Functional Magnetic
Resonance Imaging (fMRI): a
            quick overview

             Matthew Baggott, Ph.D.
              Matthew@baggott.net
Section Overview
Goal is to understand fMRI well enough to make some sense of studies

We will discuss the basics of fMRI
   – Indirect measure of neural activity
   – (Actually measures local amount of de-oxygenated blood)
   – Slow
   – Noisy
   – Often analyzed by subtracting different conditions (or by
      correlating data)

In the next section, we will use what we learn to evaluate recent work
on the mechanisms of visuals in ayahuasca

Then we will discuss more general theories about psychedelic visuals
If you put someone in a very
homogeneous magnetic field…




   (image source: Wikipedia)
The protons in their body line up
  Protons = hydrogen nucleus, abundant in water and fat




          (like spinning tops aligned with gravity)
…then you can send a RF pulse, and
  get the protons to briefly turn 90
degrees away from the magnetic field
before they „relax‟ and realign
     like spinning tops that pop back up after a push




(Their relaxation and return to equilibrium can be divided into the
components that are parallel and perpendicular to the magnetic field,
which take different times (called T1 and T2) and can be used to make
slightly different images
Their movement gives off a
    signal that you can pick up
                                                (Head coil for
                                                measuring
                                                signal from
                                                head)




(They emit energy at the same radio frequency they received until they
return to their equilibrium state)
If you want to know where the signal
is coming from…           Gradient coils let us create magnetic
                          fields in any direction
                                           Z Coil
                 Y Coil
                                                     X Coil




                                 fnord
                   transceiver


make the field uneven and use pulses
 of different strengths and directions
Proton density and uneven magnetic fields
alter the image intensity

                       Proton density: density of
                       fat and water (important
                       for structural scans)

                       Uneven magnetic fields:
                       If each proton
                       experiences a slightly
                       different magnetic field,
                       the energy they give off
                       when they „relax‟ partly
  Image from Harvard   cancels out
  whole brain Atlas
Magnetic field irregularities from
hemoglobin

 Deoxyhemoglobin is a
 significantly more
 paramagnetic (with four
 unpaired electrons) than
 oxygenated hemoglobin

 The amount of deoxy-
 hemoglobin in each part of the
 image alters the image.
                                  Hemoglobin, which carries
                                    oxygen in the blood

                                    (image source: Wikipedia)
The magnetic difference between deoxy- and
oxygenated hemoglobin is the basis of the
Blood Oxygen Level Dependent (“BOLD”)
signal, used by almost all fMRI
The magnetic difference between deoxy- and
oxygenated hemoglobin is the basis of the
Blood Oxygen Level Dependent (“BOLD”)
signal, used by almost all fMRI




                           This is what we indirectly measure
The Blood Oxygen Level Dependent
(“BOLD”) signal is slow
                           ON = Checkboard shown




Visual cortex activity changes in response to a flashing checkboard in an early
event-related fMRI study (Blamire et al. 1992)
The Blood Oxygen Level Dependent
(“BOLD”) signal is slow
                           ON = Checkboard shown




Visual cortex activity changes in response to a flashing checkboard in an early
event-related fMRI study (Blamire et al. 1992)
The Blood Oxygen Level Dependent
(“BOLD”) signal is slow
                           ON = Checkboard shown




 Response peaks around 6 seconds after stimulus


Visual cortex activity changes in response to a flashing checkboard in an early
event-related fMRI study (Blamire et al. 1992)
The Blood Oxygen Level Dependent
(“BOLD”) signal is slow and ‘noisy’
                           ON = Checkboard shown




Seemingly random fluctuations are ~40% as big as the response to the stimulus

Visual cortex activity changes in response to a flashing checkboard in an early
event-related fMRI study (Blamire et al. 1992)
Most fMRI images show statistical maps
rather than raw changes in the BOLD signal
These statistical maps take into account the fact that different parts of the brain
have more variable signals.




                                     (Beauregard & Paquette 2006)
Blurry BOLD signal is projected onto high
resolution structural data or an average
brain
Resolution of functional data      Voxels with statistically
(volume pixel or voxel initially     significant changes
around 2 cm wide)
Blurry BOLD signal is projected onto high
resolution structural data or an average
brain
Resolution of functional data      Voxels with statistically
(volume pixel or voxel initially     significant changes
around 2 cm wide)
Blurry BOLD signal is projected onto high
resolution structural data or an average
brain
Resolution of functional data      Voxels with statistically
(volume pixel or voxel initially     significant changes
around 2 cm wide)
Blurry BOLD signal is projected onto high
resolution structural data or an average
brain
Resolution of functional data      Voxels with statistically
(volume pixel or voxel initially     significant changes
around 2 cm wide)
Blurry BOLD signal is projected onto high
resolution structural data or an average
brain
Resolution of functional data      Voxels with statistically
(volume pixel or voxel initially     significant changes
around 2 cm wide)
fMRI analyses are usually based on
subtraction of conditions
Most fMRI studies use a task-activation approach:
  – participants do a task
  – scientists look for which areas become more active

But “more active” compared to what?
      (the brain is always active)

Best comparison is usually another similar task

Early studies compared Tasks vs. “Quiet rest”
   – glossing over the fact that “Quiet rest” actually
     involves very active minds
What was
compared to
what?




              (Beauregard & Paquette 2006)
What was
compared to
what?




Mystical: Memory of Intense Closeness to God

Baseline: Memory of Intense Closeness to a Person

(This could go wrong in a lot of ways -- though you have to
start somewhere)                       (Beauregard & Paquette 2006)
So much data & so many comparisons, you need to
make sure you aren‟t finding activity due to
chance




         Neural correlates of interspecies perspective taking in the post-
         mortem Atlantic Salmon: An argument for multiple comparisons
         correction (Bennett, Baird, Miller, and Wolford 2009 poster)
         See Craig M. Bennett‟s blog post here:
         http://prefrontal.org/blog/2009/09/the-story-behind-the-
         atlantic-salmon/
Visual stimulation is often used to identify visual
maps in an individual‟s brain




                                 (Dougherty et al. 2003)
Specialized visual areas




(You can make identical images using WebCaret, online software provided by the
Van Essen lab at Washington University)
Really specialized visual areas
     Bodies
     Faces
     Houses
     Other objects




(After Op de Beek, Haushofer, & Kanwisher 2008)
Specialized areas can be used to study neural
correlates of conscious perception

                                         Face
                                         area




                                   House area




Left: When this image is viewed with red-green glasses, awareness switches
randomly between the face and house.
Right: BOLD signal in face and house sensitive areas change along with
consciousness
                                     (Tong, Nakayama, Vaughn, Kanwisher 1998)
The brain is 2% of the
                             body but uses 20% of the
                             energy




(Shulman et al. 2004; Raichle and Mintun 2006; Photo by Ben Chenoweth)
The brain is 2% of the
                             body but uses 20% of the
                             energy
                              Task-related fluctuations
                              are a small part (<5%) of
                              the brain‟s overall activity




(Shulman et al. 2004; Raichle and Mintun 2006; Photo by Ben Chenoweth)
The brain is 2% of the
                             body but uses 20% of the
                             energy
                              Task-related fluctuations
                              are a small part (<5%) of
                              the brain‟s overall activity
  Differences between normal and
  pathological populations in task-
  related changes are even smaller
  (often <1%)
(Shulman et al. 2004; Raichle and Mintun 2006; Photo by Ben Chenoweth)
What is the rest of the
                         activity?




(Shulman et al. 2004; Raichle and Mintun 2006; Photo by Ben Chenoweth)
Most of the energy used by the brain
really is used to support ongoing
neuronal signaling




       (Atwell & Laughlin 2001; Shulman et al. 2004; Raichle and Mintun 2006)
Large decreases in brain activity are
  produced by anesthesia

                    Awake                                Anesthesia (Isoflurane)




                                          mg/100gm/min




                     Awake                                              Anesthetized



Cerebral metabolism measured by 18FDG-PET (Hot colors indicate higher glucose use)


                                                                              (Alkire et al. 1997)
Can we find a way to analyze seemingly
random signal fluctuations?




                            (Blamire et al. 1992)
Yes, instead of subtracting activity
between tasks, you can correlate fMRI
signal between voxels




A “Default network” is active when research participants aren‟t told what to do.
Blue shows regions most active in “passive tasks” in a meta-analysis of PET data
                                  (Buckner, Andrews-Hanna, & Schacter 2008)
This default network is                           Default Network

similar to networks
active in internally
focused tasks
      Autobiographical Memory               Thinking about Others‟ Beliefs




       Envisioning the Future                   Moral Decision Making




                                (Buckner, Andrews-Hanna, & Schacter 2008)
Default network is anticorrelated with an
externally focused network




          Regions that negatively correlate with the
          default network are shown in cool colors; those
          that positively correlate are shown in warm
          colors
                            (Buckner, Andrews-Hanna, & Schacter 2008)
Default network is anticorrelated with an
externally focused network
   (When one network gets more active, the other gets less active)



                      % Signal Change




   (Buckner, Andrews-Hanna, & Schacter 2008; time course from Fox and Greicius 2010)
Looking for correlated activity between brain
areas is a powerful way to identify coordinated
brain networks.

                     Default


                                L. FEF




           Parietal Attention




           Ventral Attention




Frontal-Parietal Task Control




                                         (Power et al. 2011)
Correlated activity during movie viewing




Similar colors indicate brain regions that respond similarly
to natural movies (Nishimoto, Huth, Vu, and Gallant 2011)
Take home
The BOLD signal is an indirect, slow measure of neural
activity

Miraculously, it works. Results are consistent with direct
electrocortical measurements, studies of brain injury, etc.

Always ask what conditions are being compared and
how/why brain activity might differ between them –the
study may not be measuring what it is trying to measure
Some tools and resources
The Whole Brain Atlas at Harvard is just what it sounds like.
http://www.med.harvard.edu/AANLIB/
BodyParts3D is an online tool for browsing (Creative Commons
licensed) gross anatomy diagrams. http://lifesciencedb.jp/bp3d/
NeuroSynth is an online platform for large-scale, automated synthesis
of functional magnetic resonance imaging (fMRI) data extracted from
published articles
brainSCANr is an online engine to search and visualize co-occurrence
of terms in the scientific literature. http://www.brainscanr.com/
WebCaret is an online tool for visualizing a database of surface and
volume fMRI data. http://sumsdb.wustl.edu/sums/

Fmri overview

  • 1.
    Functional Magnetic Resonance Imaging(fMRI): a quick overview Matthew Baggott, Ph.D. Matthew@baggott.net
  • 2.
    Section Overview Goal isto understand fMRI well enough to make some sense of studies We will discuss the basics of fMRI – Indirect measure of neural activity – (Actually measures local amount of de-oxygenated blood) – Slow – Noisy – Often analyzed by subtracting different conditions (or by correlating data) In the next section, we will use what we learn to evaluate recent work on the mechanisms of visuals in ayahuasca Then we will discuss more general theories about psychedelic visuals
  • 3.
    If you putsomeone in a very homogeneous magnetic field… (image source: Wikipedia)
  • 4.
    The protons intheir body line up Protons = hydrogen nucleus, abundant in water and fat (like spinning tops aligned with gravity)
  • 5.
    …then you cansend a RF pulse, and get the protons to briefly turn 90 degrees away from the magnetic field
  • 6.
    before they „relax‟and realign like spinning tops that pop back up after a push (Their relaxation and return to equilibrium can be divided into the components that are parallel and perpendicular to the magnetic field, which take different times (called T1 and T2) and can be used to make slightly different images
  • 7.
    Their movement givesoff a signal that you can pick up (Head coil for measuring signal from head) (They emit energy at the same radio frequency they received until they return to their equilibrium state)
  • 8.
    If you wantto know where the signal is coming from… Gradient coils let us create magnetic fields in any direction Z Coil Y Coil X Coil fnord transceiver make the field uneven and use pulses of different strengths and directions
  • 9.
    Proton density anduneven magnetic fields alter the image intensity Proton density: density of fat and water (important for structural scans) Uneven magnetic fields: If each proton experiences a slightly different magnetic field, the energy they give off when they „relax‟ partly Image from Harvard cancels out whole brain Atlas
  • 10.
    Magnetic field irregularitiesfrom hemoglobin Deoxyhemoglobin is a significantly more paramagnetic (with four unpaired electrons) than oxygenated hemoglobin The amount of deoxy- hemoglobin in each part of the image alters the image. Hemoglobin, which carries oxygen in the blood (image source: Wikipedia)
  • 11.
    The magnetic differencebetween deoxy- and oxygenated hemoglobin is the basis of the Blood Oxygen Level Dependent (“BOLD”) signal, used by almost all fMRI
  • 12.
    The magnetic differencebetween deoxy- and oxygenated hemoglobin is the basis of the Blood Oxygen Level Dependent (“BOLD”) signal, used by almost all fMRI This is what we indirectly measure
  • 13.
    The Blood OxygenLevel Dependent (“BOLD”) signal is slow ON = Checkboard shown Visual cortex activity changes in response to a flashing checkboard in an early event-related fMRI study (Blamire et al. 1992)
  • 14.
    The Blood OxygenLevel Dependent (“BOLD”) signal is slow ON = Checkboard shown Visual cortex activity changes in response to a flashing checkboard in an early event-related fMRI study (Blamire et al. 1992)
  • 15.
    The Blood OxygenLevel Dependent (“BOLD”) signal is slow ON = Checkboard shown Response peaks around 6 seconds after stimulus Visual cortex activity changes in response to a flashing checkboard in an early event-related fMRI study (Blamire et al. 1992)
  • 16.
    The Blood OxygenLevel Dependent (“BOLD”) signal is slow and ‘noisy’ ON = Checkboard shown Seemingly random fluctuations are ~40% as big as the response to the stimulus Visual cortex activity changes in response to a flashing checkboard in an early event-related fMRI study (Blamire et al. 1992)
  • 17.
    Most fMRI imagesshow statistical maps rather than raw changes in the BOLD signal These statistical maps take into account the fact that different parts of the brain have more variable signals. (Beauregard & Paquette 2006)
  • 18.
    Blurry BOLD signalis projected onto high resolution structural data or an average brain Resolution of functional data Voxels with statistically (volume pixel or voxel initially significant changes around 2 cm wide)
  • 19.
    Blurry BOLD signalis projected onto high resolution structural data or an average brain Resolution of functional data Voxels with statistically (volume pixel or voxel initially significant changes around 2 cm wide)
  • 20.
    Blurry BOLD signalis projected onto high resolution structural data or an average brain Resolution of functional data Voxels with statistically (volume pixel or voxel initially significant changes around 2 cm wide)
  • 21.
    Blurry BOLD signalis projected onto high resolution structural data or an average brain Resolution of functional data Voxels with statistically (volume pixel or voxel initially significant changes around 2 cm wide)
  • 22.
    Blurry BOLD signalis projected onto high resolution structural data or an average brain Resolution of functional data Voxels with statistically (volume pixel or voxel initially significant changes around 2 cm wide)
  • 23.
    fMRI analyses areusually based on subtraction of conditions Most fMRI studies use a task-activation approach: – participants do a task – scientists look for which areas become more active But “more active” compared to what? (the brain is always active) Best comparison is usually another similar task Early studies compared Tasks vs. “Quiet rest” – glossing over the fact that “Quiet rest” actually involves very active minds
  • 24.
    What was compared to what? (Beauregard & Paquette 2006)
  • 25.
    What was compared to what? Mystical:Memory of Intense Closeness to God Baseline: Memory of Intense Closeness to a Person (This could go wrong in a lot of ways -- though you have to start somewhere) (Beauregard & Paquette 2006)
  • 26.
    So much data& so many comparisons, you need to make sure you aren‟t finding activity due to chance Neural correlates of interspecies perspective taking in the post- mortem Atlantic Salmon: An argument for multiple comparisons correction (Bennett, Baird, Miller, and Wolford 2009 poster) See Craig M. Bennett‟s blog post here: http://prefrontal.org/blog/2009/09/the-story-behind-the- atlantic-salmon/
  • 27.
    Visual stimulation isoften used to identify visual maps in an individual‟s brain (Dougherty et al. 2003)
  • 28.
    Specialized visual areas (Youcan make identical images using WebCaret, online software provided by the Van Essen lab at Washington University)
  • 29.
    Really specialized visualareas Bodies Faces Houses Other objects (After Op de Beek, Haushofer, & Kanwisher 2008)
  • 30.
    Specialized areas canbe used to study neural correlates of conscious perception Face area House area Left: When this image is viewed with red-green glasses, awareness switches randomly between the face and house. Right: BOLD signal in face and house sensitive areas change along with consciousness (Tong, Nakayama, Vaughn, Kanwisher 1998)
  • 31.
    The brain is2% of the body but uses 20% of the energy (Shulman et al. 2004; Raichle and Mintun 2006; Photo by Ben Chenoweth)
  • 32.
    The brain is2% of the body but uses 20% of the energy Task-related fluctuations are a small part (<5%) of the brain‟s overall activity (Shulman et al. 2004; Raichle and Mintun 2006; Photo by Ben Chenoweth)
  • 33.
    The brain is2% of the body but uses 20% of the energy Task-related fluctuations are a small part (<5%) of the brain‟s overall activity Differences between normal and pathological populations in task- related changes are even smaller (often <1%) (Shulman et al. 2004; Raichle and Mintun 2006; Photo by Ben Chenoweth)
  • 34.
    What is therest of the activity? (Shulman et al. 2004; Raichle and Mintun 2006; Photo by Ben Chenoweth)
  • 35.
    Most of theenergy used by the brain really is used to support ongoing neuronal signaling (Atwell & Laughlin 2001; Shulman et al. 2004; Raichle and Mintun 2006)
  • 36.
    Large decreases inbrain activity are produced by anesthesia Awake Anesthesia (Isoflurane) mg/100gm/min Awake Anesthetized Cerebral metabolism measured by 18FDG-PET (Hot colors indicate higher glucose use) (Alkire et al. 1997)
  • 37.
    Can we finda way to analyze seemingly random signal fluctuations? (Blamire et al. 1992)
  • 38.
    Yes, instead ofsubtracting activity between tasks, you can correlate fMRI signal between voxels A “Default network” is active when research participants aren‟t told what to do. Blue shows regions most active in “passive tasks” in a meta-analysis of PET data (Buckner, Andrews-Hanna, & Schacter 2008)
  • 39.
    This default networkis Default Network similar to networks active in internally focused tasks Autobiographical Memory Thinking about Others‟ Beliefs Envisioning the Future Moral Decision Making (Buckner, Andrews-Hanna, & Schacter 2008)
  • 40.
    Default network isanticorrelated with an externally focused network Regions that negatively correlate with the default network are shown in cool colors; those that positively correlate are shown in warm colors (Buckner, Andrews-Hanna, & Schacter 2008)
  • 41.
    Default network isanticorrelated with an externally focused network (When one network gets more active, the other gets less active) % Signal Change (Buckner, Andrews-Hanna, & Schacter 2008; time course from Fox and Greicius 2010)
  • 42.
    Looking for correlatedactivity between brain areas is a powerful way to identify coordinated brain networks. Default L. FEF Parietal Attention Ventral Attention Frontal-Parietal Task Control (Power et al. 2011)
  • 43.
    Correlated activity duringmovie viewing Similar colors indicate brain regions that respond similarly to natural movies (Nishimoto, Huth, Vu, and Gallant 2011)
  • 44.
    Take home The BOLDsignal is an indirect, slow measure of neural activity Miraculously, it works. Results are consistent with direct electrocortical measurements, studies of brain injury, etc. Always ask what conditions are being compared and how/why brain activity might differ between them –the study may not be measuring what it is trying to measure
  • 45.
    Some tools andresources The Whole Brain Atlas at Harvard is just what it sounds like. http://www.med.harvard.edu/AANLIB/ BodyParts3D is an online tool for browsing (Creative Commons licensed) gross anatomy diagrams. http://lifesciencedb.jp/bp3d/ NeuroSynth is an online platform for large-scale, automated synthesis of functional magnetic resonance imaging (fMRI) data extracted from published articles brainSCANr is an online engine to search and visualize co-occurrence of terms in the scientific literature. http://www.brainscanr.com/ WebCaret is an online tool for visualizing a database of surface and volume fMRI data. http://sumsdb.wustl.edu/sums/

Editor's Notes

  • #10 a hemisphere is about 372 (310-465) in^2