Opportunities in Cognitive Neuroscience Research: Neuroimaging and Beyond

NINDS, NIA and NICHD Joint Workshop

November 28-29, 2000

Table of Contents

• I. Introduction
• II. Background
• III. Discussion/ Research Approaches in Cognitive Neuroscience
• IV. Conclusions/Recommendations
• Gaps in Scientific Knowledge
• Infrastructure and Funding Mechanisms
• V. Participants

---

I. Introduction

The National Institute of Neurological Disorders and Stroke (NINDS), along with the National Institute on Aging (NIA) and the National Institute of Child Health and Human Development (NICHD), recently conducted a workshop to examine research approaches integrating multiple technologies in cognitive neuroscience research. Cognitive Neuroscience, by its very nature, involves the interface of physiological, psychological, and computational approaches to understanding cognition. Currently, there are a number of technologies that have improved our ability to get a better view into the brain and gain insight on the neural substrates of cognition. New advances have been possible with positron emission tomography (PET), magnetic resonance imaging (MRI) and now functional MRI (fMRI). Each of these techniques measures a different aspect of brain function, and has a different accuracy and resolution. Combining these modalities is complex and far from complete. These tools have provided us with only limited data on the relationship between higher order cognitive function, the location of brain activity and the definition of brain systems and structures. Integrated across many techniques and species, these technologies can provide us with powerful tools that can bring together multiple scales, measurements, and source of imaging. Such investigation also requires new scientific perspectives, new paradigms of conducting research, collaboration and mechanisms of funding.

This workshop was unique since its focus was on what is needed to foster this integration and advance our understanding of the circuits and pathways of cognitive function. To be specific, the goal of the workshop was to encourage the development of research programs in cognitive neuroscience that use dynamic neuroimaging that integrates many methodologies. The particular issues were:

• Temporal, spatial and interpretive limits of imaging techniques:
  

• The relationship between neuronal activity and hemodynamic response.

• Best design neuroimaging strategies in cognitive neuroscience research:
  

• Event-related imaging, diffusion imaging.
  

  Protons to pathways and event-related optical signals (EROS).

• Paradigm shift to combinations of neuroimaging and psychophysical measures:
  

• fMRI
  

  magnetoencephalogram (MEG)
  

  electroencephalogram (EEG)
  

  event related potentials (ERP)
  

  transcranial magnetic stimulation (TMS)
  

  functional and effective connectivity mapping

• Validation of neural and behavioral models using neuroimaging and multi-array recording
  

• Linking knowledge from humans and other animals.

II. Background

The burgeoning of technology in the neuroimaging/cognitive neuroscience community has produced an onslaught of information. For instance, MRI machines have become available at many non-clinical institutions and scientists for whom imaging has been unavailable, have begun incorporating imaging experiments in their research programs. Long term, longitudinal studies are possible on learning, development, sleep, complex pharmacology, and animal models.

Effective combinations of the various methodologies measuring brain activity in vivo, requires new design and analysis considerations. For example, new strategies for combining the data sets need to be devised. These can range from simply comparing patterns of electrophysiological and hemodynamic effects from parallel studies, to directly incorporating hemodynamically identified activation areas into the ERP/MEG source-localization modeling analyses as a priori constraints. Another critical issue is the coordination of paradigms between methodologies. The recent development of event-related fMRI, especially at faster stimulus rates, can greatly facilitate the methodological integration of electrophysiological and hemodynamic imaging by enabling the fMRI paradigms to be more closely matched to those in ERPs/MEG, as well as the behavioral cognitive literature. In order to accommodate integration and combination of multiple methodologies, there is a need for new guidelines and mechanisms for reviewing grant applications, and ways of assessing the cost/benefit ratio of integrating multiple modalities. Some of the critical questions are:

• The new imaging techniques are expensive; is there a less expensive way to do the same research?
• Should the imaging equipment be centralized or distributed?
• Are there enough trained personnel?
• To what extent are investigators obliged to make their imaging data available to the scientific community?
• How will NIH facilitate cognitive neuroscience research that emphasizes integration of multiple technologies?

III. Discussion/ Research Approaches in Cognitive Neuroscience

Presentations by several scientists concentrated on cutting edge research strategies that would expand the scope of cognitive neuroscience research. Currently, scientists infer brain activity based on relative levels of input into imaging techniques. Therefore, the interpretation of imaging data requires sophisticated mathematical techniques both in the development of the imaging technique and in its interpretation. First, it was noted that, most of the common approaches to mapping reveal almost nothing about the functional/spatial relation of experimental events. Most brain function "maps" show colored voxels on a grey background of a brain x-ray. These colored voxels do not truly indicate the location where activity and function are related, and do not indicate where the voxels are statistically far above the background activity and when they are near the cutoff.

Second, different brain areas show different sensitivity in the imaging readouts even when the function of interest is absent. The issue of different sensitivity must be made explicit, or scientists will be looking at their data through a distorted lens, and would not know where true areas of interest are located. For both cases, statistical analyses need to be more rigorously applied to imaging techniques, and cognitive neuroscientists need more technical training in mathematics, physics and computer science.

Electrical activity can be coupled with the increase in oxygenated blood in the brain. EEG/MEG directly measures the electrical charge across neuronal membranes that are aligned and acting in concert. Near infrared spectroscopy measures the light differentially absorbed by oxygenated and non-oxygenated blood, and, with mathematical manipulations, can produce a kind of tomography. When EEG/MEG are combined with optical signals and fMRI, they provide a coupling between the electrical activity and the hemodynamic signal.

More work is needed to provide a better understanding of what these techniques can provide, ways to reduce distortions and artifacts, more precise models of both electrical activity and hemodynamic/metabolic measures, validation in animals and humans, quantification of the uncertainty, and a development of rational models of neural circuits with which to interpret the imaging signals.

Ongoing studies have demonstrated that it is possible to investigate complex behaviors with a top-down approach. In an imaging study of the behavioral correlates of "driving", scientists have used general linear modeling, which involved estimates of the hemodynamic response, as well as independent component analysis (ICA). It was possible to compare active virtual "driving" with passive driving while the subjects were scanned by MRI. Moreover, it was possible to compare driving with watching and establish what brain activation was involved in active driving, and to investigate the effects of alcohol and THC on driving. In general, multiple analysis methods can provide insight into complex data.

In order to look at brain function during complex reasoning tasks, investigators can choose a matching problem approach.. By varying the complexity of the matching, the study design and the time allowed for each trial, investigators were able to identify activity specific to the special process of reasoning engaged for higher level of thought. In addition, by subtracting one trial result from another, and applying mathematical manipulation, it is possible to identify ways in which specific cortical areas are functionally distinct and selectively involved.

Investigators can also use multiple parameters for a noninvasive macroscopic observation of integrated brain function. It is possible to make fine-grained models for brain activity based on electrophysical and magnetic field distributions by describing the inside of the head as a series of spheres, and by using finite difference methods. Diffusion tensor MRI, current density MRI and electrical impedance tomography (EIT) enhance the methods' accuracy. It is clear that the correlation of physical processes with electrophysical, optical,and hemodynamic imaging is based on effective modeling. It allows us to say something about where electromagnetic activity arises, connects neural network models to variables observed at the head surface and allows us to investigate cognition.

With current imaging protocols, investigators find that with hemodynamic imaging they know where something occurred, but not exactly when. With electrophysiologic imaging, they know when, but not exactly where. It now possible to design particular trial events, including a no-stimulation event. Then, one can effectively combine event-related fMRI with the event-related potentials (ERP)/MEG to establish the brain response to a particular visual stimulation and action.

It is possible to reconstruct the cortical surface from MRI data and use it as a constraint on the MEG/EEG inverse solution. fMRI of particular tasks may be used to show areas activated in that task. The resulting spatiotemporal 'movies' of brain activation in cognitive tasks are similar to those that can be inferred from previous direct intracranial measurements in epileptic patients. This method has been used in language, memory and perceptual tasks to resolve issues regarding the timing and modularity of brain processes. It may also be useful in localizing dysfunctional brain in epilepsy and other neuropsychiatric diseases.

Scientists have focused on the localization of brain activation during any given task. However the timing of this activation is important in a way similar to the timing of a symphony playing music, rather than noise. EEG, MEG and fMRI can be used as complementary measures and integrated in carefully directed research to develop millisecond by millisecond timing. With correlational analyses, for instance, investigators can establish the activation timing and location of areas for a to-be-attended stimulus and a to-be-ignored stimulus.

Such study design with event-related tools could allow for modeling of large scale neural networks, correctional analysis in space and time, descriptive models and the theoretical modeling of the rules that govern such brain activity. The caveat for such an approach is the need to train a new generation of scientists capable of using such integrative techniques.

A great deal of discussion focused on transcranial magnetic stimulation (TMS), a technique that uses a strong magnetic field to induce an electrical current within the brain without breaking the skull. Investigators can combine TMS with other imaging technologies. When TMS stimulates a brain area, there is a lag time before that area can function again. By stimulating with TMS, then asking the subject to perform a task, investigators are able to establish the particular cortical region necessary for a given cognitive operation. Using TMS-PET and TMS-EEG combination, they are able to assess connectivity and excitability in the human cerebral cortex.

Another approach has been the use of electrophysiological recording combined with neuroanatomical techniques in non human primates to examine the responses of the posterior parietal cortex and the lateral sulcus during active movements. It appeared likely that similar areas are involved in hand activity in the human. The spatial maps of the cortical areas that match body parts such as tail, foot, body, leg, hip are similar for many mammals--monkeys, rodents, even the platypus. Using combined fMRI and MEG techniques and parallel stimulus paradigms, scientists can establish which areas are homologous to the monkey and therefore likely to share common connection patterns, and which are unique to humans.

By looking at the intrinsic optical signal and sub-threshold voltage changes in the rat, scientists are able to identify particular interactions between groups of neurons. Rats have orderly column and row whiskers, or vibrissae on their snouts, each of which has a one-to-one correlation with a "barrel" of neuronal tissue in a similar arrangement in their brains. When a particular vibrissae is stimulated, its barrel responds, and barrels within the adjacent two rows and two columns show a lesser but still robust response. Despite the focal anatomy, the sub-threshold response in rat neurons is extensive. This has led o the conclusion that the spread and amplitude of the signal was better correlated with sub-threshold than with action potential activity.

Investigators are using optical imaging to map visual representations such as orientation, size, and color in the visual cortex. This method of data analysis appears sensitive and robust. The comparison and calibration of the optical imaging data with direct electrical recordings, provide good correlation between the two approaches, uncovering imaging relationships essential to understanding the brain.

IV. Recommendations/Conclusions

The workshop participants identified a number of areas of emphasis related to research initiatives and infrastructure development:

Gaps in Scientific Knowledge

• Physiologic basis of signal generation: Scientists needs to tie together the neural bases of hemodynamic measures in fMRI, PET, optical imaging and the neural activity measured by electrophysiology at the cell biology level.
• Modeling: It is vital that we develop models at different scales, from spikes in neuronal tissue through the stimulating or inhibitory integration that occurs before the spike, to networks and connectivity maps within an integrated brain and finally to models of brain activity and behavior in humans. Animal models, from the simplest up through primates and humans, are essential. Connectivity data in humans needs to be investigated in primates, and in humans using diffusion tensor imaging (DTI).
• Statistical methods: To accurately reflect physiological activity, the information from fMRI, PET, EEG, MEG, ERP, and DTI data requires more sophisticated statistical analysis than current standard practices. Statistical models need to be developed to incorporate and integrate results from different laboratories across all the methods.
• Developmental neurobiology at the systems level: Noninvasive research is necessary to understand the development of the various systems of the brain. Scientists who work at different levels on different systems must develop common vocabulary and communicate between scientific disciplines.
• Calibration: Measurement techniques that assess indicators of brain function, that relate to other imaging techniques, and ones that smear blocks of data need to be calibrated and validated to the underlying physiology, to each other, or to different techniques in different laboratories. We need to know what the filter is for each of these powerful techniques. We need to develop new tools, validate them and collaborate on their use. An attack on these basic validation questions may not be hypothesis-driven in the traditional sense. It may need to be viewed differently from the cognitive neuroscience community to ensure that novel breakthroughs are possible.

Infrastructure and Funding Mechanisms

• Training across disciplines: Neuroscientists need training in anatomy, computer science, computational theory and its applications, and neurobiology at the systems levels. Other specialists such as mathematicians, computer scientists, physicists, biologists, electrophysiologists, and engineers are essential to the correlation of imaging methods, the development of new imaging, modeling, calibration, validation, and psychophysiological research. These professionals need training in neuroscience, and need to recognize how their specialty can be applied to the field of neuroscience. Workshop participants were particularly anxious to increase the number of neuroscientists who understand mathematics, and physics and to attract mathematicians into the field.
• Dissemination of training: There must be alternative training approaches in addition to traditional training programs. These could be web-based, offered in summer institutes, or on compact disks or videos that would be available through a library. They must be well-designed, with recordings of training sessions plus auxiliary aids such as tutorials, online documentation, interactive video, workbooks, examples, troubleshooting practice, and technical support.
• Funding for calibration and validation: NIH should emphasize basic projects that calibrate and validate the correlation of different imaging methods and physiological brain events. Institute review of such projects should be by experts who understand the constraints of validation research and would apply a somewhat different viewpoint from that used for the usual cognitive neuroscience grants. It may be sufficient for review committees to include experts on calibration and validation, and to have instructions that make the scientific goals clear.
• Collaborations and drill deep grants: New grant mechanisms are necessary to foster collaboration beyond the present model. Granting mechanisms should reward specialists from other fields who came into cognitive neuroscience with a focus on funding the building of bridges between specialties. The workshop participants recognized that "drill deep-type grants" are needed to facilitate integrative research. Such grants would require large groups working closely on different levels and would redefine the field by blending in trained scientists from multiple outside specialties. Drill deep virtual projects could be structured across many institutions. Funding might be for the collaboration infrastructure itself, and not for the research. A recommendation was made to convene a meeting of cognitive neuroscience experts to select goals and actions for potential large-scale collaborative research projects.

V. Participants

Anders Dale, Ph.D.
NMR Center
Massachusetts General Hospital
149 13th Street, Room 2301
Charlestown, MA 02129
Phone: (617)- 724-9771
Fax: (617) 726-7422
Email: dale@nmr.mgh.harvard.edu

Vincent Calhoun, M.S.
Behavioral Science & Mental Hygiene
Division of Psychiatry Neuro-Imaging
Johns Hopkins University, Meyers 3-166
600 N. Wolfe Street
Baltimore, MD 21287-7362
Phone: (410) 955-7861
Fax: (410) 614-3676
Email: vcalhoun@jhu.edu

Emmeline Edwards, Ph.D.
Systems and Cognitive Neuroscience
National Institute of Neurological Disorders and Stroke (NINDS)
Neuroscience Center, Room 2109
6001 Executive Boulevard
Bethesda, MD 20892-9521
Tel: 301-496-9964
Fax: 301-402-2060
E-mail: ee48r@nih.gov

Gerald Fischbach, M.D.
Director
National Institute of Neurological Disorders and Stroke (NINDS)
Building 31, Room 8A54
31 Center Drive
Bethesda, MD 20892
Phone: (301) 496-9746
Fax: (301) 496-0296
Email: gdfisch@ninds.nih.gov

John Gabrieli, Ph.D.
Department of Psychology
Stanford University
Jordan Hall
Stanford, CA 94305
Phone: (650) 725-2430
Fax: (650) 725-5699
Email: gabrieli@psych.stanford.edu

Michael S. Gazzaniga, Ph.D.
Director
Center for Cognitive neuroscience
Dartmouth College
6162 Moore Hall - Room 452
Hanover, NH 03755-3569
Phone: (603) 646-1182
Fax: (603) 646-1181
Email: michael.gazzaniga@dartmouth.edu

John S. George, Ph.D.
Physics Division/ Biophysiology Group (P-21)
Los Alamos National Laboratory
MS D-454
Los Alamos, NM 87545
Phone: (505) 665-2550
Fax: (505) 665-4707
Email: jsg@lanl.gov

Eric Halgren, Ph.D.
NMR Center
Massachusetts General Hospital
149 13th Street
Charlestown, MA 02129
Phone: (617)-726-5122
Fax: (617) 726-7422
Email: halgren@nmr.mgh.harvard.edu

Terry L. Jernigan, Ph.D.
Clinical Psychologist & Professor
Department of Psychiatry & Radiology
VA San Diego Healthcare System
University of California, San Diego
Brain Image Analysis Laboratory
9500 Gilman Drive, 0949
La Jolla, CA 92093-0949
Phone: (858) 622-5882
Fax: (858) 622-5890
Email: tjernigan@ucsd.edu

Ehud Kaplan, Ph.D.
Opthamology/Physiology & Biophysics Departments
Mount Sinai School of Medicine
One Gustave Levy Place
New York, NY 10034
Phone: (212) 241-5874
Fax: (212) 289-5945
Email: kaplane@rockvax.rockefeller.edu

Leah A. Krubitzer, Ph.D.
Center for Neuroscience
University of California, Davis
1544 Newton Ct.
Davis, CA 95616
Phone: (530) 757-8868
Fax: (530) 757-8827
Email: lakrubitzer@ucdavis.edu

Guy McKahnn, M.D.
National Institute of neurological Disorders and Stroke (NINDS)
Building 31, Room
31 Center Drive
Bethesda, MD 20892
Phone: (301) 496-9746
Fax: (301) 402-2775
E-mail: gmckhan@ninds.nih.gov

Christopher I. Moore, Ph.D.
MIT Department of Brain & Cognitive Sciences
NE 20-392 MIT
Cambridge, MA 02139
Phone: (617) 258-0765
Fax: (617) 253-1504
E-mail: cim@ai.mit.edu

Tomas Paus, M.D., Ph.D.
Department of Neurology
Montreal Neurological Institute
McGill University
3801 University
Montreal, Quebec H3A 2B4
Canada
Phone: (514) 398-8481
Fax: (514) 398-1338
Email: tomas@bic.mni.mcgill.ca

Marcus E. Raichle, M.D.
Department of Radiology
Washington University School of Medicine
4525 Scott Avenue East Building
Campus Box 8225
St. Louis, MO 63125
Phone: (314) 362-6907
Fax: (314) 362-6110
Email: marc@npg.wustl.edu

Gregory V. Simpson, Ph.D.
Director
Dynamic Neurology Lab
Department of Radiology
University of California, San Francisco
521 Parnassus Avenue
San Francisco, CA 94960
Phone: (415) 502-6522
Fax: (415) 514-0405
Email: greg.simpson@radiology.ucsf.edu

Molly V. Wagster, Ph.D.
Neuroscience and Neuropsychology of Aging Program
National Institute on Aging
Gateway Building, Suite 3C307
7201 Wisconsin Ave.
Bethesda, MD 20892-9205 (20814 for overnight delivery)
Phone: (301) 496-9350
Fax: (301) 496-1494
E-mail: wagsterm@exmur.nia.nih.gov

Michael Weinrich, M.D.
National Institute of Child Health and Human Development
6100 Executive Blvd.
Room 2A03, MSC 7510
Bethesda, MD 20892-7510*
Phone: (301) 402-4201
Fax: (301) 402-0832
E-mail: weinricm@mail.nih.gov

Marty G. Woldorff, Ph.D.
Center for Cognitive Neuroscience
Duke University, Box 90999
LSRC Building - Room B203
Durham, NC 27708-0999
Phone: (919) 681-0604
Fax: (919) 681-0815
Email: woldorff@duke.edu