Understanding Behavior and Brain

1

• Understanding Behavior and Brain
• Activity Using fMRI Imaging
• Simone Campbell, Jim DeLeo, Carl Leonard
• Aron Barbey and Jordon Grafman
• National Institutes of Health
• Bethesda, Maryland
• Scientific Computing Section, NIHCC
• Cognitive Neuroscience Section, NINDS

2
Objective

• To study achievements in cognitive neuroscience
  that have led to better under-standing of how
  mental representations map into neural activity
  patterns and to suggest that new advancements in
  computer science may be useful in furthering this
  work, particularly with regard to bridging the
  gap between basic research and patient care - the
  main goal of translational medicine.

3
What is Functional MRI?

• Functional Magnetic Resonance Imaging (fMRI) is a
  neuroimaging technique that measures dynamic
  regulation of blood flow (hemodynamics) related
  to neural activity in the brain.
• The difference between an MRI and an fMRI scanner
  is that MRI scanners depict anatomical structure
  of the brain and other body areas, whereas fMRI
  depicts the metabolic activity within that
  anatomical structure in the brain.

MRI scan depicts only the structural make up of
the brain.
fMRI scan colored regions represent areas of
increased metabolic activity.
A typical fMRI scanner.
Source http//www.dmoma.org/lobby/movies/images/w
_logan_fry_axial_section.jpeg
4
Energy Use in the Brain

• The brain requires energy in the form of glucose
  and oxygen in order to function. This energy is
  supplied by the blood, which carries oxygen in
  the protein hemoglobin. This energy is used by
  neurons, the basic information processing cell in
  the brain.

5
Neural Activation
Source http//www.scholarpedia.org/article/Neurov
ascular_coupling
Neural activity requires energy in the form of
Adenine Triphosphate (ATP). ATP is produced by
glycolysis, which is the breaking down of
glucose, and by oxidative phosphorylation, which
requires oxygen to produce a large amount of ATP.
Blood vessels dilate in response to activation,
increasing blood flow and oxygen and glucose
delivery to the brain.
6
Energy Delivery

• Regions of increased neural activity require more
  energy, thus more blood flow to that active
  region. When a brain region is activated, a
  hemodynamic response is triggered, which is an
  increase in blood flow and blood volume in the
  brain. This action is the foundation for fMRI
  acquisition.

Resting vs. activated regions of activity
Vessels dilate to facilitate an increase in blood
flow to the active region.
Source http//www.fmrib.ox.ac.uk/education/fmri/in
troduction-to-fmri/images/BOLDeffect.png
7
Magnetic Properties

• A difference in the magnetic properties of
  oxygenated and deoxygenated hemoglobin is the
  starting point for acquiring fMRI scans.
  Oxygenated hemoglobin is diamagnetic, slightly
  repelling a magnetic field, while deoxygenated
  blood is paramagnetic, slightly attracting a
  magnetic field.

8
BOLD Response

• Blood Oxygen Level Dependent (BOLD) signal
  measured represents the total amount of
  deoxygenated hemoglobin present.

There will be more magnetic resonance (MR) signal
where blood is highly oxygenated and less MR
signal where blood is highly deoxygenated.
This image depicts the changes in MR signal
depending on the level of oxygenated vs.
deoxygenated blood present
Source http//www.cmrr.uic.edu/sections/fmri20ma
terials/images/Bold_contrast
9
Precession

• MRI scanners apply a powerful magnetic field to
  the subject. This causes the nuclei of a
  particular atom to undergo precession. Precession
  occurs when nuclei alter their orbit and align
  with the induced magnetic field. Some atoms have
  a positive spin, while others have a negative
  spin. Positive spin atoms always cancel out
  negative spin atoms. At room temperature there
  are more positive spin atoms than negative spin
  atoms, leaving unmatched positive spin atoms.

Source http//hyperphysics.phy-astr.gsu.edu/hbas
e/mechanics/imgmech/topp.gif
Precession refers to a change in the
direction of the axis of a rotating object. The
axis of the spinning object wobbles when a
magnetic field is applied to the atom. All
spinning objects, like the spinning top above,
can undergo precession.
10
Electron
How Signal is Detected
Nucleus
Atom
Energy levels
Positive spin atom in resting state.
Radio frequency pulse
Radio frequency pulse applied, causing atom to
resonate
Radio frequency pulse absorbed as a photon,
raising energy level
Radio frequency pulse removed. MRI machines
detect the photon energy emitted by atoms as they
transit from excited to resting
state.
Unmatched positive spin atoms are used by MRI
machines. Positive spin atoms are in a low energy
state. Magnetic energy perpendicular to the main
magnetic field is introduced in the form of a
radio frequency pulse specific to the atom,
causing the atom to jump up an energy level. When
this radio frequency pulse is removed, the atom
returns to its resting energy state. MRI machines
detect the photon energy emitted by the atom
while it is dropping back down to its resting
energy level after the radio frequency pulse has
been removed.
Positive spin atom in resting state.
11
fMRI Imaging
fMRI imaging provides thin tomographic vol-
ume slice images that reflect neuronal activity
in the brain.
12
Example of fMRI Image

• Increased blood flow is represented in fMRI
  volume elements (voxels) in three dimensional
  space. Colored regions represent voxel patterns
  where blood flow has increased. This corresponds
  to increased neural activity. Analyzing images
  like these has been important in cognitive
  neuroscience for achieving a better understanding
  of brain function.

Source http//www.nature.com/jcbfm/journal/v26/n1
0/fig_tab/9600274f1.html
13
Cognitive Neuroscience And Neural Activity

• Cognitive neuroscience is providing new knowledge
  about how mental representations map onto
  patterns of neural activity. Much of this
  success is related to the use of
  pattern-classification algorithms applied to
  voxel patterns of functional MRI data with the
  goal of decoding the information that is
  represented in the subjects brain at a
  particular point in time. These multi-voxel
  pattern analysis (MVPA) methods are useful for
  advancing understanding of neural information
  processing.

14
Method of Analysis for fMRI Results

• MVPA can be used to distinguish between
  cognitive states. Upon analysis of the results,
  categories of stimuli (ex. houses, scissors and
  chairs) can be with a reliable pattern of
  activity in a region of the brain. MVPA can be
  effective in distinguishing the category of a
  viewed object.

15
Patterns of Response
The image to the left shows the results of
analyses of fMRI scans used to measure patterns
of response in the ventral temporal cortex while
subjects viewed pictures of houses, chairs, and
other manmade products. Distinct patterns of
response were detected for each category of
stimuli. These distinct patterns were seen as
well when that particular region was excluded
from the analysis. Red and dark blue bars
represent analysis involving ventral temporal
cortex, an area most responsive to the stimuli.
Orange and light blue bars represent analysis
excluding the ventral temporal cortex.
16
Clinical Applications of fMRI

• Functional MRI (fMRI) has had a major impact in
  cognitive neuroscience. It now has an increasing
  role in clinical applications. (See Table 1.)
  These applications may be realized more fully
  with the support of modern computer science
  methodology.

17
Table 1. Clinical Applications for fMRI
pre-symptomatic diagnoses understanding
functional brain disorders understanding diseases
in general developing individualized
therapies clinical management fMRI-guided
neurosurgical planning fMRI-guided neurosurgery
understanding plasticity in recovery discovery
and development of new therapies

• pre-symptomatic diagnoses
• understanding functional brain disorders
• understanding diseases in general
• developing individualized therapies
• clinical management
• fMRI-guided neurosurgical planning
• fMRI-guided neurosurgery
• understanding plasticity in recovery
• discovery and development of new therapies

18
Presurgical Mapping

• fMRI guided neurosurgery is a method by which
  physicians map the brain prior to surgery, in
  order to localize cerebral functions in tissue
  near brain regions that are going to be removed
  during surgery. The primary goal would be to
  spare tissue that would limit recovery if damaged
  or removed from the brain. fMRI would be a
  manageable alternative to other imaging
  techniques such as magnetoencephalo-graphy.

19
Functional Characterization of Disease

• fMRI has the potential to be able to characterize
  the features of functional brain disorders.
  Functional brain disorders are disorders that
  arent clearly associated with any structural
  abnormalities in the brain.

20
Modern Computer Science

• Modern computer science provides many
  computational tools and strategies that may be
  relevant here. For example evolutionary
  computing (EC) provides techniques for evolving
  specific computational tools for specific
  purposes. Borrowing on ideas from Darwinian
  evolution, solutions evolve guided by optimizing
  one or more fitness functions. Within the frame
  of the EC paradigm, artificial neural networks,
  support vector machines, expert systems, fuzzy
  logic etc. may be used. Data mining (DM) is a
  major theme in computer science today. DM may be
  guided or unguided. In guided DM, particular
  questions are asked. In unguided DM, the machine
  is basically set loose to find interesting
  patterns that may reveal new knowledge or suggest
  new hypotheses worthy of exploration.
  Interactive data visualization is also an
  important and useful tool provided by modern
  computer science. We are interested in exploring
  possible applications of these techniques.

21
Conclusions

• We have accomplished our objectives in this
  project which were (1) to study the work in
  cognitive neuroscience that has lead to better
  understanding of how mental representations map
  into neural activity patterns and (2) to suggest
  how modern computer science may be used in
  advancing this work particularly with regard to
  bridging the gap between basic research and
  clinical applications - the main goal of
  translational medicine.

22
References

• 1. Bayesian decoding of brain images
• Karl Friston, Carlton Chu, Janaina
  Mourão-Miranda, Oliver Hulme,
• Geraint Rees,Will Penny and John Ashburnera
• Wellcome Trust Centre for Neuroimaging,
  Institute of Neurology, UCL
• 2. Analyzing for information, not activation, to
  exploit high-resolution Fmri
• Nikolaus Kriegeskorte and Peter Bandettini
• Section on Functional Imaging Methods, Lab of
  Brain and Cognition,
• National Institute of Mental Health, USA
• 3. Predicting human brain activity associated
  with the meanings of nouns.
• Mitchell TM, Shinkareva SV, Carlson A, Chang
  KM, Malave VL, Mason
• RA, Just MA
• Science. 2008 May 30320(5880)1191-5
• 4. Identifying natural images from human brain
  activity.
• Kay KN, Naselaris T, Prenger RJ, Gallant JL.
• Nature. 2008 Mar 20452(7185)352-5. Epub
  2008 Mar 5
• 5. Functional Magnetic Resonance Imaging. Scott
  A. Huettel, Allen W.

23
References

• 6. Applications of fMRI in translational
  medicine and clinical
• practice. Paul M. Matthews, Garry D. Honey,
  Edward T.
• Bullmore
• 7. Computational aspects of cognition and
  consciousness in
• intelligent devices.
• Robert Kozma Hrand Aghazarian, Terry
  Huntsberger, Eddie
• Tunstel, Walter J. Freeman
• 8. Beyond mind-reading multi-voxel pattern
  analysis of fMRI data
• Kenneth A. Norman, Sean M. Polyn, Greg J.
  Detre and James
• V. Haxby
• 9. Distributed and overlapping representations
  of faces and objects
• in ventral temporal cortex
• James V. Haxby, M. Ida Gobbini, Maura L.
  Furey, Alumit Ishai,
• Jennifer L. Schouten, Pietro Pietrini
• 10. wikepedia.com

24
Acknowledgements

• I am thankful to the Department of Clinical
  Research Informatics (DCRI) in the NIH Clinical
  Center (CC) for allowing me the opportunity to
  learn and experience as an intern here. I would
  like to thank Jim DeLeo and Carl Leonard for
  being supportive and helpful mentors throughout
  this project. My thanks also goes out to Dr.
  Jordon Grafman and Aron Barbey for sharing a
  depth of knowledge that was truly fascinating and
  intriguing.