Brain Mapping seminar abstracts 2009-10


Anna Moore, Ph.D.
Director, Molecular Imaging Laboratory
Associate Professor in Radiology, MGH
Wednesday, 09/16/09, noon, MGH bldg 149

Molecular Imaging: Methods and Clinical Potential for Neurological Applications
This presentation will focus on the application of molecular imaging modalities (MRI, PET, optical) for diagnosis and monitoring of various neurological disorders in small animals. Projects related to molecular imaging of the brain that are under way at the Martinos Center will be discussed as well.


Clas Linnman, Ph.D.
Research Fellow, Department of Psychiatry, Harvard Medical School
Wednesday, 09/23/09, noon, MGH bldg 149

Neuroimaging of the substance P system in pain
The Neurokinin 1 (NK1) or substance P system is involved in pain and anxiety behaviours, as evidenced from animal studies. Less is known about central alterations in this receptor system in humans. I will briefly review the current literature on PET imaging of the NK1 system, and then present new results from a study on chronic pain patients:
With multiple tracer PET, using both [15]-Oxygen labeled water and a [11]-Carbon labeled NK1 receptor antagonist, we demonstrate attenuated NK1 receptor availability throughout the medial pain network in patients with chronic pain. The differences between healthy subjects and chronic pain patients were most pronounced in the ventromedial prefrontal cortex. By combining blood flow measurements and receptor availability measurements, we found that prefrontal NK1 attenuations display opposing influences on rCBF in the anterior cingulate and the insula. Further, NK1 availability in the vmPFC was also related to fear and avoidance of movement. We speculate that NK1 receptors in the vmPFC may, in concert with the anterior cingulate and the insula, modulate the motor control output of the basal ganglia, thereby influencing fear and avoidance of movement in chronic pain.


Drazen Prelec, Ph.D.
Digital Eqipment Corp. LFM Professor of Management, M.I.T.
Wednesday, 09/30/09, 4pm, MIT bldg 46

A neuroeconomic study of self-deception
A motivated belief is one adopted partly to provide good news about one’s future prospects, or about an underlying personal trait or characteristic. The existence of motivated beliefs — motivated, for example, by fear or desire — is a psychological fact, amply demonstrated by experiments. What is less clear is how the brain accomplishes the trick. I propose that two distinct brain mechanisms combine to produce any volitional act: A mechanism responsible for overt belief expression, and second interpretive mechanism that draws inferences from that expression and generates an emotional response consistent with the inference. The two mechanisms work in parallel; hence, their interaction may be naturally modeled as a simultaneous (economic) signaling game. I describe a recent fMRI study (with T. Hedden, J. Gabrieli, and D. Mijovic-Prelec) that provides some further clues about localization and functioning of the conjectured mechanisms.


Oliver Hinds, Ph.D.
Postodoctoral Associate, Department of Brain and Cognitive Science, M.I.T.
Wednesday, 10/07/09, noon, MGH bldg 149

Real-time fMRI for behavior modification
An individual's brain state carries information about behavioral ability. Reliable measurement of moment to moment changes in brain state could be incorporated into behavioral training programs and brain-computer interfaces, could be used as a tool to test neuroscientific hypotheses, and could be applied to assess or guide treatments for disorders of the nervous system. I will both discuss methods for the reliable measurement of brain state using fMRI as well as present results of recent experiments that use brain state measurements to change behavior.


Moshe Bar, Ph.D.
Associate Professor, Department of Radiology, Harvard Medical School
Wednesday, 10/14/09, noon, MGH bldg 149

A cognitive neuroscience hypothesis of mood and depression
Although mood has a direct impact on mental and physical health, our understanding of the mechanisms underlying mood regulation is limited. I propose that there is a direct, reciprocal relation between the cortical activation of associations and mood regulation, whereby positive mood promotes associative processing, and associative processing promotes positive mood. This relation might stem from an evolutionary pressure for learning and predicting. Along these lines, one can think of mood as a reward mechanism that guides us to use our brains in the most productive manner. The proposed framework has many implications, most notably for diagnosing and treating mood disorders such as depression, for elucidating the role of inhibition in the regulation of mood, for contextualizing adult hippocampal neurogenesis, and for a general, non-invasive improvement of well-being.


Georg Langs, Ph.D.
Research Scientist, C.S.A.I.L., M.I.T.
Wednesday, 10/28/09, 4pm, MIT bldg 46

A Functional Geometry of the Brain
The talk will be about work exploring the functional geometry of cognitive processes. This geometry captures the global interaction pattern within the brain by mapping it to a space so that proximity reflects the functional relation. The map is obtained from fMRI data based on a diffusion process defined on the set of BOLD signals. It establishes a means of exploring the entirety of functional interactions, and the mutual roles of individual regions during particular tasks or conditions. As an example for the use as an exploratory tool, I will discuss results for a reward processing study. We compared control subjects, and cocaine abusers, who are hypothesized to have different sensitivity to reward. The results indicate that the functional geometry can be used to characterize and compare subject groups, and that it allows for insights in the mutual relations of brain regions.


Paul Yushkevich, Ph.D.
Assistant Professor, Department of Radiology
Penn Image Computing And Science Laboratory, University of Pennsylvania, PA
*Special date MONDAY, 11/02/09, noon, MGH bldg 149

Structure-Specific Techniques for Neuroimaging Analysis
In multi-subject structural, functional or diffusion-weighted MRI studies, it is common to analyze the imaging data using whole-brain approaches, such as voxel-based or deformation-based morphometry. Whole-brain analysis is a powerful tool for exploratory research, but in studies where we start with a priori hypotheses about individual anatomical structures, a more structure-centric approach to image analysis may be appropriate. This is especially true for complex structures such as the hippocampus, which tend to not be adequately normalized by whole-brain techniques. My talk will focus on recent work at Penn that uses deformable geometrical models as a framework for shape-based normalization, smoothing and analysis of individual anatomical structures. I will discuss two main applications that fall into this general structure-specific framework. The first is our work on the segmentation and analysis of hippocampal subfields, which leverages a detailed anatomical atlas derived from postmortem imaging. The second is our work on tract-specific analysis of diffusion-weighed MRI, which offers a parametric surface-based model for analyzing sheet-like white matter tracts.


Randy Gollub, M.D., Ph.D.
Associate Professor, Department of Psychiatry, Harvard Medical School
Associate Director, Neuroimaging Research Program, MGH
Wednesday, 11/04/09, noon, MGH bldg 149

How the Harvard Catalyst can help you
In this presentation the valuable clinical translational research resources available to the Martinos Center community and how to use them will be described. The Harvard Catalyst Clinical Translational Science Center (http://catalyst.harvard.edu/) is a new NIH/NCRR grant to Harvard to support exactly the work that is done by all investigators at Martinos- those done in humans (healthy subject and patient studies) and animals as well as developments that enable these studies. Including:
* Pilot study funding opportunities- current successes at Martinos and future plans
* How to use the free consulting services for Imaging, Genetics, Biostatistics and more
* Demonstration of new scientific networking tools
* Catalogues of resource directories and educational opportunities

The Harvard Catalyst evolved from its precursors, the GCRCs at the Harvard affiliated Academic Health Centers (MGH, BWH, BIDMC, CHB, Dana Farber Cancer Center and MIT). This transition has simplified and improved the experience of using our own Biomedical Imaging Core (BIC) facility and the associated resources (www2.massgeneral.org/crc/bic/index.htm). The BIC is located on the 2nd floor of CNY as you walk from Brain Map towards Bruce Fischl's office.

Highlights from GCRC/Harvard Catalyst supported studies done here at Martinos will be used to illustrate the resources available to investigators (e.g. space for your research subject study visits, physiological monitoring and/or blood draws during scanning and sample preparation, funds for lab studies). Harvard Catalyst funded faculty and staff will be introduced to you to let you know how they can help you. This includes our colleagues Giorgio Bonmassar and Vitaly Napadow who share their biomedical engineering expertise; Mark Vangel who provides comprehensive biostatistical consulting services for your imaging studies; Nancy Shearer our new dedicated research nurse; and the invaluable Kashawna Harling our Operations Associate who ensures your studies run smoothly.

An important new addition to our facility is the Transcranial Magnetic Stimulation (TMS) system now available for use by the Martinos Community. The Director of the TMS program, Tommi Raij, will describe the TMS system located in the BIC and the guidelines for using this new technology at Martinos.

Last, but not least, the brand-new web-based protocol submission system will be demonstrated (so simple! no Turbo!!).

The BIC will be open for touring after the talk to those interested in seeing the resources themselves. Interested? You can learn even more at: http://www.na-mic.org/Wiki/index.php/Collaboration:Harvard_CTSC


Seok Lew, Ph.D.
Research Fellow, Department of Radiology, Harvard Medical School
POSTPONED, MGH bldg 149

Finite element volume conductor and source analysis: dipole model, tissue conductivity estimation, and MNE-NeuroFEM integration
EEG and MEG provide useful means for identifying brain bioelectric sources in neuroscience and clinical applications. Subject-specific volume conductor models, constructed from sets of MRI or CT images can improve the accuracy of computational source localization. The finite element method (FEM) makes it possible to use the realistic geometry from the subject's images and to assign tissue conductivity in a flexible way. Three topics will be presented in this talk. First, how we model an equivalent current dipole in the finite element approach has an impact on the accuracy of forward solution. Three dipole models (Venant, partial integration and subtraction methods) are introduced and investigated for the overall forward accuracy. Second, we have developed a method to individually optimize the tissue conductivities in a realistically-shaped FEM volume conductor on the basis of MEG and EEG data. Our study shows that using somatosensory evoked potentials and realistic geometry, the method was able to simultaneously reconstruct both the brain and the skull conductivity. Third, NeuroFEM, a software package of finite element based source analysis, is being integrated with the MNE analysis software. The current development is briefly introduced.


Frank Guenther, Ph.D.
Professor, Department of Cognitive and Neural Systmes, Boston University
Wednesday, 11/18/09, noon, MGH bldg 149

The neural control of speech
Speech production involves coordinated processing in many regions of the brain. To better understand these processes, our laboratory has designed, tested, and iteratively refined a neural network model whose components correspond to brain regions involved in speech. Babbling and imitation phases are used to train neural mappings between phonological, articulatory, auditory, and somatosensory representations. After learning, the model can produce syllables and words it has learned by commanding movements of an articulatory synthesizer. Because the model’s components correspond to neurons and are given precise anatomical locations, activity in the model’s cells can be compared to neuroimaging data. Computer simulations of the model account for a wide range of experimental findings, including data on acquisition of speaking skills, articulatory kinematics, and brain activity during normal and perturbed speech. Furthermore, “damaged” versions of the model are being used to investigate a number of communication disorders, including stuttering, apraxia of speech, and spasmodic dysphonia. Finally, the model has been used to guide development of a brain-machine interface (BMI) aimed at restoring speech output to profoundly paralyzed individuals. A volunteer suffering from locked-in syndrome was able to use the BMI to control movements of a speech synthesizer in order to produce vowel sounds, reaching a level of 70% accuracy after 5-10 practice attempts of each vowel sound. These results were obtained from a single cone electrode with only two input channels; significant improvements in performance are expected in future systems utilizing more electrodes and optimized synthesizers.


Jing Jean Chen, Ph.D.
Research Fellow, Department of Radiology, MGH
Wednesday, 12/02/09, noon, MGH bldg 149

Dynamic and Noninvasive Measurement of Cerebral Venous Blood Volume Changes using MRI
Changes in cerebral venous blood volume (CBVv) constitute a critical component of the BOLD fMRI signal, but its role has remained unexplored in relation to that of total CBV changes, predominantly because measuring CBVv changes non-invasively is challenging. This talk will describe the development and application of the venous refocusing for volume estimation (VERVE) technique to non-invasively measure CBVv changes in humans at 3 Tesla. First, I will introduce the MR properties underlying VERVE contrast and summarize their characterization at 3 T. Next, I will focus on the use of VERVE in potentially elucidating some fundamental questions of brain physiology. In particular, I will discuss the measurement of the BOLD-specific blood flow-volume relationship pertaining to neuronal activation as well as vascular (CO2) challenges, which are key to the accurate quantification of cerebral oxygen metabolic rate using BOLD fMRI. Furthermore, I will discuss the contribution of venous CBV changes, as observed using VERVE, to the widely documented but still controversial post-stimulus BOLD undershoot.


David Silbersweig, M.D.
Chairman of Psychiatry, Brigham and Women's Hospital
Wednesday, 12/09/09, noon, MGH bldg 149

Failure of Frontal-Limbic Inhibitory Control in the Context of Negative Emotion in Borderline Personality Disorder
Abstract to follow.


Se Young Chun, Ph.D.
Research Fellow, Department of Radiology, MGH
Wednesday, 12/16/09, noon, MGH bldg 149

Regularization designs in motion-compensated image reconstruction and nonrigid motion estimation
Many medical imaging applications often require relatively long image acquisition times to form high-SNR images. However, long scan times can lead to motion artifacts. Conventional acquisition and reconstruction methods must sacrifice enough measurements for less motion artifacts or vice versa. Motion-compensated image reconstruction (MCIR) methods use all collected measurements, but reduce motion artifacts by incorporating (separately or simultaneously estimated) motion information into the image reconstruction framework. However, estimation problems of images and nonrigid motions are usually ill-posed and require regularization designs for both image and motion estimation. This talk presents two different regularization designs for nonrigid motion estimation and motion-compensated image reconstruction. First of all, we investigated methods for motion regularization. The usual choice for a motion regularizer in MCIR has been an elastic regularizer. Recently, there has been much research on regularizing nonrigid deformations with diffeomorphic motion priors. Conventional methods that enforce deformations to be locally invertible require high computational complexity and large memory. We developed a sufficient condition that guarantees the local invertibility and proposed a simple regularizer based on that sufficient condition. Secondly, we studied how motion affects the spatial resolution and noise properties of MCIR. We designed spatial regularizers to provide approximately uniform spatial resolution for MCIR. These regularizers enabled different MCIR methods to approximately have the same resolution. Noise properties were compared based on these regularizers.


Jonathan R Polimeni, Ph.D.
Research Fellow, Department of Radiology, MGH
Wednesday, 1/13/10, noon, MGH bldg 149

Laminar analysis of the human visual cortex using 7T fMRI
Accelerated image encoding and the increased sensitivity afforded by highly-parallel receive arrays and ultra high-field (7T) systems have recently enabled sub-millimeter isotropic resolution studies of the functional architecture of the human brain, the spatial accuracy of which is increasingly limited by the biological point-spread of the BOLD signal. However, the size distributions and densities of blood vessels are heterogeneous within the cortical gray matter and include vessels oriented randomly (capillaries), radial to the cortical surface (diving venules and arterioles), and tangential to the cortical surface (pial vessels). The asymmetry between the radial direction and the tangential direction leads us to hypothesize that the biological point-spread of BOLD may be different in the tangential and radial direction and vary as a function of laminar depth. For example, we show that a high degree of immunity to spatial localization errors from larger pial veins can be achieved by performing high-isotropic-resolution EPI (e.g., 1x1x1mm3) with a targeted sampling of the deeper cortical layers. To demonstrate this, we designed a spatial resolution stimulus to produce a desired spatial pattern of activity over primary visual cortex using a recent visuotopic mapping model, and compared the spatial fidelity of the pattern across several cortical depths.

A novel surface-based laminar analysis allowed for the selective sampling of different cortical depths across large extents of the folded cortical surface. The use of fast head gradients and a highly-parallel 32-channel receive coil limited susceptibility-induced EPI distortions to the point where a Boundary-Based Registration algorithm could accurately align the EPI data to the undistorted T1 volumes used for cortical surface reconstruction. Although sampling near the pial surface provided the highest signal strength, it also introduced the most spatial error. Thus avoiding surface laminae by restricting analysis to deeper layers improves spatial localization. Although standard gradient-echo EPI is broadly sensitive to extravascular signal changes surrounding all vessel sizes, by selectively sampling only from the deeper cortical layers BOLD signal changes appear to be predominantly driven by small parenchymal capillaries. Thus these acquisition and analysis techniques can together provide increased spatial accuracy for future functional measurements of both columnar and laminar organization.


Jing Jean Chen, Ph.D.
Research Fellow, Department of Radiology, MGH
Wednesday, 1/20/10, noon, MGH bldg 149

Dynamic and Noninvasive Measurement of Cerebral Venous Blood Volume Changes using MRI
Changes in cerebral venous blood volume (CBVv) constitute a critical component of the BOLD fMRI signal, but its role has remained unexplored in relation to that of total CBV changes, predominantly because measuring CBVv changes non-invasively is challenging. This talk will describe the development and application of the venous refocusing for volume estimation (VERVE) technique to non-invasively measure CBVv changes in humans at 3 Tesla. First, I will introduce the MR properties underlying VERVE contrast and summarize their characterization at 3 T. Next, I will focus on the use of VERVE in potentially elucidating some fundamental questions of brain physiology. In particular, I will discuss the measurement of the BOLD-specific blood flow-volume relationship pertaining to neuronal activation as well as vascular (CO2) challenges, which are key to the accurate quantification of cerebral oxygen metabolic rate using BOLD fMRI. Furthermore, I will discuss the contribution of venous CBV changes, as observed using VERVE, to the widely documented but still controversial post-stimulus BOLD undershoot.


Tammy Riklin-Raviv, Ph.D.
Postdoctoral Associate, C.S.A.I.L., M.I.T.
Wednesday, 01/27/10, 4pm, MIT bldg 46

Segmentation of Image Ensembles via Latent Atlases
The images acquired via medical imaging modalities are frequently subject to low signal-to-noise ratio, bias field and partial volume effects. These artifacts, together with the naturally low contrast between image intensities of some neighboring structures, make the extraction of regions of interest (ROIs) in clinical images a challenging problem. Probabilistic atlases, typically generated from comprehensive sets of manually labeled examples, facilitate the analysis by providing statistical priors for tissue classification and structure segmentation. However, the limited availability of training examples that are compatible with the images to be segmented renders the atlas-based approaches impractical in many cases. In the talk I will present a generative model for joint segmentation of corresponding regions of interest in a collection of aligned images that does not require labeled training data. Instead, the evolving segmentation of the entire image set supports each of the individual segmentations. This is made possible by iteratively inferring a subset of the model parameters, called the spatial parameters, as part of the joint segmentation processes. These spatial parameters are defined in the image domain and can be viewed as a latent atlas, that is used as a spatial prior on the tissue labels. Our latent atlas formulation is based on probabilistic principles, but we solve it using partial differential equations (PDEs) and energy minimization criteria. We evaluate the method successfully for the segmentation of cortical and subcortical structures within different populations and of brain tumors in a single-subject multi-modal longitudinal experiment.


Thomas Yeo, Ph.D.
Research Fellow, Department of Psychology, Harvard Univeristy
(presenting work done at the Medical Vision Group @ MIT-CSAIL and Lab for Computational Neuroimaging @ Martinos Center)
Wednesday, 2/03/10, noon, MGH bldg 149

Learning Task-Optimal Image Registration for Localizing Cytoarchitecture and Functions in the Cerebral Cortex
Image registration is typically formulated as an optimization problem with multiple tunable, manually set parameters. We present a principled framework for learning thousands of parameters of registration cost functions, such as a spatially-varying tradeoff between the image dissimilarity and regularization terms. Our approach belongs to the classic machine learning framework of model selection by optimization of cross-validation error. This second layer of optimization of cross-validation error over and above registration selects parameters in the registration cost function that result in good registration as measured by the performance of the specific application in a training data set. Much research effort has been devoted to developing generic registration algorithms, which are then specialized to particular imaging modalities, particular imaging targets and particular post-registration analyses. Our framework allows for a systematic adaptation of generic registration cost functions to specific applications by learning the ``free'' parameters in the cost functions. Here, we consider the application of localizing underlying cytoarchitecture and functional regions in the cerebral cortex by alignment of cortical folding. Most previous work assumes that perfectly registering the macro-anatomy also perfectly aligns the underlying cortical function even though macro-anatomy does not completely predict brain function. In contrast, we learn (1) optimal weights on different cortical folds or (2) optimal cortical folding template in the generic weighted Sum of Squared Differences (wSSD) dissimilarity measure for the localization task. We demonstrate state-of-the-art localization results in both histological and fMRI data sets.


BJ Casey, Ph.D.
Director of Sackler Institute for Developmental Psychobiology,
Cornell Univeristy
SPECIAL DATE & TIME: ** FRIDAY, 2/05/10, 2PM **, MGH bldg 149

The Adolescent Brain: Insights from human imaging to mouse genetics
The characterization of adolescence as a time of “storm and stress” remains an open debate. This presentation will cover the latest work linking adolescent brain development to changes in behavioral regulation. A testable neurobiological model that suggests an imbalance in the development of cortical control regions relative to subcortial limbic regions will be described. The model emphasizes the importance of examining developmental transitions into and out of adolescence as well as environmental and genetic factors that may put an individual at greater risk for “storm and stress” during this period. Empirical support for this model will be provided from pediatric imaging, imaging genetics and genetic mouse studies.


M. Justin Kim
PhD Candidate, Department of Psychological & Brain Sciences
Dartmouth College, Hanover, NH
Wednesday, 2/17/10, noon, MGH bldg 149

The functional and structural integrity of amygdala-prefrontal circuitry predicts individual differences in anxiety
Neurobiological theories of anxiety emphasize the central role of the amygdala and medial prefrontal cortex (mPFC) in the generation and experience of fear and anxiety. Building on the existing anxiety literature, I will present a combined fMRI and DTI study showing that a) functional activity within the amygdala can be used to identify a white matter amygdala-prefrontal pathway and b) the structural integrity of this pathway is inversely correlated with self-reported anxiety within the normal range. Further, from a subsequent experiment using resting state fMRI, I will present evidence that the strength of amygdala-mPFC functional connectivity during rest is correlated with individual differences in anxiety. Specifically, a dissociation between dorsal and ventral mPFC was observed, such that, amygdala-dmPFC functional connectivity was positively correlated with anxiety, whereas amygdala-vmPFC functional connectivity was negatively correlated with anxiety. Taken together, these data suggest that anxiety is related to both the structural and functional integrity of amygdala-mPFC circuitry in healthy individuals, even in the absence of anxiety-inducing stimuli. Implications of these findings for understanding normal fluctuations in anxiety levels as well as pathological anxiety (e.g., PTSD), which is hypothesized to involve abnormalities of this circuitry, will be discussed.


Satrajit Ghosh, Ph.D.
Research Scientist, Research Laboratory of Electronics, M.I.T.
Wednesday, 02/24/10, 4pm, MIT bldg 46

Nipype: Opensource platform for unified and replicable interaction with existing neuroimaging tools
Current neuroimaging software offer users an incredible opportunity to analyze their data in different ways, with different underlying assumptions. However, this has resulted in a heterogeneous collection of specialized applications without transparent interoperability or a uniform operating interface. Nipype, an open-source, community-developed initiative under the umbrella of Nipy, is a Python project that solves these issues by providing a uniform interface to existing neuroimaging software and by facilitating interaction between these packages within a single workflow. Nipype provides an environment that encourages interactive exploration of algorithms from different packages (e.g., SPM, FSL, FreeSurfer), eases the design of workflows within and between packages, and reduces the learning curve necessary to use different packages. Nipype is creating a collaborative platform for neuroimaging software development in a high-level language and addressing limitations of existing pipeline systems. Nipype is available on NITRC and SourceForge (http://nipy.sf.net/nipype) and is written in Python, a free high-level language accessible to both programmers and non-programmers with extensive scientific computation capabilities.


Seok Lew, Ph.D.
Research Fellow, Department of Radiology, Harvard Medical School
Wednesday, 3/3/10, noon, MGH bldg 149

Finite element volume conductor and source analysis: infant forward model, tissue conductivity estimation and MNE-NeuroFEM integration
EEG and MEG provide useful means for identifying brain bioelectric sources in neuroscience and clinical applications. Subject-specific volume conductor models, constructed from sets of MRI or CT images can improve the accuracy of computational source localization. The finite element method (FEM) makes it possible to use the realistic geometry from the subject's images and to assign tissue conductivity in a flexible way. Three topics will be presented in this talk. First, major features of FEM conductor model are introduced with a pipeline of FEM model build. As an example, the effect of infant fontanelle on the EEG/MEG field will be presented with a neonate FEM model. Second, we have developed a method to individually optimize the tissue conductivities in a realistically-shaped FEM volume conductor on the basis of MEG and EEG data. Our study shows that using somatosensory evoked potentials and realistic geometry, the method was able to simultaneously reconstruct both the brain and the skull conductivity. Third, Simbio-NeuroFEM, a software package of finite element based source analysis, is being integrated with the MNE analysis software. The current development is briefly introduced.


Danial Lashkari
Ph.D. Candidate, EECS-CSAIL, M.I.T.
Wednesday, 3/10/10, noon, MGH bldg 149

Unsupervised Learning for Discovery of Structure in the Brain Functional Organization from fMRI Data
Functional MRI experiments robustly localized regions in the ventral cortex that are selective for single visual stimulus categories (e.g., faces, bodies, scenes). However, further efforts to characterize the ventral visual stream with conventional hypothesis-based methods have encountered fundamental methodological challenges. Exploring the space of possible patterns of category-selectivity using common pairwise comparison tests becomes intractable as we consider more numbers of categories in an experiment. Another likely reason for our failure in finding novel selective regions is that our hypothesized categories do not match with the intrinsic structure of the visual system. In order to circumvent these issues, we developed an exploratory analysis method that automatically searches for coherent patterns of fMRI response that consistently appear across subjects. The method does not assume any spatial correspondence across subjects to avoid the shortcomings of voxel-wise spatial normalization techniques. The analysis is more generally applicable to fMRI studies of brain functional organization concerned with a large number of distinct tasks/stimuli. We verify this method on different visual fMRI experiments including one that presents 69 distinct images to subjects. The results demonstrate good agreement with the findings based on prior hypothesis-driven analyses, in spite of the fact that no category-related information is provided to the method. This finding suggests that our novel approach enables discovering consistent and complex patterns of fMRI response in presence of rich sets of tasks/stimuli.


Erkki Somersalo, Ph.D.
Professor, Department of Mathematics,
Case Western Reserve University, Cleveland, OH
Wednesday, 3/17/10, noon, MGH bldg 149 (NOTE: Conference Room A)

Bayesian hierarchical methods with application to MEG
In this talk, we discuss inverse problems in the Bayesian framework. Instead of a single estimate, the solution of the problem is the posterior probability distribution, and complementary information about the unknown augmenting the measured data is implemented in the form of prior distribution. When the prior information is qualitative in nature, hierarchical prior models turn out to be powerful tools to import the information into the model. In this talk, the basic ideas of hierarchical modeling are reviewed and an application to the MEG problem is discussed.


Stephanie Jones, Ph.D.
Instructor, Department of Radiology, Harvard Medical School
Wednesday, 3/24/10, noon, MGH bldg 149

From neurons to perception: using computational neural modeling to study the neural dynamics of human imaging signals
Elucidating the role that neural dynamics play in perception, cognition and action is a key challenge of modern neuroscience. While a wealth of information exists on this topic from human MEG and EEG studies, relating these data to the rich mechanistic understanding at the level of circuits and individual neuron types possible with animal models is a central missing link. In this talk, I will describe how biophysically principled neural modeling can be used to bridge this crucial divide, and provide novel, biophysically interpretable predictions on the neural origin of evoked and rhythmic cortical activity and their modulation with perception.

Specifically, we have examined in detail the relation between rhythmic and evoked activity in human primary somatosensory neocortex (SI). These signals show modulation that predicts tactile detection, changes with cued attention, and varies across age (even in the young adult range, from 20-40 years). To study the neural dynamics generating these signals, we developed a laminar cortical column model of an SI circuit, containing excitatory and inhibitory neurons, and feedforward and feedback inputs. Model activity maps directly onto the recorded signals and generates specific predictions of cellular level neural events that mediate the observed SI activity. I will describe a novel prediction on the origin of the beta frequency component of the commonly observed SI mu rhythm (containing 10Hz and 20Hz bands), and present initial data from an electrophysiologic study in awake rodents that supports the model predictions. Our findings may ultimately help to understand the neuropathology of these rhythms and their influence on sensory evoked responses in disease.


Janet M. Baker, Ph.D.
Chair, Saras Institute, West Newton, MA
Wednesday, 03/31/10, 4pm, MIT bldg 46

Speech and Language Knowledge Engineering Inside the Brain
There are now a number of brain recording and imaging techniques enabling the observation of brain functioning while subjects perform diverse speech and language tasks. This talk presents a variety of these studies (MEG, EEG, intracranial, etc.), both published and unpublished, that show where and when certain types of lexical, semantic, and syntactic processing occur. In recent MEG “brain movies” demonstrating word discrimination/recognition analyses, it is evident that speech and language information is highly distributed, both spatially and temporally. Some intriguing comparisons and contrasts with state-of-the-art speech recognition systems will be discussed.


M. Dylan Tisdall, Ph.D.
Research Fellow, Department of Radiology, Harvard Medical School
Wednesday, 4/7/10, noon, MGH bldg 149 (NOTE: Conference Room A)

Prospectively Motion-Corrected Anatomical Imaging: Available on a Scanner Near You
3D-encoded MPRAGE and T2SPACE scans are used routinely to produce high-quality T1- and T2-weighted images of neuroanatomy. However, because of their extended duration, these scans are particularly sensitive to subject motion. We present our recent work on making these sequences less sensitive to motion by introducing EPI-based navigators and using the PACE registration algorithm currently used in fMRI sequences. Our system allows users to perform motion-corrected MPRAGE and T2SPACE scans and have their images reconstructed on the scanner just as with the standard MPRAGE and T2SPACE sequences.

Prospective motion correction systems attempt to modify the scanning coordinates "on the fly" so that imaging occurs in consistent "patient coordinates" regardless of subject motion during the scan. Currently most users will be familiar with prospective motion correction via the PACE system for fMRI. In fMRI the scanner collects a succession of volumes with similar contrast. As the subject moves, the PACE algorithm attempts to prospectively motion-correct the imaging coordinates by registering the most recently acquired image volume to the first image volume. This gives an estimate of where the subject was last located and the scanning coordinates can be updated so that the image volumes are (ideally) aligned as they are acquired, instead of requiring alignment in post-processing.

We have extended this concept to anatomical imaging by embedding complete EPI volume acquisitions in each TR of the MPRAGE and T2SPACE sequences. This can be done at no cost in scan time because both sequences already have substantial "dead times" required to produce their desired contrast properties. As in the fMRI case, our sequences register the EPI volume acquired during each TR back to the volume captured in the first TR using the same PACE registration algorithm. We can then update the scan coordinates each TR to ensure that the coordinates of the anatomical imaging sequence remain consistent with the subject throughout the duration of the scan.

As these sequences are now available in Bays 3 and 4, we will discuss some of the practical issues associated with their use by groups at the center, as well as our future plans for the development of motion-insensitive anatomical imaging.


David Glahn, Ph.D.
Associate Professor, Yale University School of Medicine
Wednesday, 4/14/10, noon, MGH bldg 149

Using Images to Find Genes for Brain Structure and Function
Coming soon


Tor Wager, Ph.D.
Associate Professor, Department of Psychology and Neuroscience
University of Colorado, Boulder
SPECIAL TIME: THRUSDAY 4/15/10, 1PM, MGH bldg 149

Brain-body communication in stress and pain: A view from neuroimaging
Whereas some cognitive scientists once thought of emotion as merely a "volume knob," the last 15 years have seen a resurgence in quantitative work on affective processes and their effects on behavior, perception, action, and physical and mental disorders. This research provides new insight into the brain representations of pain, pleasure, and affective/motivational context. These processes, and the brain circuits that underlie them, are critical for healthy social and emotional functioning and are likely to play a central role in a variety of mental health disorders. In the first part of the talk, I draw on meta-analytic evidence to describe a vertically integrated, functional system that spans the medial prefrontal cortex, subcortical telencephalon, and brainstem. I will argue that this system is intimately involved in a number of functions central to the "self," including the representation of positive and negative outcomes, autobiographical memory, expectations about the future, and context-driven regulation of the autonomic and endocrine systems. In the second part of the talk, I present examples from recent functional neuroimaging studies that demonstrate a role for this system in the context-based regulation of pain, stress, and negative emotion by abstract contextual information. These studies suggest that medial prefrontal-brainstem pathways mediate placebo effects in pain, mental stress-induced changes in autonomic physiology, and successful self-generated regulation of emotion. These studies demonstrate that healthy function in circuits is likely to be important for a variety of mental health disorders and physical disorders with a neurogenic component.


Zhe Phillip Sun, Ph.D.
Assistant Professor, Department of Radiology, Harvard Medical School
Wednesday, 4/21/10, noon, MGH bldg 149

Quantitative chemical exchange saturation transfer (CEST) MRI and its in vivo application
Chemical exchange saturation transfer (CEST) MRI is a versatile imaging technique that provides a remarkably sensitive method that allows imaging of dilute proteins/peptides and microenvironment properties, in complementary to conventional relaxation and structural MRI. In fact, the sensitivity of CEST MRI is significantly enhanced so polypeptide, microenvironment pH and local temperature can be imaged, and remains promising for in vivo applications and ultimately, clinical translation. CEST MRI contrast, however, is complex. The experimentally obtainable CEST MRI not only varies with pH and local concentration of CEST agents, but also depends on parameters such as magnetic field strength and RF irradiation amplitude. It is important to note that recent progress in quantitative CEST MRI suggested that fundamental parameters could be derived from the apparently complex CEST contrast. Such development, if validated, may significantly advance the field of CEST MRI. Specifically, we have developed a novel and scalable numerical solution to describe multi-pool CEST MRI, which can be easily extended to simulate CEST contrast of unlimited labile proton groups. In addition, we modified dual 2-pool CEST model and elucidated the amide proton transfer (APT) MRI of ischemic stroke, and further optimized it for clinical translation. Moreover, iopamidol, a commonly used X-ray contrast agent, has been explored for ratiometric CEST MRI, potentially important for imaging renal pH in vivo. Topics including simplified numerical solution, ratiometric CEST MRI, pH-sensitive APT imaging of acute stroke, as well as preliminary clinical translation of CEST MRI will be presented.


Thomas Nichols, PhD.
Principal Research Fellow and Head of Neuroimaging Statistics
Department of Statistics & Warwick Manufacturing Group
University of Warwick, United Kingdom
Wednesday, 5/5/10, noon, MGH bldg 149

Advances in Statistical Methods in Imaging Genetics
There is growing interest in using neuroimaging measures as phenotypes in genetic analyses and, conversely, using genetic variables in imaging analyses. Since quantitative imaging phenotypes are more physiologically based than traditional phenotypes (like dichotomous disease state), the imaging associations should provide greater sensitivity and interpretability. In this talk I will give an overview of my work in this area, beginning with an overview of strategies currently used and how they can be improved. I consider how to define optimal imaging phenotypes, comparing ROI averages to voxel-wise searches; surprisingly, even when focal activations are expected, voxel-wise searches can be more sensitive than ROIs even after accounting for multiple comparisons. I also will present work on measuring the accuracy of inferences for VBM associations, with particular focus on nonstationary cluster size inference by Random Field Theory, and the accuracy of heritability inferences in twin studies, showing the dangers of the (widely used) Falconer's estimate. Finally, I will discuss the direction that I consider to have the greatest promise, simultaneous multivariate modelling of imaging and genetic data. In joint work with Giovanni Montana & Maria Vounou (Imperial College), we have proposed joint modelling with sparse multivariate methods, finding linear combinations of SNPs that correlated with brain regions. Based on detailed imaging and genetic simulations, we find superior power over linear models for the whole-brain, whole-genome setting.


Rajeev Raizada, PhD.
Research Assistant Professor
Neukom Institute for Computational Science
Darmouth College, NH
SPECIAL DATE/PLACE: Monday, 5/10/10, noon, MGH bldg 149 (NOTE: 6th Floor: 6033)

Using pattern-based fMRI to relate the structure of people's neural representations to their behavioural performance.
A key aim of neurocience is to understand how people's mental representations are structured, and how they give rise to behaviour. However, fMRI has had very little to say about this issue: merely showing which part of the brain lights up for a particular task does not in itself tells us anything about the mechanisms via which that task is carried out. I will present recent work aimed at addressing this problem. Drawing upon machine-learning research, we can treat brain scans not just as a collection of individual voxels lighting up, but instead as multivariate distributions of spatial patterns. These distributions have structure, which we can quantify and then seek to relate to behaviour. In a study of Japanese and English speakers listening to /ra/ and /la/, we found that the statistical separability of the neural patterns elicited by these sounds predicted individual differences in people's ability behaviourally to tell the sounds apart. Moving to the domain of numerical cognition, we have found that the separability of people's fMRI patterns in a "number sense" task correlated with their scores on standardised math tests. Both of these results hold true even when the whole brain is analysed at once in a single statistical test, without the need for any selection of a region-of-interest. These results, if they turn out to hold more generally, could allow fMRI to serve as a diagnostic tool, distinguishing neural representational competence from behaviourally measured performance. Possible implications for the diagnosis and remediation of learning disabilities will be discussed.


Joshua Roffman, MD.
Assistant Professor of Psychiatry, Harvard Medical School
Wednesday, 5/12/10, noon, MGH bldg 149 (NOTE: 6th Floor: 6033)

Psychotherapy and the brain: new perspectives from neuroimaging
Curiosity about the interface of psychodynamics and brain function stretches as far back as psychotherapy itself. In 1895, Sigmund Freud embarked upon his Project for a Scientific Psychology, an attempt to define the mechanisms of the unconscious mind in physiologic terms. A century later, functional neuroimaging studies have provided the first proof of concept for how psychotherapy changes brain function. A growing literature is now revealing convergent and mechanistically sensible effects of psychotherapy on neural activity across a range of psychiatric disorders, identifying in many instances brain networks that are uniquely responsive to talk therapies. This work could have important implications on the potential clinical use of brain imaging biomarkers in psychiatry, as a means to improve the efficiency of treatment assignment. We will discuss ongoing efforts at MGH and elsewhere that aim to bring psychotherapy into 21st century medicine through the use of brain imaging technologies.


Vivek Jay Srinivasan, PhD.
Research Fellow, Department of Radiology, Harvard Medical School
Wednesday, 5/19/10, noon, MGH bldg 149

Optical Coherence Tomography for in vivo, deep tissue imaging of the neurovascular unit
Optical Coherence Tomography (OCT) is an optical imaging modality that can perform micron-scale, imaging of tissue microstructure. With resolutions of 1-10 microns and penetration depths of >1 mm in highly scattering tissue, OCT occupies an important niche between macroscopic optical imaging such as diffuse optical tomography and microscopic optical imaging such as two photon microscopy. While OCT has become widespread in clinical ophthalmology, it is not yet widely used in neuroscience research. In this talk I will review the fundamentals of OCT as well as recent technological advances. I will discuss methods of Doppler OCT to measure quantitative relative and absolute blood flow and OCT angiography to perform rapid mapping of microvasculature. Applications related to neurovascular coupling and cerebrovascular physiology from our research group will be shown. Finally, I will present a new method of Optical Coherence Microscopy (OCM) that enables label-free in vivo imaging of cells and synaptic networks at depths of 1.3 mm in the rat cortex, capabilities which exceed current state-of-the-art two photon microscopy.


Contact the Webmaster