Functional MRI

Spring 2007Spring 2007

University of KentuckyUniversity of Kentucky

CS689CS689 Computational Medical Imaging Processing Computational Medical Imaging Processing

CS 689 Computational Medical Imaging Processing •2•. 4/3/2007 University of

Kentucky

Outline

1. fMRI vs. MRI 2. Procedures of fMRI3. Medical Significance of fMRI4. Methods of fMRI5. Types of fMRI

CS 689 Computational Medical Imaging Processing •3•. 4/3/2007 University of

Kentucky

If the MRI experiment is done while a mental task is given to a subject, a so-called functional magnetic resonance image (fMRI) is generated.

fMRI is used to map different sensor, motor and cognitive functions to specific regions in the brain.

a Simple Definition:

CS 689 Computational Medical Imaging Processing •4•. 4/3/2007 University of

Kentucky

fMRI vs. MRI

MRI studies brain anatomy.Seeing brain structure3Dbrain anatomy

Functional MRI (fMRI) studies brain function.

CS 689 Computational Medical Imaging Processing •5•. 4/3/2007 University of

Kentucky

Procedures of fMRI1. a series of baseline images are taken of the

brain region of interest when the subject is at rest, as A

2. the subject performs a task

3. a second series of images is taken, as A’

4. the first set of images is subtracted from the second, as B=A’-A

5. the areas that are most visible in the resulting image, B, are presumed to have been activated by the task.

CS 689 Computational Medical Imaging Processing •6•. 4/3/2007 University of

Kentucky

One example

On the basis of the differences in timing activation in the brain, the areas responsible for hearing (in the middle of the brain in grey) and vision (in the back of the brain in white) could be localized by fMRI (Image courtesy of Dr. S. Smith from www.fmrib.ox.ac.uk).

A healthy subject was asked to listen to sentences being spoken while watching a screen with a flashing checkerboard presented. The sentences started and stopped at slightly different times than the flashing picture was turned on and off.

Visual cortex

Auditory cortex

CS 689 Computational Medical Imaging Processing •7•. 4/3/2007 University of

Kentucky

a Experiment Timing:

It consists of a set of trials, and the data is partitioned into trials. For some of these intervals, the subject simply rested, or gazed

at a fixation point on the screen. For other trials, the subject was shown a picture and a sentence, and instructed to press a button to indicate whether the sentence correctly described the picture.

For these trials, the sentence and picture were presented in sequence, with the picture presented first on half of the trials, and the sentence presented first on the other half of the trials.

Forty such trials are available for each subject. The timing within each such trial is as follows:

CS 689 Computational Medical Imaging Processing •8•. 4/3/2007 University of

Kentucky

The first stimulus (sentence or picture) was presented at the begining of the trail (image=1).

4 seconds later (image=9) the stimulus was removed, replaced by a blank screen.

4 seconds later (image=17) the second stimulus was presented. This remained on the screen for 4 seconds, or until the subject pressed the mouse button, whichever came first.

A rest period of 15 seconds (30 images) was added after the second stimulus was removed from the screen. Thus, each trial lasted a total of approximately 27 seconds (approximately 54 images).

Imaging parameters: Images were collected every 500msec. Only a fraction of the brain of each subject was imaged. The data is marked up with 25-30 anatomically defined regions (called "Regions of Interest", or ROIs).

CS 689 Computational Medical Imaging Processing •9•. 4/3/2007 University of

Kentucky

Medical Significance of fMRI

Different tasks activate different parts of the brain When listening to music, a specialized area in the

auditory cortex along the sides of the brain shows an increased signal

The locations vary for different cases and individuals.

CS 689 Computational Medical Imaging Processing •10•. 4/3/2007 University of

Kentucky

Medical Significance of fMRI a diagnostic method

learning how a normal, diseased or injured brain is working

assessing the potential risks of surgery or other invasive treatment of the brain.

planning brain surgery -- monitor normal brain function as well as any disturbed brain function.

While research is still ongoing, it appears that fMRI can also help assess the effects of stroke, trauma or degenerative disease (such as Alzheimer's) on brain function.

CS 689 Computational Medical Imaging Processing •11•. 4/3/2007 University of

Kentucky

fMRI? It aims to determine the neurobiological

correlation of behavior by identifying the brain regions (or “functioning modules”) that become “active” during the performance of specific tasks in vivo.

It extends traditional anatomical imaging to functional imaging. observe both the structures and which structures

participate in specific functions improving our understanding of a variety of brain

pathologies. such as the addictive behaviors of gambling or

drug abuse, are without structural brain changes.

CS 689 Computational Medical Imaging Processing •12•. 4/3/2007 University of

Kentucky

the principle:

when a brain region is being used, arterial oxygenated blood will redistribute and increase to this area.

CS 689 Computational Medical Imaging Processing •13•. 4/3/2007 University of

Kentucky

Knowledge required The science of applying fMRI is quite complicated and multi-

disciplinary. It involves: Physics: Statistics: Because the signals are very subtle, correct application of

statistics is essential in the statistical analysis of results to "tease out" observations and avoid false-positive results.

Psychology: When conducting fMRI on humans it is essential to employ carefully designed psychophysical experiments which allow the precise measurement of the neural effect under consideration.

Neuroscience: For a non-invasive scan, MRI has moderately good spatial resolution, but relatively poor temporal resolution. Increasingly, it is being combined with other data collection techniques such as electroencephalography (EEG) or magnetoencephalography (MEG), which have much higher recording frequencies.

Neuroanatomy: Anatomy is critical in understanding the location (and role) of the signals which fMRI is able to detect.

CS 689 Computational Medical Imaging Processing •14•. 4/3/2007 University of

Kentucky

3. Types of functional MRI: BOLD fMRI

Measures blood oxygenation, i.e, regional differences in oxygenated blood

perfusion fMRI measures regional cerebral blood flow,i.e. the rate at which blood

is delivered to tissue.

diffusion-weighted fMRI measures random movement of water molecules in tissue. It can detect acute brain infarction within 1 to 2 hours ***Infarct: 梗塞 ,如由于血栓或栓子的原因，局部血液供应不畅而引发局部组织坏死

Magnetic Resonance Spectroscopic Imaging (MRSI) measure certain cerebral metabolites noninvasively. phase encoding is used to obtain spectra from multiple regions

across a field of view.

CS 689 Computational Medical Imaging Processing •15•. 4/3/2007 University of

Kentucky

3.1 BOLD fMRI Blood Oxygen Level Dependent Contrast Based on:

(1) different magnetic properties of deoxy- and oxyhemoglobin

(氧合血红蛋白 ) (2) coupling of oxygenated blood flow and neuronal

activity High spatial and temporal resolution 3-6 second delay in hemodynamic(血液动力学 )

response ---limits optimal temporal resolution. Compares images taken during active and rest

states within a single session

CS 689 Computational Medical Imaging Processing •16•. 4/3/2007 University of

Kentucky

MRI vs. fMRI

neural activity blood oxygen fMRI signal

MRI fMRI

one image

many images (e.g., every 2 sec for 5 mins)

high resolution(1 mm)

low resolution(~3 mm but can be better)

fMRI Blood Oxygenation Level Dependent (BOLD) signal

indirect measure of neural activity

…

CS 689 Computational Medical Imaging Processing •17•. 4/3/2007 University of

Kentucky

Slice Thicknesse.g., 6 mm

Number of Slicese.g., 10

SAGITTAL SLICE IN-PLANE SLICE

Field of View (FOV)e.g., 19.2 cm

VOXEL(Volumetric Pixel)

3 mm

3 mm6 mm

Slice Terminology

Matrix Sizee.g., 64 x 64

In-plane resolutione.g., 192 mm / 64

= 3 mm

CS 689 Computational Medical Imaging Processing •18•. 4/3/2007 University of

Kentucky

Hemoglobin

Figure Source, Huettel, Song & McCarthy, 2004, Functional Magnetic Resonance Imaging

CS 689 Computational Medical Imaging Processing •19•. 4/3/2007 University of

Kentucky

Magnetic properties of deoxy- and oxyhaemoglobin

Deoxyhaemoglobin is paramagnetic The presence of deoxyhaemoglobin in vessel

causes a susceptibilitysusceptibility difference between the vessel and its surrounding tissue.

Then causes dephasing of MR proton signal Leading to a reduction in the value of T2*, which

causes a darkening of the image

oxyhaemoglobin is diamagnetic

CS 689 Computational Medical Imaging Processing •20•. 4/3/2007 University of

Kentucky

Susceptibility artifacts The magnetic susceptiblity of a material :::: a measure of how much magnetization is produced within

it when it is placed in a magnetic field. Susceptibility differences between tissues can lead to

signal loss in MR scans, especially in EPI scans.

The susceptibility difference between deoxygenated and oxygenated blood is the basis of the BOLD effect used to detect fMRI signals.

CS 689 Computational Medical Imaging Processing •21•. 4/3/2007 University of

Kentucky

HDR (Hemodynamics Response) the time courses of the changes in blood flow, blood

volume and blood oxygenation that occur in the brain in response to brain activity.

In the brain, neuronal activity is thought to cause a local increase in blood flow (CBF), which leads to an increase in blood oxygenation and blood volume (CBV).

Upon activation, oxygen is extracted by the cells, increasing the level of deoxyhaemoglobin in the blood. And it is compensated for by an increase in blood flow in the vicinity of the active cells, leading to a net increase in oxyhaemoglobin.

Signal changes

CS 689 Computational Medical Imaging Processing •22•. 4/3/2007 University of

Kentucky

T2*-weighted imaging T2*-weighted images are performed which take

advantage of the different magnetic properties of deoxy- and oxyhaemoglobin.

Because of the magnetic properties of the deoxyhaemoglobin molecule which causes rapid dephasing, T2* signal is retained longer in a region when it has more oxygenated blood compared to when there is less oxygenated blood.

Thus, an area with more oxygenated blood will show up more intense on T2*-weighted images compared to when there is less oxygenated blood around.

CS 689 Computational Medical Imaging Processing •23•. 4/3/2007 University of

Kentucky

BOLD Time Course

CS 689 Computational Medical Imaging Processing •24•. 4/3/2007 University of

Kentucky Hu et al., 1997, MRM

Evolution of BOLD Response

CS 689 Computational Medical Imaging Processing •25•. 4/3/2007 University of

Kentucky

Initial Dip (Hypo-oxic Phase) Transient increase in oxygen consumption, before change in blood

flow Menon et al., 1995; Hu, et al., 1997

Smaller amplitude than main BOLD signal 10% of peak amplitude (e.g., 0.1% signal change)

Potentially more spatially specific Oxygen utilization may be more closely associated with neuronal

activity than positive response

Slide modified from Duke course

CS 689 Computational Medical Imaging Processing •26•. 4/3/2007 University of

Kentucky

Rise (Hyperoxic Phase) Results from vasodilation of arterioles, resulting in a

large increase in cerebral blood flow Inflection point can be used to index onset of processing

Slide modified from Duke course

CS 689 Computational Medical Imaging Processing •27•. 4/3/2007 University of

Kentucky

Peak – Overshoot Over-compensatory response

More pronounced in BOLD signal measures than flow measures

Overshoot found in blocked designs with extended intervals Signal saturates after ~10s of stimulation

Slide modified from Duke course

CS 689 Computational Medical Imaging Processing •28•. 4/3/2007 University of

Kentucky

Sustained Response Blocked design analyses rest upon presence of sustained

response Comparison of sustained activity vs. baseline Statistically simple, powerful

Problems Difficulty in identifying magnitude of activation Little ability to describe form of hemodynamic response May require detrending of raw time course

Slide modified from Duke course

CS 689 Computational Medical Imaging Processing •29•. 4/3/2007 University of

Kentucky

Undershoot Cerebral blood flow more locked to stimuli than

cerebral blood volume Increased blood volume with baseline flow leads to

decrease in MR signal More frequently observed for longer-duration stimuli

(>10s) Short duration stimuli may not evidence May remain for 10s of seconds

Slide modified from Duke course

CS 689 Computational Medical Imaging Processing •30•. 4/3/2007 University of

Kentucky

BOLD fMRI BOLD = convolution of neuronal activity and

haemodynamic transfer function (gamma)

Neuronal Activity

Haemodynamic Function

BOLD Signal

Time

TimeSlide from Matt Brown

CS 689 Computational Medical Imaging Processing •31•. 4/3/2007 University of

Kentucky

BOLD SummatesNeuronal Activity BOLD Signal

Slide from Matt Brown

CS 689 Computational Medical Imaging Processing •32•. 4/3/2007 University of

Kentucky

BOLD Overlap and Jittering

Closely-spaced haemodynamic impulses summate.

Constant ITI causes tetanus.Burock et al. 1998.

CS 689 Computational Medical Imaging Processing •33•. 4/3/2007 University of

Kentucky

Design TypesBlock

Design

Slow ERDesign

RapidCounterbalanced

ER Design

RapidJittered ER

Design

MixedDesign

= null trial (nothing happens)

= trial of one type (e.g., face image)

= trial of another type (e.g., place image)

CS 689 Computational Medical Imaging Processing •34•. 4/3/2007 University of

Kentucky

* Example

With prior knowledge of the activation timing, we can perform a statistical test on the data to determine which areas of the brain are active, then overlay this statistical map (shown in color) on a high resolution MR image so that one can visualize the functional information in relation to relevant anatomical landmarks.

There are a wide variety of different software packages that facilitate processing, analysis and display of fMRI data in addition to many different stimulus delivery packages: http://www.fmri-world.de/ The choice of each depends largely on the onsite resources and the specific application.

A typical BOLD time course with 4 “active” states and 4 “resting” states.

CS 689 Computational Medical Imaging Processing •35•. 4/3/2007 University of

Kentucky

* Functional mapping : Statistical Parametric Mapping

Statistical parametric mapping entails the construction of spatially extended statistical processes to test hypotheses about regionally specific effects (Friston et al. 1991).

image processes with voxel values that are distributed according to a known probability density function, usually T or F distributions.

CS 689 Computational Medical Imaging Processing •36•. 4/3/2007 University of

Kentucky

General Linear Model (GLM) To estimate some parameters that could explain the spatially continuous

data in exactly the same way as in conventional analysis of discrete data.

CS 689 Computational Medical Imaging Processing •37•. 4/3/2007 University of

Kentucky

Gaussian random field

GRF theory is used to resolve the multiple-comparisons problem that ensues when making inferences over a volume of the brain.

GRF theory provides a method for adjusting p values for the search volume of an SPM to control false positive rates.

CS 689 Computational Medical Imaging Processing •38•. 4/3/2007 University of

Kentucky

Bayesian inferences

Bayesian inferences about spatially extended effects use posterior probability maps (PPMs).

CS 689 Computational Medical Imaging Processing •39•. 4/3/2007 University of

Kentucky

* Two issues on BOLD fMRI

No prior knowledge of activation timing

Temporal limits : HDR delay

Spatial limits of fMRI

CS 689 Computational Medical Imaging Processing •40•. 4/3/2007 University of

Kentucky

1. When the temporal response is unknown…

new fMRI analysis methods to detect activation without prior knowledge of activation timing.

One of these methods is called Independent Component Analysis (ICA)

It involves the development of novel ICA algorithms that are specific to fMRI data, the development of new stimulus designs that are appropriate for ICA and the use of ICA in patient populations to remove noise and motion artifacts.

CS 689 Computational Medical Imaging Processing •41•. 4/3/2007 University of

Kentucky

2. HDR delay & Limited temporal resolution A time delay from the onset of the neural activity to the

change of BOLD signal. a time delay of 3-6 seconds between when a brain region

is activated and blood flow increases to it.

During this time, the activated areas experience a relative decrease in oxygenated blood as oxygen is extracted by the active regional neurons. Afterward, the amount of blood flowing to the area far outweighs the amount of oxygen that is extracted so that oxygenated blood is now higher.

CS 689 Computational Medical Imaging Processing •42•. 4/3/2007 University of

Kentucky

Limited temporal resolution

Although images can be acquired every 100 msecs with EPI, this predictable but time-varied delayed onset of the BOLD response limits the immediate temporal resolution to several seconds instead of the 100 msec.

In the future, researchers may be able to improve the temporal resolution of fMRI by measuring the initial decrease in oxygenated blood with activation.

CS 689 Computational Medical Imaging Processing •43•. 4/3/2007 University of

Kentucky

3.What Limits Spatial Resolution noise

smaller voxels have lower SNR head motion

the smaller your voxels, the more contamination head motion induces temporal resolution

the smaller your voxels, the longer it takes to acquire the same volume 4 mm x 4 mm at 16 slices/sec OR 1 mm x 1 mm at 1 slice/sec

vasculature depends on pulse sequences

e.g., spin echo sequences reduce contributions from large vessels some preprocessing techniques may reduce contribution of large

vessels (Menon, 2002, MRM)

CS 689 Computational Medical Imaging Processing •44•. 4/3/2007 University of

Kentucky

The Initial Dip

The initial dip seems to have better spatial specificity However, it’s often called the “elusive initial dip” for a reason

fMRI for Dummies

CS 689 Computational Medical Imaging Processing •45•. 4/3/2007 University of

Kentucky

EXCEPT when the activated region does not fill the voxel (partial voluming)

Voxel Size

3 x 3 x 6= 54 mm3

e.g., SNR = 100

3 x 3 x 3= 27 mm3

e.g., SNR = 71

2.1 x 2.1 x 6= 27 mm3

e.g., SNR = 71

isotropic

non-isotropic

non-isotropic

In general, larger voxels buy you more SNR.

fMRI for Dummies

CS 689 Computational Medical Imaging Processing •46•. 4/3/2007 University of

Kentucky

Partial Voluming

The fMRI signal occurs in gray matter (where the synapses and dendrites are)

If your voxel includes white matter (where the axons are), fluid, or space outside the brain, you effectively water down your signal

fMRI for Dummies

CS 689 Computational Medical Imaging Processing •47•. 4/3/2007 University of

Kentucky

Partial Voluming

This voxel contains mostly gray matter

This voxel contains mostly white matter

This voxel contains both gray and white matter. Even if neurons within the voxel are strongly activated, the signal may be washed out by the absence of activation in white matter.

Partial voluming becomes more of a problem with larger voxel sizes

Worst case scenario: A 22 cm x 22 cm x 22 cm voxel would contain the whole brain

Partial volume effects: The combination, within a single voxel, of signal contributions from two or more distinct tissue types or functional regions (Huettel, Song & McCarthy, 2004)

fMRI for Dummies

CS 689 Computational Medical Imaging Processing •48•. 4/3/2007 University of

Kentucky

3.2. Diffusion and Perfusion MRI Diffusion MRI measures the molecular mobility of water in

tissue, while perfusion MRI measures the rate at which blood is delivered to tissue.

Therefore, both these techniques measure quantities which have direct physiological relevance.

Diffusion in biological systems is a complex phenomenon, influenced directly by tissue microstructure, and that its measurement can provide a large amount of information about the organization of this structure in normal and diseased tissue.

Perfusion reflects the delivery of essential nutrients to tissue, and so is directly related to its status.

CS 689 Computational Medical Imaging Processing •49•. 4/3/2007 University of

Kentucky

3.3. MRS Magnetic Resonance Spectroscopy. The data gathered

in MRS is presented as a spectrum. (i.e. the strength of the magnetic resonance signal is plotted as a function of resonant frequency).

Because of the way magnetic resonance works, the chemical environment of the nucleus being scanned will vary its resonant frequency. Hence, by observing the position of peaks in MR spectroscopic data it is possible to determine some of the molecules present in the sample.