Functional Magnetic Resonance Imaging or fMRI

1
Functional Magnetic Resonance Imaging or fMRI
How does a brain get red and yellow spots ?
? Center for Complex Systems and Brain Sciences,
2000
Created by Armin Fuchs
2
fMRI What is does

• fMRI is a modern noninvasive imaging
  technology which can be used to identify regions
  in the brain that get activated when participants
  perform certain tasks. In contrast to
  conventional MRI scans showing brain structure,
  fMRI provides information about brain function.

3
fMRI How it works

• In a nutshell The oxygen carrier in blood is the
  hemoglobin molecule. When this molecule releases
  oxygen to a cell its magnetic properties change
  from a so-called diamagnetic to a paramagnetic
  state. This difference can be found in image
  series taken by the scanner. With the assumption
  that in activated regions more oxygen is released
  compared to inactive regions, blood serves as a
  natural contrast agent for detecting brain
  activity.

4
So we detect the deoxygenated hemoglobin ?

• Well, not exactly. Because the blood flow in the
  brain overcompensates the actual need of oxygen,
  there is more oxygenated hemoglobin inside the
  active regions than can be used. So what we
  detect is the unused blood or the oxygenated
  hemoglobin. This is known as the Blood Oxygen
  Level Dependent or BOLD contrast.

5
How do we do that ?
MRI scanners can collect a variety of different
images depending on the scan sequence like the
ones below
T1 weighted images
Proton density images
T2 weighted images
6
Inversion recovery images
Echo planar images
7

• T1 weighted images are used to obtain brain
  structure for reconstructions of the cortical
  surface or as overlay images for for functional
  activity. They show fluid in dark and the brain
  right, i.e. white matter bright and gray matter
  gray.
• T2 weighted images are used in most medical
  scans. Fluid appears bright and gray matter is
  brighter than white matter.
• Inversion recovery scans allow for suppressing
  tissue with certain properties and for optimal
  control of the contrast.
• Echo planar images certainly dont have the same
  resolution as the other scans. In fact, they look
  pretty bad, but they can be taken extremely fast
  and make functional imaging possible.

8
Typical Scanning Times

• The time a scan takes depends on many things like
  resolution, intended signal to noise ratio and
  more. The following times are for standard
  sequences we typically use.
• T1 weighted full head scan (120 slices, 256x256
  pixel per slice with a voxel size of 1x1x2mm3)
  takes about 10min. T2 weighted and inversion
  recovery scans are comparable.
• The echo planar sequence we typically use takes
  20 slices with 64x64 pixels and a voxel size of
  3.5x3.5x7.5mm3
• IN 3 (THREE) SECONDS !!!

9
O.K., but what do we see on them ?

• By just looking at them not much. Below are two
  sets of echo planar images taken while the
  subject was moving a finger (top) and during rest
  (bottom).

10
So how does it work then?
Well, every three seconds we collect a volume of
20 or so slices and we tell our participant in
the scanner to keep her eyes closed and just
rest. We then scan for 30sec which gives us ten
volumes of the brain as a base line. Now we tell
our subject to perform a task, say move your
right index finger back and forth, and we keep
her doing that for another 30sec or ten volumes.
Then we tell her to rest again. This pair of
rest and task 30sec each we call a block. The
participant will perform four blocks of task and
rest lasting four minutes during which we scan
the volume 80 times or collect 1600 slices.
11
Now we have a 4-dimensional dataset which three
spatial dimensions (the volumes) and one
dimension in time. Each volume consists of 20
slices consisting of 64x64 pixels (for picture
elements) or 20x64x6481920 little volumes of
3.5x3.5x7.5mm3 each called voxels (for volume
elements). We have taken 80 of these volumes and
now we look at how the activity in individual
voxels changes as a function of time, i.e.
whether they are related to the periods of rest
and task the subject performed during the scan. A
typical example of time series of voxels from a
single slice looks like this
12
(No Transcript)
13
What did I see in this picture ?
On the left you see the time series corresponding
to the the voxels inside the green box on the
slice in the middle right. Obviously, certain
voxels follow the task function, i.e.
off-on-off-on plotted in red above the time
series in the middle. This function is delayed by
about 6sec to account for the time it takes the
hemo-dynamic system to respond. The bottom right
blow-up of one voxel shows that this curve and
the intensity recorded from the voxel match quite
well. Now the correlation between the task
function and the time series from all 81920
voxels are calculated. If the correlation is less
than 0.5 it is ignored. If the value is greater
than 0.5 the voxel gets a color corresponding to
the intensity between its min and max.
14
Why didnt we see this earlier? There was no
difference !
Actually, there was. Its just that the
difference between the task and the rest is very
small compared to the baseline (only a few ) so
we couldnt see it. After subtracting this
baseline the differences are quite striking.
Certain locations in the brain (the active
voxels) follow the task-rest cycles, others are
completely unaffected. Calculating the
correlation between the task function and the
voxel time series is one way to have a computer
find active regions (another method is to compare
the distributions during the task and rest
periods by a t-test to find the voxels where the
difference is significant).
15
Now what about these color spots?
Of course, the images dont come out of the
scanner with these colors. They are marks of
regions where the computer found high
correlations and represent the difference between
the average intensity in the task condition
compared to rest.
In a last step the low-resolution echo planar
image is replaced by a T1 weighted image (while
keeping the color spots) which is a much better
representation on the underlaying anatomy. Here
we see a finger movement task which involves
motor cortex (on the left) and a region called
supplementary motor area (or SMA) in the middle.
16
And thats it ???
Essentially, yes. Now we just have to study the
activity related to different tasks and find the
active regions inside the brain. We show a few
simple examples of what can be done with this
technology.
17
Right, left and bimanual
The next slide is an example for finger
sequencing. In the top row the task is performed
with the right hand, in the middle row with the
left hand and in the bottom row bimanually. Top,
the activity over sensory-motor areas is
strongest on the right-hand side of the image
(which is actually the left side of the brain,
the way radiologists like to have their images
because when they look at their patients they see
the left side on the right and vice versa). In
the middle row the activity is strongest on the
left (the right hemisphere of the brain), and on
the bottom the activation is approximately the
same in both hemispheres.
18
right
left
both
19
Motor versus Sensory
The brain regions where movement is controlled
for a certain limb are very close to those which
get the sensory inputs from that limb (and there
are probably good reasons for that). They are,
however, on different sides on the central
sulcus. The next slide shows in the right column
the activity from a finger movement, in the
middle a sensory stimulation of the same finger
(simply by rubbing), and on the right moving this
finger against an obstacle. It is evident that in
the first case the activity is more anterior
(towards the nose), in the second case it is more
posterior (further back), and the third row shows
simply both, i.e. motor and sensory activity.
20
movement
both
sensory
21
Syncopate and Synchronize
Brain activity not only depends on the task
itself but also on the context in which this task
is executed. The task in next slide was press a
small air pillows held between the thumb and
index fingers of both hands at a rate of 1.33Hz
paced by an auditory metronome. In the top row
the instruction was to press the pillow on the
beat, i.e. to synchronize with the metronome. For
the middle row the task was to execute the
movement in the middle between two consecutive
beats, i.e. to syncopate. The metronome was the
noise made by the scanner where we recorded four
slices every three second leading to pings at
1.33Hz. The bottom row shows the relation between
the metronome beats (red bars) and the average
movements (black lines), measured as pressure
changes in the pillows for both hands and task
conditions.
22
Syncopate
Synchronize
23
Syncopate and Synchronize
The slices are taken in the coronal plane because
we were mainly interested in activation in the
cerebellum, a structure known to be involved in
timing tasks. It is know that syncopation is a
more difficult task then synchronization. In
fact, if subjects are asked to syncopate at
higher movement rate (like 2.5Hz) they are unable
to to that whereas synchronization at this
frequency is no problem. We see the difference in
task difficulty in the cerebelar activity which
is much higher during syncopation than
synchronization.
Finally a little movie
24
Right Hand Figure Sequencing Task

• Subject continually touched thumb to the finger
  of the right hand in the sequence 4235 during the
  on period.
• This movie shows the comparison of on and off
  period of the functional scans rendered with a
  whole head volume of the same subject.
• Note the bilateral activation of the pre-central
  region (area M1) which is strongest in the
  contralateral hemisphere.
• Also note the strong activation in SMA and the
  ipsilateral cerebellum.

25
Thats it, thanks for watching.