 |
  |
|
|
|
11．生理研研究会「機能的MRI研究会」2003年11月27日−11月28日
【参加者名】 【概要】
（1） Simultaneous measurement of hemoglobin concentration change and BOLD signal response during visual stimulation by NIRS and fMRI豊田浩士（生理研究所，福井大学・高エネルギー医学研究センター） 健常者を対象に視覚刺激に対する hemodynamic responseをnear infrared spectroscopy (NIRS) とfunctional MRI (fMRI) の信号変化として同時計測し比較することを目的とした。視覚刺激として，8 Hzで反転するチェッカーボードパターンを20秒間の休止期間をはさんで1, 2, 4, 8秒間提示するsessionを繰り返した。NIRS装置のoptodeを後頭部に配置し，oxy-，deoxy-hemoglobin (Hb) 濃度を計測し，同時に3T-MRI装置にてBOLD信号を計測した。高分解能MRI上でoptodeペアの先端位置からNIRS信号源を推定し，比較するfMRIデータの関心領域とした。subject毎に，最大刺激応答の見られたNIRS信号受信チャネルからの信号とその信号源推定部位に対応するBOLD信号を，time-lockedで加算平均して比較した。oxy-, deoxy-, total-Hbの各濃度のtime courseにおける最大信号変化率および曲線下面積は刺激提示時間を1, 2, 4, 8秒と変化させることに対応していずれも増加するが，total-Hb濃度が比較的linearな増加傾向を示すのに対し，deoxy- Hb濃度の増加ではBOLD信号の場合に似た飽和傾向が見られた。
（2）脳賦活時のCBF, CMRO2, CBVのカップリング柏倉健一（福井大学・高エネルギー医学研究センター） 本研究は，MRI及び近赤外装置を用いて脳負活時のCBF, CMRO2, CBVの関係を非侵襲的に求める方法論を開発し，評価することを目的とした。 脳血流量と脳酸素代謝量の関係に関しては，以前より，脳血流量と脳酸素代謝量，脳ブドウ糖代謝量との間には，強いカップリング関係にあると考えられていた。これに対して，1980年代半ばにFoxらにより脳賦活時には脳血流量と脳酸素代謝量はuncouplingしているという報告がされた。この問題については現在まで議論が続いている。一方，脳血流量と脳血液量との関係に関しては，古くセントルイスのグループから脳血液量は脳血流量の関数で表されるという報告がされた。ただし，この報告はサルに対して二酸化炭素負荷を行うという手技で行われたものであり，ヒトで脳賦活時の脳血流量と脳血液量との関係を調べた報告はほとんどない。 本論では，CBFとBOLD，CBFとCMRO2，BOLDとCBVとの関係を求めることを目的とし，BOLD信号はMRIで，CBFはMRIのFAIR法で，CMRO2はMRIのFAIR法とBOLD信号との組み合わせから，さらにCBVは近赤外装置で求めることを試みた。 この結果，BOLD信号とFAIR信号の信号変化率の関係は，若干のばらつきはあるが，ほぼリニアに相関した。次に，CMRO2に関して，CBFとCMRO2との関係はほぼ直線関係にあることがわかった。さらに近赤外装置を用いて測定したDeoxy-Hbとoxy-Hb, total-Hbとの関係はきれいな直線関係になり，またBOLDとtotal-Hbとの関係もほぼリニアという結果になった。以上の結果は，脳賦活時の脳血流量，脳酸素代謝量，脳血液量の変化は相互に強いカップリング関係にあることを示唆したものと考える。
（3）Statistical analysis of functional near infrared spectroscopy time series -adjusting or modeling temporal autocorrelation-Takanori Kochiyama (Kagawa University) Near Infra-Red Spectroscopy (NIRS) has been used in functional activation studies as a non-invasive technique to detect changes in cerebral oxygenation in human adults and infants. By a high sampling rate and simultaneous multi channel measurements, NIRS instruments have the ability of revealing temporal and spatial structure of brain hemodynamic response, as the same as fMRI or PET. However, contrasted to the improvement of the measuring instrument and the wide application to research, there are not well-organized analytical methods compared with the other imaging technique at present. The purpose of our study is the establishment of a method to detect the functional brain activation with robustness and without losing sensitivity in NIRS measurement. In this reports, we focus on the topics about the modeling and correcting the temporal autocorrelation in NIRS time series data. We revealed that the autoregressive (AR) process of NIRS time series exhibited higher order and more complex form than that of fMRI time series. We examined some methods of the statistical signal processing techniques based on the generalized linear model (GLM), which were developed in fMRI time series analysis: (1) General least square estimation (GLS) of GLM with swamping approach, (2) GLS of GLM with whitening approach, (3) Variational bayesian estimation (VBE) of GLM. We evaluate these methods in terms of the ability to control the type-1 error (false-positive)rate and propose the appropriate methods and their parameter's settings for NIRS time series analysis.
（4）An Introduction to the Time Series Approach in fMRI Data AnalysisToru Ozaki (Institute of Statistical mathematics, Tokyo) We would like to present a time series approach for the analysis of fMRI data. We have been working on fMRI data analysis since 2000 when Prof. P. Valdes and I visited Prof. N. Sadato in his lab in Fukui Medical School. The most widely used standard method of fMRI data analysis is the SPM method developed by K. Friston and his group (Friston et al. (1995) ) . We think the SPM method does not fully exploit the dynamic information involved in fMRI data. In the present talk, we would like to show how useful spatio-temporal information, such as localization and connectivity, can be extracted from the data using our time series modeling approach with two types of experimental data: one is visual stimulus data from Prof. Sadato’s lab of National Institute of Physiological Sciences, and the other is motor task data from Prof. Kawashima’s lab in NICHe, Tohoku University.
（5）Evaluating effective connectivity: Multivariate time series approach Okito Yamashita (The graduate university for advanced studies) Integration within a distributed system is well understood in terms of 'effective connectivity'. 'Effective connectivity' is defined as "the influence that one neural system exerts over another" by K. Friston. In order to evaluate effective connectivity, several statistical methods have been suggested : structural equation model (McIntosh and Gonzalez 1994) , regression based analysis (Friston et.al. 1995) and so on. A disadvantage of these methods is that analysts need to prespecify the direction of influence. However this disadvantage can be overcome by causal inference based on multivariate time series approach (Harrison et.al. 2003) . We have suggested an extension of Harrison's method so that both direction and magnitude of effective connectivity can be evaluated. Akaike's relative power contribution is a key tool for evaluation. The result for the data of random dot stimulus will be presented. Reference: McIntosh A.R. and Gonzalez-Lima F. (1994) Structural equation modeling and its application to network analysis in functional brain imaging. Hum. Brain Mapp. 2: 2-22 Friston K. Ungerleider L.G. Jezzard P. and Turner R. (1995) Characterizing modulatory interactions between V1 and V2 in human cortex with fMRI. Hum. Brain Mapp. 2: 211-224 Harrison L. Penny W.D. and Friston K. (2003) Multivariate autoregressive modeling of fMRI time series. NeuroImage 19: 1477-1491
（6）Correction of residual motion effects using ICATakanori Kochiyama (Kagawa University) In functional MRI, movement-related effects are extant as residual motion effects even after perfect realignment (Friston et al. 1996) . As movements of subject's head or tissues are accompanied with task performance, the residual motion effects tend to exhibit the task-correlated features. In such a case, the existing methods, e.g. the voxel-by-voxel ANCOVA including the estimated motion parameters, fail to correct the task-correlated residual motion effects because of the violation of the orthogonality assumption between the task-related effects of interest and the motion-related effects of no interest. To be free from a temporally constraint for univariate time series analysis, we use independent component analysis (ICA). ICA is multivariate statistical technique, which can decompose fMRI data into spatially independent components (ICs) with an associated time course temporal-blindly. ICA could spatially separate neuronal activity-related and residual motion artifactual process into different ICs regardless of the temporal orthogonality among them if they have a mutually different spatial distribution. In this report, we propose the new procedure to automatically remove the task-correlated residual motion effects using ICA. In particular, to characterize the IC that represents the task-related motion effect, we adopt two criteria: (1) task-correlation, and (2) heteroscedasticity. Heteroscedasticity is the change in residual variance for a regression across serial observations, which is most likely to be induced by task-related motion. We demonstrate the efficiency of our proposed method by correcting the simulated data sets contaminated with task-correlated head motions.
（7）Hemodynamic, electrophysiological and structural interaction involved in human face processing. Evidence from a combined fMRI-ERP-VBM studyITetsuya Iidaka*, Atsushi Matsumoto*, Junpei Nogawa*, Tomohisa Okada**, Norihiro Sadato** (*Department of Psychology, Nagoya University, Graduate School of Environmental Studies, Nagoya, Japan, **Department of Cerebral Research, National Institute for Physiological Sciences, Okazaki, Japan) fMRI and ERP experiments were conducted in the same group of subjects and with an identical task paradigm to investigate the possible interaction of hemodynamic and electrophysiological responses. During the task the subjects judged whether visually presented stimuli were faces or houses. fMRI identified face- and house-related regions in the lateral and medial part of the fusiform gyrus, respectively, while ERP showed significantly greater N170 negativity in the temporo-occipital electrodes for face than for house. These results were consistent with the previous studies, and add evidence that the difference in N170 amplitude between the conditions is associated with a topographic difference in activation in the inferior temporal lobe. Correlation analysis between the BOLD signal and ERP parameter demonstrated that the subjects with a long N170 latency had greater activation in the fusiform gyrus than those with a short latency under the face condition. The magnitude of face- related activation in the fusiform gyrus correlated with the increase of N170 negativity. Furthermore, voxel-based morphometry (VBM) revealed that the grey matter density in the right fusiform gyrus had a positive correlation with the N170 latency for face. The widely distributed correlation between the signal and latency appears to be explained by the fact that the demanding cognitive process for face prolongs N170 latency and increases the signal in the fusiform regions. The present study suggests that an integrative analysis of spatial, temporal, and anatomical information regarding the brain and cognition can be achieved by a combined fMRI-ERP-VBM study.
（8）MRI撮像音強度の変化に伴う聴覚野血流反応時間の加齢性変化岡田知久，本田 学，定藤規弘（生理学研究所） これまで脳賦活による血流増加時間は加齢に伴い，視覚野ではより速くなるのに対して，感覚運動野では有意な変化はないと報告されているが，聴覚野に関しては明らかにされていない。昨年度われわれは本研究会で一次聴覚野は新たな刺激入力が生じた場合のみでなく，それが減少した場合にも，定常状態よりも大きな賦活があることを示した。この手法は常に一定の強さの聴覚刺激を正確なタイミングで与えることが出来る。これを用いて，53人の被験者（年齢：18-73歳）に対して撮像音強度を変化させる刺激を与えて，聴覚野における血流反応の加齢性変化を検討した。血流反応の大きさについては年齢による有意な増大・減少傾向は認められなかったが，血流反応が最大となる時間は加齢とともに延長していることが判明した。
（9）Application of MRI movie to kinematic analysis of articulatory and orofacial movementsShinobu Masaki, Yasuhiro Shimada, Ichiro Fujimoto (ATR Brain Activity Imaging Center) The purpose of installation of MRI scanner at ATR Brain activity Imaging Center (BAIC) was to explore not only functional brain imaging but also motor behavioral studies. The focus of this talk is to introduce how an MRI movie technique was established and has been applied for motor behavioral studies, especially in speech production researches. The technique used at BAIC is based on the synchronized sampling method which was originally invented to visualize heart movements. In the original method the gating pulse was generated from electrocardiac signal to synchronize MR scan to cardiac movement. In order to apply the method to observing articulatory movements, the pulses for MRI scanner and tone bursts for a subject were generated using two-channel signal generating software in a personal computer. During a data acquisition experiment, a subject repeats the spoken material so as to synchronize the presented tone bursts. Since the method allows us to reproduce three-dimensional moving images of speech organs, it has been used to accumulate data to investigate human speech production mechanisms. Currently, this technique is being applied to the field of rehabilitation, such as in investigating the mechanisms of pathological speech production and swallowing.
（10）FREQUENCY OF REVERSING CHECKERBOARDS AND THE BOLD SIGNAL IN HUMAN PRIMARY VISUAL CORTEX: A HIGH-RESOLUTION FMRI STUDY Pei Sun, Kenichi Ueno, R. Allen Waggoner, Keiji Tanaka, and Kang Cheng Previous PET and fMRI studies have shown that rCBF or BOLD signal in human primary visual cortex (V1) increases with the stimulus rate and peaks at ~8 Hz. To date, most of these studies were conducted using visual patterns with a fixed on-duration and variable intervals, where the averaged luminance was a function of stimulus rate (but see Fox and Raichle,1985). We have performed an fMRI study using white/black checkerboards (7.6° in diameter; checker size ~0.8°) with contrast reversal frequency of .05, .75, 2, 4, 8 and 16 Hz, but averaged luminance was kept constant across frequencies. For each scan, checkerboards reversed at a randomly selected frequency were presented in blocks alternated with baseline conditions (a homogeneous gray background) . The checkerboards were placed at 225°/315° orientation (7.6° from a central fixation cross) in the left or right lower visual field through a pair of fiber optic glasses. During fMRI scans, the subject detected the color change of fixation cross (eye positions were monitored). Experiments were conducted on a Varian 4T system with a quadrature surface coil and a segmented EPI pulse sequence (8 segments; volume TR, 4.6s; TE, 25ms). Six continuous slices (thickness, 3mm; in-plane resolution, .94x.94mm), parallel to the calcarine sulcus, were prescribed based on V1/V2 borders determined in a separate experiment, and covered the dorsal V1 of the targeted hemisphere. In all subjects and across all frequencies, circumscribed activations were observed in expected retinotopic loci within V1. There were no significant differences in activated voxel number and percent BOLD signal change between different frequencies. These results raise the question regarding the stimulus rate dependency, and suggest that at least in V1, at a population level, neurons preferring varied reversing frequencies are distributed roughly equally. Recently, in a series of high-resolution (in-plane resolution, .75x.75mm) study, it was revealed that whole-field checkerboards of high (15Hz) and low (.75Hz) reversal frequencies preferentially activated spatially segregated patches in V1. These preliminary results indicate that there may exist a distinct functional architecture in V1, where visual information of different temporal frequencies is represented.
（11）Contrast adaptation and the BOLD signal in primary visual cortexJ.L. Gardner, P. Sun, R.A. Waggoner, K. Ueno, K. Tanaka, K. Cheng Adaptation of neuronal responses is a ubiquitous property of visual cortical neurons. Adaptation of neuronal responses is thought to underlie the astounding ability of the brain to process sensory stimuli over many orders of magnitude of stimulus strength. Understanding adaptation processes is also critical for the proper interpretation of the BOLD signal because non-linearities of temporal summation of the BOLD signal are thought to be at least partially due to underlying adaptation of neuronal responses and not necessarily due to inherent nonlinearities of the coupling of neuronal responses to hemodynamics. With these motivations in mind, we have used BOLD imaging to examine the contrast-response functions of early visual cortical areas for several contrast adaptation states. We used an event-related paradigm in which we presented a 7.5Hz contrast-reversing checkerboard stimulus at a single contrast (base contrast) . Every 8-12 seconds we would increase or decrease the contrast of the checkerboard for a brief period (3 seconds) . The changes in the BOLD signal in early visual cortical areas to these brief changes in contrast were measured for three different base contrast levels. From these measurements we constructed contrast-response curves for the BOLD signal. We have found evidence that similar to cat primary visual cortex, humans exhibit adaptation of contrast response functions as primarily horizontal shifts of contrast response curves. These horizontal shifts of contrast response curves show that early cortical visual response in humans displays contrast response gain changes with adaptation, a process that is thought to underly an adaptive mechanism allowing the brain to efficiently encode and process stimuli with various average contrast levels.
（12）機能的光トポグラフィーによる乳児視覚野の反応の経時的変化 中井昭夫，眞弓光文 （福井大学医学部 小児科） 我々は以前fMRIにて視覚野のHbの反応が生後60日で逆転することを報告したが，fMRIはBOLD効果をみているに過ぎない。一方，近赤外線分光法(NIRS) は酸化，還元Hbを各々測定でき，時間分解能も高い。今回NIRSを用いた光トポグラフィーにて乳児視覚野の反応について検討した。対象は保護者から同意を得られた正常満期産児。自然睡眠中8Hz Flashによる5秒ON，15秒OFFの光刺激に対する酸化，還元Hbを日立メディカルETG-100にて生後1ヶ月，3ヶ月，7-9ヶ月と経時的に測定した。結果，光刺激に対し酸化Hbが増加，還元Hbが減少し，このパターンは生後1ヶ月，3ヶ月，7-9ヶ月とも同様であった。以前のfMRI研究は薬物による鎮静下で，そのパラダイムは30秒ON，30秒OFFであった。今後更に例数を増やすとともに，様々な刺激時間によるfMRIとの同時測定などの検討が必要と考えられる。
（13）Model-free nonlinear regression analysis in fMRI using an artificial neural network Masaya Misaki, Satoru Miyauchi We developed a new method for analyzing fMRI data using an artificial neural network. In most fMRI studies, searches are made looking for a correlation between certain events/tasks and change in fMRI signal. It is known that change in fMRI signals nonlinearly relates to events. We employed a three-layered back-propagation neural network, given a series of events as inputs and an fMRI signal as an ideal output, for performing a nonlinear regression analysis. Although this method is almost equivalent to the analysis using the Volterra series, it is more flexible and model-free because no assumptions about the shape of kernel function and its order are needed. The method can detect any correlations between events and fMRI signals even when an unknown response mediates those correlations. The example results indicated that our model-free nonlinear regression analysis has advantages of flexibility and detectability, but has also weakness for false positive errors. Thus, this method is not just right for a substitute of other model-based methods like the general linear model, rather it is suitable for a more exploratory analysis such as to detect unknown brain activities or hemodynamic responses which are not a priori expected.
|
||||
