Tejaswini Kavallappa¹, Steven Roys², Anindya Roy³, Joel Greenspan², Rao Gullapalli², and Alan McMillan^2
1Dept. of Nuclear Medicine and Diagnostic Radiology, Univeristy of Maryland School of Medicine, Baltimore, MD, United States, ²University of Maryland School of Medicine, ³University of Maryland Baltimore County

The reliability of functional and effective connectivity (using structural equation modeling [SEM]) was examined. Reliability assessment was performed on four datasets that were subject to various types of physiological and mean brain signal filtering. Path coefficients in effective connectivity demonstrated a higher degree of variability compared to correlation coefficients in functional connectivity, regardless of filtering method used. The high variability from test-retest results from SEM analysis suggests caution when interpreting results from such analysis.

Yi Chia Li¹, and Jyh Horng Chen^2
1Graduate Institute of Biological Engineering and Bioinformatics, National Taiwan University, Taipei, Taiwan, ²Interdisciplinary MRI/MRS Lab, Department of Electrical Engineering, National Taiwan University, Taipei, Taiwan

Up to 10 functional networks contributed by low frequency fluctuations (LFFs) have been reliably identified to consistently exist in human resting brains. These networks consist of regions which are known to be involved in function of motor, vision, execution, auditory, pain perception, language, cerebellum, and the so called default-mode network (DMN). In our present work, we analyzed resting-state fMRI data of 11 healthy participants to further investigate functional networks which consistently exist in resting brains. The functional networks obtained in our work largely corresponded to the findings in prior literatures. Additionally, we discovered two new functional networks: spatial cognition network and facial sensory network. Spatial cognition network consisted predominantly of superior and inferior parietal gyrus (BA 7/40), which were crucial in visuo-spatial processing during cognition-Chinese-language paradigms (reading and writing). Facial sensory network covered pons and medial temporal pole, which served to process sensory information from human faces such as the sense of smell and taste.

Tuomo Starck^1,2, Juuso Nissilä³, Antti Aunio³, Ahmed Abou Elseoud^1,2, Jukka Remes¹, Juha Nikkinen¹, Markku Timonen^4,5, Timo Takala⁶, Osmo Tervonen^1,2, and Vesa Kiviniemi^1,2
1Diagnostic Radiology, Oulu University Hospital, Oulu, Finland, ²Diagnostic Radiology, Oulu University, Oulu, Finland, ³Valkee Ltd, Finland, ⁴Department of Psychiatry, Oulu University, Finland,⁵Institute of Health Sciences, Oulu University, Finland, ⁶ODL Health ltd, Oulu, Finland

Based on literature about opsin proteins and anecdotal evidence of inherent light-sensitivity of the human brain tissue we tested the hypothesis that brain activity would alter during bright light stimulation via ear canal. ICA dual regression analysis was performed for full band resting state BOLD fMRI data between constant light stimulus (n=24) and sham controls (n=26). Lateral visual IC was significantly different between the groups, light stimulus subjects demonstrated slowly increasing activity around the visual cortex and related regions. Results suggest brain tissue to be inherently light sensitive.

Elizabeth Quattrocki Knight^1,2, Xiaoying Fan³, Blaise Frederick⁴, Marc Kaufman⁴, and Bruce Cohen^2,3
1Psychiatry, McLean Hospital, Belmont, MA, United States, ²Psychiatry, Harvard Medical School, Boston, MA, United States, ³Frazier Research Institute, McLean Hospital, Belmont, MA, United States, ⁴Brain Imaging Center, McLean Hospital, Belmont, MA, United States

Although numerous fMRI studies have examined visual processing, less work has focused on the auditory system. With the exception of sparse sampling techniques, interference from scanner noise can impede the study of auditory processing. Independent component analysis (ICA), by isolating and removing components in the data representing extraneous sources of noise, can facilitate fMRI data analysis. Here, we compare results of a self-organizing group level ICA (SogICA) to a random effects (RFX) general linear model in an auditory listening study. SogICA identifies not only more extensive task-related activity, but also reveals underlying resting state networks.

Yasuhiro Funakoshi¹, Tomomi Sumiyoshi², Masafumi Harada³, and Hitoshi Kubo^3
1Medidcal Imaging, University of Tokushima, Tokushima, Tokushima, Japan, ²University of Tokushima, ³Health Biosciences, University of Tokushima

The default mode neural network would be interfered or localized in the smaller area by the stimulation from the outside and subject’s attention. In the case of clinical application to patients, the stimulation from the outside should be removed and it is considered that the psychological resting state is important to apply this technique for the clinical diagnosis.

Tzu-chen Yeh^1,2, Chou-ming Cheng¹, Bi-yu Hsu¹, and Jo-mei Huang^2
1Department of Medical Research and Education, Taipei Veterans General Hospital, Taipei, Taiwan, ²Insitute of Brain Science, National Yang-Ming University, Taipei, Taiwan

Prehension is defined as the capacity to reach and grasp which involves complex neuro-coginitive architectures. Hand prehension is composed of selection of grasp model and transformation of motor command. Wide network of motor hierarchy involves action initiation of parietal cortices, premotor, supplementary motor area and primary motor cortex. A visuo-motor flanker task was implemented for the fMRI study to validate the prehension control model via dorsal stream. By lag correlation of the prehension-related components derived from spatial independent component analysis, the functional network of hand prehension echoed the theoretical construct and demonstrated major and minor connectivity of functional correlates.

Xinyuan Miao¹, Thomas Zeffiro², and Yan Zhuo^1
1Institute of Biophysics, Chinese Academy of Sciences, Beijing, China, People's Republic of, ²Neural Systems Group, Massachusetts General Hospital, United States

In this preliminary study, we used ROI-based functional connectivity of resting-state functional MRI to investigate changes in inter-regional correlations in women’s different menstrual phases. Our results showed that the functional connectivity of the left hippocampus and parahippocampus with the bilateral superior occipital gyrus and cuneus, and the right middle frontal gyrus was higher in the early follicular phase than in the mid-luteal phase. The patterns of functional connectivity shown in this study may provide new clues for understanding the mechanism of how spatial abilities are modulated by hormone during menstrual cycle.

Jingyuan Chen¹, Catie Chang², Kui Ying¹, Yan Zhu¹, and Gary Glover^2
1Tsinghua University, Beijing, Beijing, China, People's Republic of, ²Stanford University, Stanford, CA, United States

In our study, we used both cluster analysis and marginal/partial correlation analysis to quantify and compare how low-frequency temporal connectivity of the brain default mode network (DMN) changes during sustained tasks that activate and deactivate major regions in the network. We found that low-frequency temporal connectivity was not extinguished but attenuated within most major regions of DMN under tasks that deactivate its nodes relative to rest, and that more prefrontal regions were engaged in the network under such task modulation. Moreover, we noticed the persistence of low-frequency temporal connectivity in subjects whose DMN was activated by external task.

Robert L Barry^1,2, Stephen C Strother^3,4, and John C Gore^1,2
1Vanderbilt University Institute of Imaging Science, Nashville, TN, United States, ²Department of Radiology and Radiological Sciences, Vanderbilt University Medical Center, Nashville, TN, United States, ³Rotman Research Institute, Baycrest, Toronto, ON, Canada, ⁴Department of Medical Biophysics, University of Toronto, Toronto, ON, Canada

A challenge with ultra-high-field fMRI is the predominance of noise associated with physiological processes unrelated to tasks of interest. This degradation in data quality may be reversed using post-acquisition algorithms designed to estimate and remove the effects of these noise sources. BOLD fMRI data acquired using 2D EPI and 3D PRESTO at 7T were processed using the Stockwell transform filter, retrospective image correction (RETROICOR), and phase regression. Data quality was evaluated via metrics of prediction and reproducibility using NPAIRS. Results demonstrate the pseudo-complementation of these algorithms and maximization of prediction and reproducibility through synergistic interactions between RETROICOR and phase regression.

Gigi Galiana¹, and Robert Todd Constable^1
1Diagnostic Radiology, Yale University, New Haven, CT, United States

Significant improvements have been realized in the reconstruction of undersampled MR angiography data simply by switching the order of the subtraction and the unfolding steps. We report a similar improvement in fMRI reconstructions simply by switching the order of the analogous steps in fMRI processing. Rather than reconstructing the individual time point images and analyzing those images for activation, we perform the GLM analysis on the undersampled k-space. The sparse activation images are then easier to unfold, allowing for better quality reconstructions at high acceleration factors.

Dietmar Cordes¹, Mingwu Jin¹, Tim Curran², and Rajesh Nandy^3
1C-TRIC and Dept. of Radiology, University of Colorado-Denver, Aurora, CO, United States, ²Dept. of Psychology and Neuroscience, University of Colorado-Boulder, Boulder, CO, United States,³Depts. of Biostatistic and Psychology, University of California-Los Angeles, Los Angeles, CA, United States

An improved method to detect activations in fMRI uses local canonical correlation analysis (CCA) to encompass a group of voxels in a 3x3 pixel neighborhood. It is customary to assign the value of the test statistic to the center voxel of the local neighborhood. However, without spatial constraints such an assignment introduces smoothing artifacts in regions of strong localized activation, which we refer to as “bleeding artifacts”. To reduce this artifact we propose different spatial constraints in CCA to enforce dominance of the center voxel and introduce a method based on mixture modeling to further reduce this artifact.

Mingwu Jin¹, Rajesh Nandy², and Dietmar Cordes^1
1University of Colorada Denver, Aurora, CO, United States, ²UCLA, Los Angeles, CA, United States

In previous work, local constrained canonical correlation analysis (cCCA) methods were proposed in order to avoid model overfitting and loss of specificity. In this work, we further investigate the performance, efficiency and possible improvement of region-growing based cCCA (cCCA-RG) methods. Using simulated data, we compare the estimation power of different cCCA-RG methods as well as the exhaustive search method (cCCA-ES). The detection power is also investigated upon real fMRI data. Our results demonstrate that cCCA-RG can significantly improve the detection power within an acceptable period of computation time.

Rajesh Ranjan Nandy^1
1Psychology and Biostatistics, University of California, Los Angeles, CA, United States

Local canonical correlation analysis (CCA) is a multivariate method that simultaneously analyzes the timecourses of a group of neighboring voxels and is more sensitive than the conventional univariate GLM approach. However, unlike the general linear model (GLM), an arbitrary linear contrast of the temporal regressors has not been so far incorporated in the CCA formalism. To address the first problem, a multivariate regression model is presented. Multivariate regression model is equivalent to CCA, but easier to interpret. Arbitrary contrasts can be used in the multivariate regression model (MRM) approach including contrasts on voxels which is impossible in the univariate framework.

Veronika Schöpf^1,2, Christian Windischberger^1,2, Simon Robinson^1,3, Christian Kasess^1,4, Florian Ph.S. Fischmeister^1,5, Rupert Lanzenberger⁴, Jessica Albrecht⁶, Anna M Kleemann⁶, Rainer Kopietz⁶, Martin Wiesmann^6,7, and Ewald Moser^1,2
1MR Centre of Excellence, Medical University Vienna, Vienna, Vienna, Austria, ²Center of Medical Physics and Biomedical Engineering, Medical University Vienna, Vienna, Vienna, Austria,³Department of Radiology, Division of Neuroradiology, Medical University Vienna, Vienna, Vienna, Austria, ⁴Division of Biological Psychiatry, Department of Psychiatry and Psychotherapy, Medical University Vi, Vienna, Vienna, Austria, ⁵Faculty of Psychology, University of Vienna, Vienna, Vienna, Austria, ⁶Department of Neuroradiology, Ludwig-Maximilians-University, Munich, Munich, Germany, ⁷Department of Neuroradiology, Technical University Aachen RWTH, Aachen, Germany

In this study we were able to show that the new ICA method FENICA reliably identifies activation in a wide variety of paradigms and stimuli types. Activation maps are in excellent agreement with those established in previous, model-based analyses. FENICA has the potential to become a valuable tool for group fMRI studies, eliminating a priori assumptions including model and HRF, and without the need to downsample data in large studies, define spatial templates or manually identify single-subjector group components. Using FENICA it is possible to analyze functional MRI data of experiments using complex stimulus design involving different modalities on a truly data-driven basis. FENICA is a single-subject based technique allowing for group statistics to be applied in a well-established framework and provides a truly exploratory, data-driven, operator independent and therefore unbiased way of identifying common patterns of activation.

Radu Mutihac^1,2, Elena Ackley¹, Jochen Rick³, Akio Yoshimoto⁴, Maxim Zaitsev³, Oliver Speck⁵, and Stefan Posse^1,6
1Department of Neurology, University of New Mexico, Albuquerque, New Mexico, United States, ²Department of Electricity & Biophysics, University of Bucharest, Bucharest, Romania,³Department of Radiology - Medical Physics, University Medical Center Freiburg, Freiburg, Germany, ⁴Polytechnic Institute of New York University, New York, New York, United States,⁵Department Biomedical Magnetic Resonance, Otto-von-Guericke-University Magdeburg, Magdeburg, Germany, ⁶Department of Physics and Astronomy, University of New Mexico, Albuquerque, New Mexico, United States

Whole brain multiple-slab echo-volumar-imaging (EVI) is a novel methodology that provides up to an order of magnitude higher temporal resolution compared to multi-slice echo-planar imaging (EPI). However, fMRI sensitivity of EVI and EPI has not yet been systematically compared using neither hypothesis driven inferential statistics like statistical parametric mapping (SPM) nor exploratory methods like spatial independent component analysis (ICA) or temporal fuzzy clustering analysis (FCA). In this study, we statistically assess the extent and maximum T-score of activation elicited by an auditory-gated visual-motor task for both modalities using SPM8. Furthermore, the finer time course information available in EVI lends itself to data-driven analysis to identify physiological noise sources and spurious activation investigated by spatial ICA.

Jolinda Carol Smith¹, and Scott H Frey^1,2
1Lewis Center for Neuroimaging, University of Oregon, Eugene, OR, United States, ²Department of Psychology, University of Oregon, Eugene, OR, United States

Regions of interest (ROIs) are frequently used in fMRI. When defining functional ROIs, investigators face a number of arbitrary choices concerning the statistical threshold to employ and the method for delineating ROI boundaries. We propose a method for defining ROIs using independent component analysis (ICA). This method avoids many of the shortcomings of general linear model based ROI definition, and is robust and easy to implement. As a demonstration, we apply this method to define ROIs in the cortex and cerebellum that respond selectively to aurally paced movements of the lips, hands, and feet.

Samir D Sharma¹, Bosco S Tjan², and Krishna S Nayak^1
1Electrical Engineering, University of Southern California, Los Angeles, CA, United States, ²Psychology, University of Southern California, Los Angeles, CA, United States

Functional magnetic resonance imaging (fMRI) with blood-oxygenation-level-dependent (BOLD) signal is fundamentally limited by the time required to acquire each volume. Standard EPI sequences with statistically advantageous TRs of 1-2s are typically restricted to acquiring no more than 16-32 slices with a typical 3 mm³ isotropic resolution. This covers only about 40-80% of the cerebral cortex. A fast imaging technique is needed for full volumetric coverage at short TRs. This work proposes a simple and fast one-step thresholding (OST) algorithm for the detection of BOLD signal activation. Results from the proposed method at 2x-acceleration demonstrate close agreement with the fully-sampled reference.

Kayako Matsuo¹, Annabel S.-H. Chen², and Wen-Yih Isaac Tseng^1
1Center for Optoelectronic Biomedicine, National Taiwan University College of Medicine, Taipei, Taipei, Taiwan, ²Division of Psychology, School of Humanities and Social Sciences, Nanyang Technological University, Singapore

Laterality index (LI) is often applied in functional magnetic resonance imaging (fMRI) studies to determine functional hemispheric lateralization. However, conventional LI methods could suffer from an outlier bias if t-values were extreme for a small number of voxels. We developed an improved method to calculate LI for fMRI called AveLI. This method considers laterality of activation at each and every t-value threshold for the task (thus this index is not limited by over stringent thresholds), and allows an intuitive comprehension of overall asymmetry. We examined the feasibility and reproducibility of AveLI by applying it to two fMRI language tasks.

Scott James Peltier^1,2, Marc G Berman³, Yash Shah², Stephen Kaplan³, and John Jonides^4
1Functional MRI Laboratory, University of Michigan, Ann Arbor, MI, United States, ²Biomedical Engineering, University of Michigan, Ann Arbor, MI, United States, ³Psychology, University of Michigan, Ann Arbor, MI, United States, ⁴Psychology, University of Michigan, Ann Arbor, MI

Multivariate analysis offers an alternative way to analyze functional MRI data, and allows applications such as biofeedback. In this study, we use support vector machines to analyze subjects viewing natural or urban scenes. We demonstrate high accuracy, even when using minimal amount of data.

Robert Cary Welsh¹, Laura Jelsone-Swain¹, Veronika Schoepf², and Scott J Peltier^3
1Radiology, University of Michigan, Ann Arbor, MI, United States, ²Radiology, Division of Neuroradiology, Medical University of Vienna, Vienna, Austria, ³Functional MRI Laboratory, University of Michigan, Ann Arbor, MI, United States

We used fMRI data in conjunction with support vector machine (SVM) classification of brain state to examine (fMRI-based) brain-computer interface stability in the amyotrophic lateral sclerosis (ALS) patient population.

Rainer Boegle^1,2, Carolin Cyran³, Stefan Glasauer^1,2, and Marianne Dieterich^2,3
1Center for Sensorimotor Research, Ludwig-Maximilians University, Munich, Germany, ²Integrated Center for Research and Treatment of Vertigo, Ludwig-Maximilians University (IFBLMU), Munich, Germany, ³Department of Neurology, Ludwig-Maximilians-University, Munich, Germany

Nowadays it is assumed that most brain functions involve a network of cortical areas. Thus it should be anticipated that the temporal dynamics of the task-related BOLD responses of the respective areas may vary. Previous studies used spatio-temporal support vector classification to produce functional brain maps revealing responses discriminating between tasks (''discrimination maps''), without having to assume a model for the task related BOLD response. Building on these we present a method for determining the class-wise contributions to these brain maps which can reveal the class(task)-wise similarities in addition to the class(task)-wise differences revealed by the ''discrimination maps''.

An Vo¹, Aziz M. Ulug^1,2, E Kozora^3,4, G Ramon⁵, J Vega⁵, R D Zimmerman⁶, D Erkan⁵, and M D Lockshin^5
1The Feinstein Institute for Medical Research, Manhasset, New York, United States, ²Department of Radiology, Albert Einstein School of Medicine, Bronx, New York, United States, ³National Jewish Health, Denver, Colorado, United States, ⁴University of Colorado Medical Center, Denver, Colorado, United States, ⁵Hospital for Special Surgery, New York, New York, United States,⁶Weill Medical College of Cornell University, New York, New York, United States

Most patients with systemic lupus erythematosus (SLE) have cognitive dysfunction suggesting but clinically important central nervous system involvement. Antiphospholipid syndrome is another autoimmune disorder defined as the presence of arterial or venous thromboses and/or pregnancy morbidity with persistent antiphospholipid antibodies (APL). Diffusion tensor imaging (DTI) and fMRI has been used to study the SLE and APL patients. The purpose of this study is to use both fMRI and DTI features to automatically classify SLEs, APLs and normal controls.

Jeremy Lecoeur¹, Feng Wang², Li Min Chen², Benoit M. Dawant¹, and Malcolm J. Avison^2
1Electrical Engineering and Computer Science, Vanderbilt University, Nashville, Tennessee, United States, ²Radiology and Radiological Science, Vanderbilt University Medical Center, Nashville, Tennessee, United States

A pipeline for the automated coregistration of sub-millimeter resolution functional maps from brain sub-volumes was implemented and tested in a non human primate model of focal somatosensory cortical activation. Preliminary studies demonstrate an accuracy of coregistration of about 30 micrometers for structural and 100 micrometers for functional maps across separate sessions.

Adam J Schwarz¹, Vesa Kiviniemi², Sara de Simoni³, Steven CR Williams³, and Mitul A Mehta^3
1Translational Medicine, Eli Lilly and Company, Indianapolis, IN, United States, ²Diagnostic Radiology, Oulu University Hospital, Oulu, Finland, ³Centre for Neuroimaging Sciences, Institute of Psychiatry, London, United Kingdom

PhMRI responses are often widespread, but the stability of the spatial localization of responses across subjects and cohorts at the voxel scale may be affected by neuroanatomical variation, motion and residual differences in spatial normalization. We show that the phMRI response to ketamine can be accurately and sensitively modeled as a linear superposition of stable, independently derived, functional units. The concept is illustrated using a high model order ICA segmentation of the brain. Such functional units can be spatially distributed and overlapping, unlike anatomical VOIs, and may be robust to high spatial frequency differences across subjects.

zikuan chen¹, Arvind Caprihan¹, and Vince Calhoun^1,2
1Mind Research Network, Albuquerque, NM, United States, ²Electrical and Computer Engineering, University of New Mexico, Albuquerque, NM, United States

In BOLD fMRI, the BOLD activity can be delineated in terms of susceptibility map reconstructed from BOLD complex image, which is a 3D ill-posed inverse problem involving 3D deconvolution and denosing. In this work, we report a solution by the split Bregman algorithm of total variation (TV) regularization, which is an iterative regularization (implemented by a 3-subproblems iteration) for image restoration from noisy blurred image. Numerical simulation and phantom experiment show that this novel TV technique outperforms the filter-truncated Fourier inverse solution.

Yongquan Ye¹, Jiani Hu¹, Jie Yang¹, and Mark Haacke^1
1Radiology, WSU, Detroit, MI, United States

The novel single shot mHASTE BOLD fMRI technique was tested using a more subtle stimuli pattern than the flashing checkerboard, i.e. the MT/V5 activation, which induced highly localized activation with medium level significance. mHASTE was demonstrated to be sensitive to the BOLD response in MT/V5 area with significantly improved functional specificity than EPI methods, indicating that mHASTE is mainly sensitive to micro-vasculature BOLD signals.

Pål Erik Goa^1,2, Peter Jan Koopmans^2,3, Benedikt Andreas Poser^2,3, Markus Barth^2,3, and David Gordon Norris^2,3
1Department of Medical Imaging, St.Olav University Hospital, Trondheim, Norway, ²Erwin L. Hahn Institute for Magnetic Resonance Imaging, University Duisburg-Essen, Essen, Germany,³Donders Institute for Brain, Cognition and Behaviour, Radboud University Nijmegen, Nijmegen, Netherlands

Non-balanced Steady-state free precession (nb-SSFP) might be an alternative for BOLD-fMRI at 7T. Here we compare the functional signal change at the pial surface and within grey matter for S1 and S2 using multi-echo nb-SSFP at high spatial resolution. Eight subjects were scanned using a checkerboard paradigm, and tissue-specific signal changes were extracted based on segmentation of MP-RAGE images. Results show that the signal change is largest at the pial surface in S1 at both short and long TE as well as in S2. This might indicate intravascular blood is contributing to the functional contrast also at 7T.

Steven Andrew Patterson^1,2, Steven Donald Beyea^1,3, and Chris Van Bowen^1,3
1Institute for Biodiagnostics (Atlantic), National Research Council Canada, Halifax, Nova Scotia, Canada, ²Physics, Dalhousie University, Halifax, Nova Scotia, Canada, ³Physics, Biomedical Engineering and Radiology, Dalhousie University, Halifax, Nova Scotia, Canada

To achieve artifact-free whole brain coverage and good temporal resolution using passband balanced SSFP (pbSSFP) functional MRI (fMRI), alternating between two steady states to eliminate banding artifact is necessary. Monte Carlo simulations of alternating steady state pbSSFP (altSSFP) fMRI were conduced to characterize BOLD signal and contrast. Results suggest altSSFP can provide artifact-free whole brain coverage with 2-3 s temporal resolution and up to 90% of the BOLD contrast observed in conventional pbSSFP if high (45-60°) flip angles and linearly increasing flip angle RF catalyzation is used. Despite off-resonance signal intensity variation, spatially-uniform sensitivity fMRI maps are anticipated.

Yu-Wei Tang¹, and Teng-Yi Huang^1
1Electrical Engineering, National Taiwan University of Science and Technology, Taipei, Taiwan

To calibrate the sensitivity-band of transition-band SSFP fMRI, a SSFP angle adjustment method based on a sweep scan with increment of SSFP angle has been proved its potential in previous studies. In this study, we proposed a real-time feedback optimization method to calibrate the SSFP angle automatically and rapidly. Through network connection of MRI scanner and an external PC, a feedback loop was created. An optimization method is proposed to search the optimal SSFP angle automatically. The results demonstrate that it exhibits similar effectiveness of the previous sweep scan method and requires much shorter calibration time.

Ute Goerke¹, and Kamil Ugurbil^1
1Radiology, Center for Magnetic Resonance Research, Minneapolis, Minnesota, United States

In recent research efforts, fMRI contrast in steady-state sequence has been investigated. Potential contributions to the fMRI signal changes in these sequences, in which a large fraction of the imaging time is used for excitation pulse, include magnetization transfer effects and/or relaxation during RF pulses. In this paper, we propose a novel approach using non-slice selective spoiled 3D GRE with chirp pulse excitation and multiple echo acquisition. Activation maps are obtained from the undersampled fMRI signal changes using the spectral side band analysis (SSBA). Initial results indicate differences in amplitude of the fMRI signal change depending on pulse bandwidth.

Cungeng Yang¹, and Victor Andrew Stenger^1
1University of Hawaii, Honolulu, Hawaii, United States

Signal loss caused by susceptibility induced intravoxel dephasing is a major limitation in high field MRI applications such as BOLD fMRI. Spectral-spatial (SPSP) pulses have been shown to be very effective at reducing through-plane signal loss in axial slices using a single excitation. SPSP pulse design assumes a linear relationship between off-resonance frequency and through-plane susceptibility gradient Gs(f)=αf. Previous studies show empirically α=-2.0 μT/m/Hz works well for many brain regions at 3T, however, no detailed measurement of α was investigated. We propose spectral decomposition technique using a spiral spectroscopic imaging sequence to directly measure α at all locations in the brain. Resultant SPSP pulses were demonstrated in T2*-weighted brain images showing reduced signal loss at 3T. Inferior slices were found to require α values of opposite sign and smaller magnitude. This indicates that using more than one pulse may improve the efficacy of the SPSP technique.

Lars Kasper^1,2, Maximilian Häberlin¹, Christoph Barmet¹, Bertram Jakob Wilm¹, Christian C. Ruff^2,3, Klaas Enno Stephan^2,3, and Klaas Paul Prüssmann^1
1University and ETH Zurich, Institute for Biomedical Engineering, Zurich, Zurich, Switzerland, ²University of Zurich, Laboratory for Social and Neural Systems Research, Zurich, Zurich, Switzerland, ³University College of London, Wellcome Trust Centre for Neuroimaging, London, London, United Kingdom

Filtering is a common post-processing step in MRI. Specifically, the analysis of fMRI data frequently includes a Gaussian smoothing of the raw images. This application of a “matched filter” improves sensitivity at the spatial scale of the BOLD response. We show that the argument already holds at the acquisition stage and propose a 2D gradient-velocity modulated EPI sequence providing a Gaussian k-space density weighting. In the case of phantom and resting state fMRI, temporal SNR is thus raised by 60-80% while preserving overall measurement duration. Furthermore, for visual stimulation, t-contrast images reveal both increases in cluster sizes and peak t-values.

Gary Chiaray Lee¹, Caroline Jordan², Carlos Ruiz³, Pamela Tiet³, Brian Hargreaves², Ben Inglis⁴, and Steven Conolly^1
1Berkeley/UCSF Bioengineering Joint Graduate Group, Berkeley, CA, United States, ²Radiology, Stanford University, ³Bioengineering, UC Berkeley, Berkeley, CA, United States, ⁴Helen Wills Neuroscience Institute, Berkeley, CA

One difficulty with current EPI BOLD fMRI is poor T₂^*-weighted analysis near areas of strong field inhomogeneities. The magnetic susceptibility difference between air and tissue produces B₀ field perturbations in the head, and may cause significant MRI artifacts, expecially for EPI. Here we develop in vivo susceptibility matching pyrolytic graphite foams for improving EPI near external air/tissue interfaces. We have verified these that foams are safe for patient use, and here we demonstrate that the foams can make partial Fourier acquisition techniques more robust, which can ultimately lead to faster acquisitions or higher resolutions for EPI BOLD fMRI.

Juliane Budde¹, Frank Mühlbauer¹, G. Shajan¹, Maxim Zaitsev², and Rolf Pohmann^1
1Max Planck Institute for Biological Cybernetics, Tuebingen, Germany, ²University Hospital Freiburg, Freiburg, Germany

EPI data in humans were acquired at 9.4 T, using a GRE EPI sequence with distortion correction, resulting in 1.1 mm isotropic voxels. A simple finger tapping paradigm was employed for functional testing. In addition, high resolution (0.2 mm x 0.2 mm x 1.1 mm) phase and susceptibility-weighted images were acquired and co-registered to the functional data. Highly significant BOLD response was found in the left motor cortex, mostly located within gray matter, with activation amplitudes up to 20%. Co-registration to susceptibility-weighted data shows strong signal contributions even from vein-free regions.

Sheeba Arnold¹, Susan Whitfield-Gabrieli², Steven Shannon¹, John DE Gabrieli², and Christina Triantafyllou^1,3
1A.A. Martinos Imaging Center, McGovern Institute for Brain Research, MIT, Cambridge, MA, United States, ²Deparment of Brain and Cognitive Sciences, Cambridge, MA, United States, ³A.A. Martinos Center for Biomedical Imaging, Department of Radiology, MGH, Charlestown, MA, United States

Amongst the existing acquisition tools for functional connectivity MRI (fcMRI), 32-channel phased array head coil is yet to be evaluated at 3T. Our results demonstrate that in comparison to most commonly used receive coils (e.g. 12-channel) highly parallel multiple channel arrays offer increased sensitivity in detecting detailed connections in resting state networks such as the default mode network. Furthermore, the 32-channel coil proved to be significantly better than the 12-channel as revealed by graph theory analysis in terms of both global and local network efficiency.

Hsu-Lei Lee¹, Benjamin Zahneisen¹, Thimo Grotz¹, Pierre LeVan¹, and Jürgen Hennig^1
1Medical Physics, University Medical Center Freiburg, Freiburg, Germany

Resting-state network analysis looks for coherent spontaneous BOLD signal fluctuations at frequencies lower than 0.1 Hz, where the most signal energy is stored. However hemodynamic signal change can occur at a faster rate and can also have different characteristics at different time scales. By using a highly under-sampled shell trajectory we were able to resolve a frequency spectrum up to 5 Hz. Preliminary tests found both coherent networks that are similar to those at 0.01~0.1 Hz frequency band, and some that differ.