitsuo Kawato4 富山県立大・工・知能デザイン工学科1, ATR・脳解析2, 京都大院・情報学研究科3, ATR・脳情報4 Dept Intelligent Systems Design Engineering, Toyama Pref. Univ, Toyama1, NIA, ATR, Kyoto2, Graduate school of informatics, Kyoto Univ, Kyoto3, CNS, ATR, Kyoto4 The measurement of MEG signals is contaminated by large magnetic artifacts stemming from eye movements, heart beats, and so on. These artifacts can be of orders of magnitude larger than those of typical brain signals, thus making cortical current estimation extremely difficult. To overcome this difficulty, we have proposed an artifact removal method employing extra dipoles, which simultaneously estimates intra- and extra-cranial source currents by placing dipoles not only on cortical surfaces but also on extra-brain locations. This method has prominent advantage, because it does not require strong a priori knowledge on electrical or magnetic data. On the other hand, two kinds of constraint were introduced in terms of prior information to ease the ill-posedness of the inverse problem: first, the prior and posterior current variances should be the same; and second, the spatial patterns of fMRI activities and estimated current power should be highly correlated. Although we previously found this method worked for simulated MEG data, in this study we examined its applicability to real MEG data during covert pursuit eye movements. Since the both eyes and the heart were assumed to be main extra-brain sources, extra dipoles were placed on their locations. We found the eye currents estimated by our method were of the consistent order of magnitude with that reported in the existing study. The magnetic field of heart beats would comprise a P-QRS-T cycle, and we observed the estimated heart currents were consistent with the cycle. The estimated cortical currents exhibited reasonable spatial-temporal patterns. These encouraging results suggest our method can reasonably resolve cortical and extra-brain source activities from obfuscating MEG signals by introducing proper prior information.
P1-2-225 DTIトラクトグラフィーネットワーク中の破損結合推定 Predicting corrupted connections in networks of DTI tractography ○倉重宏樹1, 磯谷悠子1, 大高洋平2,3, 大須理英子1 ○Hiroki Kurashige1, Yuko Isogaya1, Yohei Otaka2,3, Rieko Osu1 ATR脳情報通信総合研1, 慶應義塾大学 医学部2, 東京湾岸リハビリテーション病院3 CNS, ATR, Kyoto1, Department of Rehabilitation Medicine, Keio University, Tokyo2, Tokyo Bay Rehabilitation Hospital, Chiba3 Diffusion MRIs, such as diffusion tensor imaging (DTI), are non-invasive imaging techniques to measure the neural fibers in brain white matter. By applying the well-established algorithms, whole-brain connectivity networks can be reconstructed (e.g. Hagmann et al. 2008). However, the data measured with diffusion MRI are dependent on the imaging conditions, including scanners, values of parameters and head-coils. Therefore, the reconstructed connectivity networks are highly affected from the conditions. That is, the connections in the networks may be overestimated or underestimated. In the present study, we treat this situation as a problem of corruption of the connection or 'link' in the network and developed a method to solve this using link prediction. In one of the known link prediction methods proposed by Clauset et al. (2008), hierarchical structures existing behind the targeted network are estimated, and then the corrupted links are predicted using the estimated hierarchical structures. Since graph theoretical analysis suggests the existence of some types of hierarchical natures in the networks of brain connectivity (Meunier et al. 2010) such method based on the hierarchical structure can be usable to reconstruct corrupted neural connection.However, the most of the link prediction methods including Clauset's are intended only for the binary (all-or-none) networks. Since the anatomical neural connection strength is usually has continuous, the way to predict the continuous values of link strength is desired.The method we propose predicts the continuous values of the links based on the Clauset's method. To test the method, we identified and corrected the corrupted connections in whole-brain connectivity networks. Our method is general and not limited to use for DTI data. It may provide a way to derive the hidden hierarchical structures and to correct the corrupted link in any network having link strength of continuous-value.	P1-2-226 変動発火レートの情報伝送 Information transmission on variable neuronal firing rate ○小山慎介1 ○Shinsuke Koyama1 統計数理研究所1 The Institute of Statistical Mathematics1 The question of how much information can be theoretically gained from variable neuronal firing rate with respect to constant mean firing rate, is investigated. For this purpose, we employ the relative entropy (Kullback-Leibler divergence) as a measure of information. We first give a statistical interpretation of this information in terms of detectability of rate variation: the lower bound of detectable rate variation, below which the temporal variation of firing rate is undetectable with a Bayesian decoder, is entirely determined by this information. We also show that the information depends not only on the variation of firing rates (i.e., signals), but also significantly on the dispersion properties of neuronal firing described by the shape of interspike interval (ISI) distribution (i.e., noise properties). Interestingly, the gamma distribution attains the theoretical lower bound of the information among all ISI distributions when the coefficient of variation of ISIs is given. With the basis of the theoretical investigations, we propose a practical procedure for estimating the information from spike trains, and report results of the real data analysis.
P1-2-227 ヒト大脳ネットワークの単純性：最大エントロピー法に基づく推定 Simplicity of human brain networks: resting-state functional brain networks can be accurately described by a pairwise maximum entropy model ○渡部喬光1, 廣瀬聡1, 和田裕之2, 今井宜雄2, 町田徹2, 白水一郎2, 小西清貴1, 宮下保司1, 増田直紀3 ○Takamitsu Watanabe1, Satoshi Hirose1, Hiroyuki Wada2, Yoshio Imai2, Toru Machida2, Ichiro Shirouzu2, Seiki Konishi1, Yasushi Miyashita1, Naoki Masuda3 東京大学大学院医学系研究科統合生理学1, NTT東日本関東病院放射線科2, 東京大学大学院情報理工学系研究科数理情報学専攻3 Department of Physiology, The University of Tokyo, School of Medicine, Tokyo, Japan1, Department of Radiology, NTT Medical Center Tokyo, Japan2, Department of Mathematical Informatics, The University of Tokyo, Japan3 During rest, the human brain shows a large amount of spontaneously fluctuating activity. Previous studies have revealed that the global brain activity during rest constitutes complex brain networks, which is called the resting-state networks (RSNs). Although the RSNs are considered to be related to fundamental cognitive functions such as memory maintenance and self reference, the level of complexity of the RSNs has not been quantified. In the present study, we evaluate the complexity by fitting a pairwise maximum entropy model (MEM) to resting-state human brain activities obtained by functional magnetic resonance imaging (fMRI). Consequently, the pairwise MEM, which takes into account region-specific activity rates and pairwise interactions, can be robustly and accurately fitted to resting-state fMRI signals. Moreover, to validate the approximation of the RSNs by the pairwise MEM, we estimate how accurately the functional interactions estimated by the pairwise MEM reflect anatomical connections. As a result, the functional interactions based on the pairwise MEM are more similar to the anatomical connections than those based on other competitive methods to evaluate functional connectivity among brain regions. These findings indicate that the complex RSNs can be described by a relatively simple second-order statistical model. In addition, the model would provide a possible method to derive physiological information about various large-scale brain networks.	P1-2-228 神経細胞集団のスパース発火活動によって表現されるスパイク高次相関 The simultaneous silence of neurons explains higher-order interactions in ensemble spiking activity ○島崎秀昭1, , 池谷裕二3, 豊泉太郎1 ○Hideaki Shimazaki1, Sadeghi Kolia2, Yuji Ikegaya3, Taro Toyoizumi1 理研・脳科学総合研究センター1, Dept. of Statistics, Columbia University2, 東京大・薬学研究科3 RIKEN Brain Science Institute, Wako-shi, Saitama, Japan1, Dept. of Statistics, Columbia University, New York, USA2, Graduate School of Pharmaceutical Sciences, The University of Tokyo, Tokyo, Japan3 Collective spiking activity of neurons is the basis of information processing in the brain. Sparse neuronal activity in a population of neurons limits possible spiking patterns and, thereby, influences the information content conveyed by each pattern. However, because of the combinatorial explosion of the number of parameters required to describe higher-order interactions (HOIs), the characterization of neuronal interactions has been mostly limited to lower-order interactions, such as pairwise interactions. Here, we propose a new model that characterizes population-spiking activity by adding a single parameter to the previously proposed pairwise interaction model. This parameter describes the fraction of time a group of neurons is simultaneously silent, which can be alternatively expressed as a specific combination of HOIs. We apply our model to groups of neighboring neurons that are simultaneously recorded from spontaneously active slice cultures from the hippocampal CA3 area. Most groups of neurons that are not adequately explained by the pairwise interaction model exhibit significantly longer periods of simultaneous silence than the chance level expected from firing rates and pairwise correlations, demonstrating that simultaneous silence is a common property coded by HOIs. To confirm that the simultaneous silence is also a major property, we systematically obtained a one-dimensional data-driven HOI term that is asymptotically optimal when added to a pairwise-interaction model. This analysis exhibited the structured HOIs expected from the simultaneous silence of neurons, i.e., positive pairwise interactions are followed by negative triple-wise interactions, and then positive quadruple-wise interactions. These results suggest that seemingly complex HOIs can be explained by simultaneous silence of multiple neurons. We discuss the implication of simultaneous silence for our understanding of the underlying circuit architecture and information coding.
P1-2-229 fNIRSにおける頭皮血流のグローバルな干渉の低減: 短距離プローブと一般線形モデルを用いた解析の検討 Reduction of global interference of scalp hemodynamics in functional near-infrared spectroscopy: methodological investigations for an analysis using short-distance probe and general linear model ○佐藤貴紀1, 武田湖太郎2,3, 南部功夫1, 大須理英子3, 和田安弘1 ○Takanori Sato1, Kotaro Takeda2,3, Isao Nambu1, Rieko Osu3, Yasuhiro Wada1 長岡技術科学大学1, 国立病院機構 村山医療センター2, 国際電気通信基礎技術研究所 脳情報研究所3 Nagaoka Univ of Tech, Niigata, Japan1, National Hospital Organization Murayama Medical Center, Tokyo, Japan2, ATR CNS, Kyoto, Japan3 Functional near-infrared spectroscopy (fNIRS) provides the non-invasive measurement of concentration changes in oxygenated and deoxygenated hemoglobin (Oxy- and Deoxy-Hb) in the cerebral cortex. Recent studies have shown that fNIRS signal is contaminated by the changes mainly originated from scalp (scalp hemodynamics). The scalp hemodynamics may cause a false positive cerebral activity detected in inactive cortical regions. To reduce the influences of the artifact, we propose a new method that estimates the global scalp hemodynamics from four Short-channels (source-detector distance of 15 mm) by using principal component analysis (PCA), and that identifies the cerebral hemodynamics by using general linear model (GLM) incorporating the estimated global scalp hemodynamics in the design matrix. To examine the feasibility of the proposed method, we performed an experiment and simulations in a standard commercial fNIRS setting of 16 probe pairs. In the experiment, we placed 18 Short-channels over motor related areas of both hemispheres and measured fNIRS signals during finger tapping task for 3 healthy subjects. Applying PCA to this data, we confirmed that the scalp hemodynamics in Oxy-Hb was globally distributed and such component can be extracted from a few Short-channels. The estimation accuracy of cerebral activity was evaluated in simulations. We generated artificial data of 43 Long-channels (source-detector distance of 30 mm) which contained cerebral and scalp hemodynamics, and four Short-channels which contained only scalp hemodynamics. As a result of applying the proposed method to this data, we found a greater improvement in the estimation of cerebral activity compared with conventional GLM constructed without scalp hemodynamics model. These results suggest that the proposed method is an effective method for scalp hemodynamics reduction. Because the method allows us to measure from broad areas in standard commercial fNIRS, it may be useful in clinical practice.	P1-2-230 粒子フィルタを用いた神経細胞集団モデルのパラメータ推定手法構築 Simulation study for estimating the parameters of neural mass model using particle filter method ○小川雄太郎1, 小谷潔1, 神保泰彦1 ○Yutaro Ogawa1, Kiyoshi Kotani1, Yasuhiko Jimbo1 東京大学大学院　新領域創成科学研究科1 Graduate School of Frontier Sciences, The University of Tokyo, Japan1 EEG (Electroencephalogram) is used in various situations such as the diagnosis of the epilepsy, the sleep stage analysis, Brain computer Interface etc. One of a method for improving the EEG analysis is estimating the physiological parameters of the neural mass from EEG. In this study we develop the estimating method for neural mass physiological parameters from time series data using the particle filter method and evaluate the estimated accuracy in simulation case. Mimic EEG data are generated from NMM (Neural Mass Model), and the parameter of inhibitory synapse gain is estimated. We use the particle filter method which is a model estimation technique based on Monte Carlo method for recursively evaluating the probabilities using Bayes' rule. With the particle filter method, we evaluated the parameter value which reproduces the time series data with the highest likelihood. The parameter is estimated at less than 11% of the rate of an average error with 10 seconds data and the time variant case is also estimated well. We evaluate a case the model changes across the bifurcation point and the parameter is estimated well. These results indicate the NMM physiological parameter is estimated well with the particle filter method from short time series data in simulation study, and also indicate the possibility of this method for the real EEG data.
P1-2-231 マウス海馬スライスにおける様々な時空間活動パターンとその非線形解析 Various spatiotemporal activity patterns of mouse hippocampal slices and the nonlinear analysis ○浜崎雄太1, 小山内裕美1, 鈴木章義1, 斎藤稔1 ○Yuuta Hamasaki1, Hiromi Osanai1, Akiyoshi Suzuki1, Minoru Saito1 日大院・総合基礎科学・相関理化学1 Graduate School of Integrated Basic Sciences, Nihon University, Tokyo, Japan1 The brain forms a complex network composed of many neurons which communicate with one another. In addition, even a single neuron shows very complex activities, and some neurons spontaneously fire due to nonlinear characteristics. Recently, functional multineuron calcium imaging, which enables us to access brain function with single-neuron resolution, has been developed. In the present study, we observed spatiotemporal activity patterns of mouse hippocampal slices by a similar technique. The slices (350 μm) were prepared from 1-week-old male ddY mouse. The slice preparation was stained with a Ca2+-sensitive dye, oregon green. The stained slice was illuminated by an Ar laser (488 nm; 532-BS-A04, Melles Griot), and the 520 nm fluorescence images were acquired through a Nipkow confocal unit (CSU-10, Yokogawa) and a CCD camera (iXon X3 897, Andor). As a result, some dozens of neurons fired incoherently in some slices, while they exhibited a completely coherent activity pattern in other slices. The coherent pattern occurred more frequently under a higher K+ concentration and the existence of penicillin which are often used to induce the epilepsy-like state. Interestingly, we also observed the irregular switching between the coherent and incoherent activity patterns in other slices. The nonlinear analysis showed deterministic dynamics in these phenomena.	P1-2-232 チャコウラナメクジ嗅覚神経系に見られる時空間活動パターンとその非線形解析（II） Spatiotemporal Patterns of Neural Activities in the Olfactory Center of the Land Slug and the Nonlinear Analysis (II) ○石田康平1, 下川智也1, 浜崎雄太1 ○Kouhei Ishida1, Tomoya Shimokawa1, Yuuta Hamasaki1 日大院・総合基礎科学・相関理化学1 Graduate School of Integrated Basic Sciences, Nihon University, Tokyo, Japan1 We examined the odor responses of the oscillatory activity in the olfactory center (procerebrum; PC) of the land slug Limax valentianus by extracellular recording, and analyzed them by wavelet analysis. The local field potential of the PC showed an oscillation of about 1 Hz, and the oscillatory activity was changed by various odor stimuli to the tentacle. Prior to the odor stimuli, the wavelet energy was distributed into some frequency ranges. After the aversive odor stimuli, the wavelet energy concentrated mainly into two ranges (0.45-0.9 Hz, 0.9-2.4 Hz), and the wavelet entropy decreased. These results suggest that the activities of neurons in the PC became more coherent in response to the aversive odor stimuli. To confirm it, we examined the spatiotemporal patterns of neural activities in the PC by fluorescence voltage imaging technique. The PC preparation was stained with a voltage-sensitive dye, Di-4-ANEPPS. The stained preparation was illuminated by a LED (530 nm; LEX2-G, Brain Vision), and the 705 nm fluorescence images were acquired through a CCD camera (iXon X3 897, Andor). As a result, an oscillation of fluorescence intensity was observed in the PC. The oscillation had a phase delay along the distal-proximal axis, which was 0.04-0.1 π. After the aversive odor stimuli, the phase delay disappeared, which was 0-0.02 π. This result also shows more coherent activities of neurons in the PC.
P1-2-233 fNIRSによる事象関連運動時の脳活動計測の検討 Investigation of event-related functional near-infrared spectroscopy for a ballistic grasp movement ○小澤拓也1, 相原孝次2, 藤原祐介2, 大高洋平3,4, 南部功夫1, 和田安弘1, 大須理英子2, 井澤淳2 ○Takuya Ozawa1, Takatsugu Aihara2, Yusuke Fujiwara2, Yohei Otaka3,4, Isao Nambu1, Yasuhiro Wada1, Rieko Osu2, Jun Izawa2 長岡技科大院・工・電1, 慶大・医3, 東京湾岸リハビリテーション病院4 Dept Electrical Eng, Nagaoka Univ of Tech, Niigata, Japan1, ATR CNS, Kyoto, Japan2, Keio Univ, Tokyo, Japan3, Tokyo Bay Rehabilitation Hosp, Chiba, Japan4 It has been considered to difficult to measure single motor-related brain activities by functional near-infrared spectroscopy (fNIRS) which measures a concentration change of oxy/deoxy hemoglobin because fNIRS was much lower spatial resolution than functional magnetic resonance imaging (fMRI). To detect a change in the hemoglobin induced by motor movements by fNIRS, most of the previous studies utilized the "block design" paradigm in which a subject conducted a set of repetitive movements during over a second and changes of the hemoglobin induced by the series of movements were accumulated. In the present study, we address whether fNIRS can detect a phasic change induced by the single ballistic movement in an "event-related design" paradigm often adopted in fMRI experiments. In this end, we designed an event-related grasping task where the subject squeezes an elastic ball for a few hundred milliseconds after each randomly presented cue beep. To increase spatial resolution, we arranged the fNIRS channels more than double density used in conventional arrangement. Measured signals of each channel were fitted by a convolution of the grasping movements and a hemodynamics response function and statistically significant changes in oxy/deoxy hemoglobin were observed. To obtain a topographic functional map on cortical surface, the statistical values (t-values) were projected onto a cerebral cortex. As a result, high event related activities were found in a contralateral sensorimotor cortex. We also compared the topographic functional map of fNIRS with the functional map given by an event-related fMRI experiment when the same subjects performed exactly the same grasping task. We conclude that the fNIRS affords the opportunity to explore a motor related brain activity even for the single ballistic movement.	P1-2-234 マウスの連続的レバー操作と運動野ニューロン活動との相互情報量推定 On the estimation of mutual information between continuous lever trajectories and neuronal activities in mouse motor cortex ○田中康裕1,2, 正水芳人1,2, 松崎政紀1,2,3 ○Yasuhiro Tanaka1,2, Yoshito Masamizu1,2, Masanori Matsuzaki1,2,3 基生研・光脳回路1, 独立行政法人科学技術振興機構,CREST2, 総研大3 Div. Brain Circuit, NIBB, Okazaki1, JST, CREST, Saitama2, The Graduate University of Advanced Studies (Sokendai), Okazaki3 Estimation of information carried by neuronal ensemble is a fundamental interest in neuroscience. Various methods have been developed to estimate mutual information between discrete stimulus properties and spike count data. However, animals can receive continuous sensory inputs and execute continuous motor outputs. In addition, neuronal data can also have a continuous nature. With existing methods for these continuous variables, the optimal parameters for the estimation sometimes could not be determined unless the true mutual information is given. Here we propose a novel method for estimating mutual information between continuous variables via copula function, which describes dependency of a multivariable distribution without the effect of marginal distributions. The empirical joint distribution was transformed into the empirical copula frequency. Obtained copula functions were divided by bins with a width that was automatically determined as optimal with respect to Akaike's information criterion. This automatic model selection is an advantage of this method over k-nearest neighborhood (k-nn) method, one of the best methods for estimating mutual information between continuous variables. We first test the proposed method for bivariate Gaussian distributions with varying their covariance matrix. With moderate sample number (more than 100), the proposed method is more precise and faster than k-nn method. We also applied the proposed method to calculate mutual information between continuous lever trajectories manipulated by head-fixed mice and simultaneously recorded intracellular calcium transients in forelimb area of the motor cortex. Calcium transients, being seen as convolution of action potentials with some continuous function, were obtained by in vivo two-photon imaging. We found that mutual information of single neurons for lever trajectories was distributed in a gamma-like heavy-tailed fashion.
P1-2-235 機能的近赤外分光法で計測された視覚運動追従課題時の脳活動に対する頭皮血流変化成分除去の検討 Investigating effects on removal of the systemic interference for learning related functional near-infrared spectroscopic signals during visuomotor tracking ○南部功夫1, 今井貴弘1, 齊藤翔太1, 佐藤貴紀1, 和田安弘1 ○Isao Nambu1, Takahiro Imai1, Shota Saito1, Takanori Sato1, Yasuhiro Wada1 長岡技術科学大学　電気系1 Dept Electrical Engineering, Nagaoka Univ of Tech, Nagaoka, Japan1 Functional near-infrared spectroscopy (fNIRS) is one of the neuroimaging techniques to evaluate brain activity. Because it has advantages such as portability and simplicity of use, it can be applied to the measurement during movements. Previous studies showed a possibility that fNIRS activity reflects learning related changes of motor tasks. However, recently, it has shown that an influence of systemic interferences is largely observed in fNIRS signals. Because the interferences change in a task-related manner, it may obscure actual learning related activity, which usually shows small changes. In this study, we examined effects of removing the systemic artifacts from fNIRS activity during a motor learning task. We measured fNIRS activity for left hemisphere while the participant performed visuomotor tracking on a touchscreen with his right hand. To reduce an influence of the artifacts, a method using signals from short distance probes (short-distance channels) and general linear model (GLM) analysis was applied to the data for standard fNIRS signals (long-distance channels). In this method, a globally distributed systemic interference was extracted from short-distance channels and was incorporated into the GLM design matrix. We found that the extracted interference was included in most of the long-distance channels. Without this global artifact in the design matrix, different brain activity patterns were estimated. The result suggests that the removal of the artifacts is necessary for evaluating motor learning related brain activity. We further discussed individual activity patterns and behavioral performances.	P1-2-236 逐次ベイズ推定に基づくオーバーラップに頑健なリアルタイムスパイク検出法の多点電極アレイへの拡張 Real-time sequential Bayesian inference algorithm for detection of overlapped spikes extended to multi-electrode array ○芳賀達也1, 深山理1, 満渕邦彦1 ○Tatsuya Haga1, Osamu Fukayama1, Kunihiko Mabuchi1 東京大学　情報理工学系研究科　システム情報学専攻1 Graduate School of Information Science and Technology, The University of Tokyo, Tokyo1 Real-time detection and sorting of extracellularly recorded neural spike to obtain multi-unit spike trains have been fundamental technology for development of brain-machine interface systems and the experimental systems with real-time feedback in neuroscience. In this technology, overlaps of spike waveforms have been one of big issues because they hid original spike waveforms and seriously impaired the performance of spike detection by the predefined threshold and traditional template-matching. We have developed the algorithm based on probabilistic modeling and the sequential Bayesian inference to solve this issue and have shown that our algorithm could robustly detect and sort spikes from single-channel neural signals containing a lot of complex overlaps of spikes in real-time. In the presentation, we will present the extension of our method to develop the real-time spike detection system for multi-electrode arrays. We constructed the probabilistic models of recording with multi-electrode arrays and derived the efficient algorithm to estimate spike timings, spike waveforms, the standard deviation of the noise in real-time. We showed the feasibility of our system by applying it to simulated neural signal and extracellular recordings from primary cortical neurons and the brain of a rat.
P1-2-237 頭皮血流モデルを組み込んだ光拡散トモグラフィ : ヒト実験による検証 Diffuse optical tomography with the scalp blood volume model: human validation study ○山下宙人1, 下川丈明1, 愛須亮太1,3, 網田孝司2, 井上芳浩2, 佐藤雅昭1 ○Okito Yamashita1, Takeaki Shimokawa1, Ryota Aisu1,3, Takashi Amita2, Yoshihiro Inoue2, Masa-aki Sato1 国際電気通信基礎技術研究所　脳情報解析研究所1, 島津製作所　医用機器事業部技術部2, 奈良先端科学技術大学院大学 情報科学研究科3 Neural Information Analysis Laboratories, ATR1, Medical Systems Division Research and Development Department, Shimadzu Corporation2, Information Science, Nara Institute of Science and Technology3 Diffuse optical tomography (DOT) is an emerging technology to improve the spatial resolution of the scalp topography obtained by the multi-channel near infrared spectroscopy (NIRS). The DOT provides three dimensional reconstruction of cortical activities that is more interpretable and more quantitative measures of brain functions. Although the DOT is promising, the tomography obtained by the current standard DOT reconstruction algorithm, the minimum-norm method, is known to be blurred and superficially-biased. In addition, none of DOT algorithm proposed in the literature is ability to segregate the scalp blood volume, which is one of major artifact components in the NIRS measurement. In the previous reports, we have proposed the hierarchical Bayesian model to improve the spatial resolution and depth accuracy of the minimum-norm method as well as to remove the scalp blood volume accurately. The validity of our method to simulation studies, phantom studies and human experimental data of a single subject has been demonstrated. In this report, we apply the proposed method to human experimental data of twelve normal subjects to show validity of our method to the group of subjects. We measured brain activities during the right-index finger tapping and hand grasping tasks with fMRI and NIRS in two different days. We compared the spatial patterns and temporal patterns of our DOT with those of fMRI. Accurate reconstructions are obtained across subjects, particularly, for data during the hand grasping task, which is higher signal-to-noise ratio than that during the finger tapping task. This research was supported by the National Institute of Information and Communications Technology.

上部に戻る

前に戻る