Neuroimaging: from image to Inference

1
Neuroimaging from image to Inference

• Chris Rorden
• fMRI limitations relative to other tools used to
  infer brain function.
• fMRI signal tiny, slow, hidden in noise.
• fMRI processing a sample experiment.
• fMRI anatomy stereotaxic space.
• See also
• http//www.biac.duke.edu/education/courses/fall05/
  fmri/

2
Modern neuroscience

• Different tools exist for inferring brain
  function.
• No single tool dominates, as each has
  limitations.
• This course focuses on fMRI.

poor (whole brain)
iap
eeg
lesions
erp
nirs
pet
tms
Spatial resolution
fmri
good (neuron)
scr
good (millisecond)
poor (months)
Temporal resolution
3
Single Cell Recording

• Directly measure neural activity.
• Exquisite timing information
• Precise spatial information
• Often, no statistics required!

• Each line is one trial.
• Each stripe is neuron firing.
• Note firing increases whenever monkey reaches or
  watches reaching.

4
Single Cell Recording

• With SCR, we are very close to the data.
• We can clearly see big effects without
  processing.
• Unfortunately, there are limitations
• Invasive (needle in brain)
• Typically constrained to animals, so difficult to
  directly infer human brain function.
• Limited field of view just a few neurons at a
  time.

5
fMRI Processing

• Unlike SCR, we must heavily process fMRI data to
  extract a signal.
• The signal in the raw fMRI data is influenced by
  many factors other than brain activity.
• We need to filter the data to remove these
  artifacts.
• We will examine why each of these steps is used.

• Processing Steps
• Motion Correct
• Spatial
• Intensity
• Physiological Noise Removal
• Temporal Filtering
• Temporal Slice Time Correct
• Spatial Smoothing
• Normalize
• Statistics

6
fMRI signal sluggish

• Unlike SCR, huge delay between activity and
  signal change.
• Visual cortex shows peak response 5s after
  visual stimuli.
• Indirect measure of brain activity

2 1 0
0 6 12 18
24
Time (seconds)
7
What is the fMRI signal

• fMRI is Blood Oxygenation Level Dependent
  measure (BOLD).
• Brain regions become oxygen rich after activity.
• Very indirect measure.

8
Lets conduct a study

• Anatomical Hypothesis lesion studies suggest
  location for motor-hand areas.
• Ask person to tap finger while in MRI scanner
  predict contralateral activity in motor hand
  area..

M1 movement
S1 sensation
9
Task

• Task has three conditions
• Up arrows do nothing
• Left arrows press left button each time arrow
  flashes.
• Right arrows press right button every time arrow
  flashes.
• Block design each condition repeats rapidly for
  11.2 sec.
• No sequential repeats block of left arrows
  always followed by block of either up or right
  arrows.

10
Data Collection

• Participant Lies in scanner watching computer
  screen.
• Taps left/right finger after seeing left/right
  arrows.
• Collect 120 3D volumes of data, one every 3s
  (total time 6min).

11
Raw Data

• The scanner reconstructs 120 3D volumes.
• Each volume 64x64x36 voxels
• Each voxel is 3x3x3mm.
• We need to process this raw data to detect
  task-related changes.

12
Motion Correction

• Unfortunately, people move their heads a little
  during scanning.
• We need to process the data to create
  motion-stabilized images.
• Otherwise, we will not be looking at the same
  brain area over time.

13
Spatial smoothing

• Each voxel is noisy
• By blurring the image, we can get a more stable
  signal (neighbors show similar effects, noise
  spikes attenuated).

14
Predicted fMRI signal

• We need to generate a statistical model.
• We convolve expected brain activity with
  hemodynamic response to get predicted signal.

Predicted fMRI signal
Neural Signal
HRF

15
Predicted fMRI signal

• We generate predictions for neural responses for
  the left and right arrows across our dataset.
• Statistics will identify which areas show this
  pattern of activity.
• Several possible statistical contrasts (crucial
  to inference)
• Activity correlated with left arrows visual
  cortex, bilateral motor.
• More activity for left than right arrows
  contralateral motor.

16
Voxelwise statistics

• We compute the probability for every voxel in the
  brain.
• We observe that right arrows precede activation
  in the left motor cortex and right cerebellum.

17
fMRI signal change is tiny, noise is high

• Right motor cortex becomes brighter following
  movement of left hand.
• Note signal increases from 12950 to 13100, only
  about 1.2
• And this is after all of our complicated
  processing to reduce noise.

18
Coordinates - normalization

• Different peoples brains look different
• Normalizing adjusts overall size and orientation

Normalized Images
Raw Images
19
Why normalize?

• Stereotaxic coordinates analogous to longitude
• Universal description for anatomical location
• Allows other to replicate findings
• Allows between-subject analysis crucial for
  inference that effects generalize across
  humanity.

20
Goals for this course

• fMRI is notoriously difficult technique
• Sluggish signal
• Poor signal/noise
• Must find meaningful statistical contrasts
• This seminar reveals how to
• Devise meaningful contrasts
• Maximize signal, minimize noise
• Control for statistical errors.

21
Safety

• MRI uses very strong magnet and radiofrequencies
• 3T x60,000 field that aligns compass
• Metal and electronic devices are not compatible.
• MRI scanning makes loud sounds
• Rapid gradient switching creates auditory noise.
• Auditory protection crucial.
• MRI scanning is confined
• Claustrophobia is a concern.

22
Summary of Lectures

1. Introduction
2. Physics I Hardware and Acquisition
3. Physics II Contrasts and Protocols
4. fMRI Paradigm Design
5. fMRI Statistics and Thresholding
6. fMRI Spatial Processing
7. fMRI Temporal Processing
8. VBM DTI subtle structural changes
9. Lesion Mapping overt structural changes
10. Advanced and Alternative Techniques

23
Which tools

• There are many tools available for analysis.
• Different strengths.
• We predominantly focus on SPM and FSL.
• These are both free, popular and have good user
  support.

Tool SFN04
SPM 78.5
AFNI 9.1
FSL 7.4
BrainVoyager 4.1
https//cirl.berkeley.edu/view/Grants/BrainPyMotiv
ation
24
Reporting findings

• How do we describe anatomy to others?
• We could use anatomical names, but often hard to
  identify.
• We could use Brodmanns Areas, but this requires
  histology not suitable for invivo research.
• Both show large between-subject variability.
• Requires anatomical coordinate system.

25
Relative Coordinates

• On the globe we talk about North, South, East and
  West.
• Lets explore the coordinates for the brain.

26
Orientation - animals

• Dorsalback

Dorsal
Rostral
Caudal
Ventral

• Cranialhead
• Rostralbeak

• Caudaltail

• Ventralbelly

27
Coordinates Dorsal Ventral

• Human dorsal/ventral differ for brain and spine.
• Head/Foot, Superior/Inferior, Anterior/Posterior
  not ambiguous.

Dorsal Ventral
Dorsal Ventral
Dorsal Ventral
28
Coordinates Human

• Human rostral/caudal differ for brain and spine.
• Head/Foot, Superior/Inferior, Anterior/Posterior
  not ambiguous.

C
R
R
R
C
C
29
Orientation

• Human anatomy described as if person is standing
• If person is lying down, we would still say the
  head is superior to feet.

30
Anatomy Relative Directions
lateral lt medial gt lateral
Posterior ltgt Anterior
Ventral/Dorsal aka Inferior/Superior aka Foot/Head
Ventral ltgt Dorsal
Anterior/Posterior aka Rostral/Caudal
Posterior ltgt Anterior
31
Coordinates - Anatomy

• 3 Common Views of Brain
• Coronal (head on)
• Axial (birds eye), aka Transverse.
• Sagittal (profile)

32
Coronal

• Corona a coronal plane is parallel to crown that
  passes from ear to ear

33
Transverse

• Transverse/Axial perpendicular to the long axis

Example cucumber slices are transverse to long
axis.
34
Sagittal

• Sagittal arrow like
• Sagittal cut divides object into left and right
• sagittal suture looks like an arrow.

top view
35
Sagittal and Midsagittal

• A Sagittal slice down the midline is called the
  midsagittal view.

midsagittal
sagittal
36
Oblique Slices

• Slices that are not cut parallel to an orthogonal
  plane are called oblique.
• The oblique blue slice is neither Coronal nor
  Axial.

Cor
Oblique
Ax