introduction to mri - oghabian to neuroimaging... · 2015. 3. 13. · – can resolve brief single...

An Introduction to Neuro-Imaging

for Neuroscience

(PhD Program)

Dr M A Oghabian

`

www.oghabian.net

1 )ناتوميكتصو يربرداري اطلاعات ا

2) يربرداري اطلاعات فيزيولوزيكتصو

3 )(تغييرات شيميائي)يربرداري اطلاعات مولكولي تصو

4 )يربرداري اطلاعات عملكرديتصو

تصويربرداري بر مبناي نوع اطلاعات

PET SPECT FUNCTION

STRUCTURE CT MRI

Structural & Functional Imaging

Nuclear Medicine is to physiology as Radiology is to anatomy

TR = 500 ms TE = 20 ms

Spin-Echo Images

TR = 2000 ms TE = 40 ms

TR = 2000 ms TE = 80 ms

Contrasts: PD, T1, T2 and T2* weighting

PD-weighting

(proton/spin density)

+ T1-weighting

(gray/white contrast)

+ T2-weighting

(bright CSF/tumor)

Which is best for brain morphometry/FreeSurfer?

FLASH 5° MPRAGE T2-SPACE

Brain Structural Analysis

1- Cortical (surface-based) Analysis.

2- Volume Analysis.

3- Voxel Based morphometry

4- Fiber Tracking

Surface and Volume Analysis

Cortical Reconstruction

and Automatic Labeling Inflation

Surface Flattening

Automatic Subcortical

Gray Matter Labeling

Automatic Gyral White

Matter Labeling

DTI

(Diffusion

Tensor

Imaging)

DTI

Fiber

Tracking

DTI Fiber Tracking and fMRI

Functional Neuro-Imaging

Methods

fMRI vs. PET

• fMRI does not require exposure to radiation

– fMRI can be repeated

• fMRI has better spatial and temporal resolution

– requires less averaging

– can resolve brief single events

• MRI is becoming very common; PET is specialized

• MRI can obtain anatomical and functional images within same

session

• PET can resolve some areas of the brain better

• in PET, isotopes can tagged to many possible tracers (e.g., glucose

or dopamine)

• PET can provide more direct measures about metabolic processes

Spatial and Temporal Resolution of

Various functional imaging methods

Perfusion measurement • Goal of perfusion MRI is to measure the blood

flow

• This flow corresponds to microcirculatory

tissue perfusion rather than the flow of the

main vascular axes

• This is expressed in milliliters per 100 gram of

tissue per minute.

• Gadolinium chelates do not cross the normal

blood-brain barrier. But in pathological

conditions they cross, and therefore produces

extravasation,

Perfusion parameters • These parameters are extracted by post-processing the signal

curves, and are:

• TA: time of arrival of the contrast agent in the slice after injection

• TTP (Time To Peak): time corresponding to the maximum

contrast variation

• MTT (mean transit time)

• Peak amplitude: percentage loss of intensity of the maximal

signal

• rCBV (regional cerebral blood volume): The area below the

decreasing signal curve

• rCBF (regional cerebral blood flow): The ratio rCBV/MTT.

• RCBV and rCBF are relative measurements, giving the

calculation of ratios between pathological zone and healthy zone

(which serves as a reference).

• Permeability (Ktrans)

FMRI – Presurgical Planning MA Oghabian: NeuroImaging and Analysis Lab- Tehran Unive of Medical Sci

Pre

-surg

ical risk e

valu

ation

Perf

usio

n M

RI

(CB

V/C

BF

/MT

T/T

TP

)

Perfusion MRI

• Perfusion MRI is a technique for evaluation of microscopic

blood flow in cerebral capillaries and venules.

• (A) Perfusion of a high grade brain tumor demonstrates areas of

increased capillary blood volume in tumor (red) corresponding to tumor

neovascularity.

• (B) Perfusion map superimposed on FLAIR image demonstrates area of

tumor with highest malignancy potential to aid biopsy planning

Diffusion MRI in acute stroke

• (a) Day 1 diffusion MRI demonstrates a small cortical a

sub-cortical infarct.

• (b) Day 2 DWI demonstrates enlargement of the stroke

(limiting Diffusion).

• (c) Perfusion MRI demonstrates reduced blood flow

corresponding to final infarct size.

MRI Spectroscopy (MRS)

• MRS is a special technique used for

characterization of the biochemistry of

tumors, infarcts, and other pathology.

•

• MRS is useful for demonstrating

aspects of physiology such as tumor

aggressiveness and anaerobic

metabolism.

• 2D MRS of the brain demonstrates

the characteristic internal biochemical

makeup of glioblastoma multiforme, a

highly aggressive brain tumor

در حضور میدان مغناطیسی TE= 30msتصویر طیف پروتون در در

7T ترکیب متفاوت بر روي 14مطالعه اي بر روي مغز انسان در میدان در

طیف قابل مشاهده بوده اند


Pre

-surg

ical risk e

valu

ation

MR

S (

Cholin

incre

ase in tum

our

• fMRI (BOLD imaging)

The Brain Before fMRI (1957)

Polyak, in Savoy, 2001, Acta Psychologica

Solution: Brain Functional studies • Imaging

– CBF

– CBV

– Perfusion study

– Diffusion study

– Metabolite evaluation (MRS)

• Creatine

• Lactate

• Phosphor (ATP)

– ……

– BOLD

• Signal measurement

• Post-Synaptic Potentials

• EEG (ERPs) and MEG

• Action Potentials

• Electrophysiology

What is fMRI?

• Functional Magnetic Resonance Imaging (fMRI): uses MRI to indirectly measure brain activity

• Known for over 100 yrs. that blood flow and blood oxygenation are linked to neural activity– only since the early 1990’s was fMRI developed (Ogawa & Kwong)

• Based on the assumption that neuronal activity requires O2

which is carried by the blood; increased blood flow and resulting hemodynamics are foundation to fMRI

Why Use fMRI?

• Clinical uses- pre-surgical planning, id’ing brain pathologies, and many future potential uses

• Research Purposes- many researchers from several disciplines are using fMRI to better understand brain function in animals and humans

Language task (WG) on left-handed

patient with Temporoparietal mass


History of fMRI

MRI

-1971: MRI Tumor detection (Damadian)

-1973: Lauterbur suggests NMR could be used to form images

-1977: clinical MRI scanner patented

-1977: Mansfield proposes echo-planar imaging (EPI) to acquire images faster

fMRI

-1990: Ogawa observes BOLD effect with T2*

blood vessels became more visible as blood oxygen decreased

-1991: Belliveau observes first functional images using a contrast agent

-1992: Ogawa et al. and Kwong et al. publish first functional images using BOLD signal

Ogawa

Necessary Equipment

Magnet Gradient Coil RF Coil

Source: Joe Gati, photos

RF Coil

4T magnet

gradient coil

(inside)

fMRI Setup

Contrast Agents in MRI

• Definition: Substances that alter magnetic

susceptibility of tissue or blood, leading to

changes in MR signal

– Affects local magnetic homogeneity: decrease in T2*

• Two types

– Exogenous: Externally applied, non-biological

compounds (e.g., Gd-DTPA)

– Endogenous: Internally generated biological

compound (e.g., deoxyhemoglobin, dHb)

BOLD contrast

FMRI – Week 6 – BOLD fMRI Scott Huettel, Duke University

Oxygenated blood?

No signal loss…

Deoxygenated blood?

Signal loss!!!

Images from Huettel, Song & McCarthy, 2004, Functional Magnetic Resonance Imaging

http://fmri4newbies.com/


Disadvantage of T2*: Susceptibility Artifacts

-In T2* images, artifacts occur near junctions between

air (sinuses, ear canals) and tissue

sinuses

ear

canals

T1-weighted image T2*-weighted image


Blood Deoxygenation affects T2* Decay

Thulborn et al., 1982

BOLD Endogenous Contrast

• Blood Oxyenation Level Dependent Contrast – Deoxyhemoglobin is paramagnetic

– Magnetic susceptibility of blood increases linearly with increasing Deoxygenation

• Oxygen is extracted during passage through capillary bed – Brain arteries are fully oxygenated

– During activation Venous (and capillary) blood has increased proportion of Doxyhemoglobin

– Then oxygen is compensated in veins

– Difference between oxy and deoxy states becomes greater for veins BOLD sensitive to venous changes


Vasculature

Source: Menon & Kim, TICS


Measuring Deoxyhemoglobin

• fMRI measurements are of amount of

oxyhemoglobin per voxels in Venus pool

• We assume that amount of deoxygenated

hemoglobin in vein (and oxyhemoglobin in

later stage) is predictive of neuronal

activity


BOLD signal

Source: Doug Noll’s primer



Stimulus to BOLD

Source: Arthurs & Boniface, 2002, Trends in Neurosciences


Basic Form of Hemodynamic Response

BOLD Time Course for repeated stimulus


A Simple Experiment

Intact

Objects

Scrambled

Objects

Blank

Screen

TIME

Condition changes every 16 seconds (8 volumes per Block), 17 block

One volume (12 slices) every 2 seconds

for 272 seconds (4 minutes, 32 seconds)

Lateral Occipital Complex

• responds when subject

views objects

What data do we start with

• 12 slices * 64 voxels x 64 voxels = 49,152 voxels

• Each voxel has 136 time points

• Therefore, for each run, we have 6.7 million data points

• We often have several runs for each experiment

…

Predicted Responses • It takes about 5 sec for the blood to catch up with the brain

• We can model the predicted activation in one of following ways: – Block activation model stimuli

1. shift the boxcar by approximately 5 seconds (2 images x 2 seconds/image = 4 sec, close enough)

2. convolve the boxcar with the hemodynamic response to model the shape of the true function as well as the delay

– Event related stimuli

– Rest activation (non-model based) response

PREDICTED ACTIVATION IN VISUAL AREA

BOXCAR

SHIFTED

CONVOLVED

WITH HRF

PREDICTED ACTIVATION IN OBJECT AREA

Why do we need stats? • move the cursor over different areas and see if any of the

time courses look interesting: Here for visual activation

Slice 9, Voxel 1, 0 Slice 9, Voxel 0, 0

Even where there’s no brain, there’s noise

Slice 9, Voxel 9, 27

Here’s a voxel that responds well whenever there’s visual stimulation


Here’s one that responds well whenever there’s intact objects


Here’s a couple that sort of show the right pattern but is it “real”?



BOLD signal in each voxel

time

Mxy

Signal

Mo sin

T2* task

T2* control

TEoptimum

Stask

Scontrol

S


Statistical Approaches in a Nutshell t-tests

• compare activation levels between two conditions

correlations

• model activation and see whether any areas show a similar pattern

Fourier analysis

• Do a Fourier analysis to see if there is energy at your paradigm frequency

Fourier analysis images

from Huettel, Song &

McCarthy, 2004,

Functional Magnetic

Resonance Imaging

Effect of Thresholds

r = 0

0% of variance

p < 1

r = .24

6% of variance

p < .05

r = .50

25% of variance

p < .000001

r = .40

16% of variance

p < .000001

r = .80

64% of variance

p < 10-33

Types of Errors

Slide modified from Duke course

Is the region truly active?

Do

es o

ur

stat

tes

t in

dic

ate

that

th

e re

gio

n i

s ac

tive?

Yes

N

o

Yes No

HIT Type I

Error

Type II

Error

Correct

Rejection

p value:

probability of a Type I error

e.g., p <.05

“There is less than a 5%

probability that a voxel our

stats have declared as

“active” is in reality NOT

active

Complications

• Not only is it hard to determine what’s real,

but there are all sorts of statistical problems

r = .24

6% of variance

p < .05

What’s wrong with these data? Potential problems: 1. data may be contaminated

by artifacts (e.g., head

motion, breathing artifacts)

2. .05 * 49,152 = 2457

“significant” voxels by

chance alone

3. adjacent voxels uncorrelated

with each other; adjacent

time points uncorrelated

with one another) are false

Source: Karl Friston

Estimating Parameters of Statistical Model

• FSL and SPM uses the Generalized Linear Model (GLM) to make parameter estimates.

Y = X . β + ε

Observed data Design matrix – model

formed of several

components which explain

the observed data

Parameters defining the

contribution of each

component of the

design matrix to the

model.

These are estimated so

as to minimize the

error, and are used to

generate the contrasts

between conditions

Error - the

difference

between the

observed

data and the

model

defined by

Xβ.

Source: http://www.fil.ion.ucl.ac.uk/spm/doc/mfd/GLM.ppt

Hypothesis vs. Data-Driven Approaches

Hypothesis-driven Examples: t-tests, correlations, general linear model (GLM)

• a priori model of activation is suggested

• data is checked to see how closely it matches components of the model

• most commonly used approach

Data-driven Example: Independent Component Analysis (ICA)

• no prior hypotheses are necessary

• multivariate techniques determine the patterns across all voxels

• can be used to validate a model

• can be inspected to see if there are things happening in your data that

you didn’t predict

• can be used to identify confounds (e.g., head motion)

Example of ICA

• hardest part is sorting the many components into meaningful ones vs. artifacts

• “fingerprints” may help

Study Design & Experimental Setup •Study Design

- Normative data in normal subjects in necessary

•Experimental Setup

- Blocked versus Event-Related Design

– Habituation, expectation, fatigue

•Data Acquisition

-Stimulus Pacing

-Matching difficulty level

•Post-Processing

-Performance monitoring

•Statistical Analysis and Validation

•Interpreting Results

- eg; Presurgical fMRI:

•Report

Major components of post-processing

and Analysis 1. Quality control (data free from noise and artifacts)

2. Motion correction

3. Slice timing correction

4. Spatial normalization (alignment into common spatial

framework)

5. Spatial smoothing

6. Temporal filtering

7. Statistical modeling (GLM & data fitting)

8. Statistical Inference (estimation of statistical

significance)

9. Visualization

Software

package for

fMRI analysis • SPM: include connectivity modeling tools, psychophysioligical

interaction, Dynamic Causal Moleding

• FSL: novel modeling techniques (eg RANDOMISE modules), ICA

for resting-state, DTI analysis, FSLview & probabilistic Atlases,

enable computing clusters

• AFNI: Powerful visualization abilities, integrating volume and

cortical surfaces

• BrainVoyager: commercial on all computing platforms

• Freesurfer: for cortical surfaces and anatomical

parcellations; can incorporate fMRI data from SPM/FSL

System requirement

1.5T

2%

signal

change

3T

4%

signal

change

Echo Planar Imaging: EPI Properties:

- One single excitation pulse ⇒ low rf energy deposition

(low SAR values).

- A train of gradient echoes, generated by very fast

switching of the defasing/refasing frequency encoding

gradients, at different values of the phase encoding

Gradient

- every echo generates one line of the image

- MR image acquired in less than 60 ms (128x 128 matrix)

- Sensitive to susceptibility effects

MRI Basis: T2*- contrast

Fast Imaging WHOLE HEAD

– SMALL voxels

• about 2.5 x 2.5 x 2.5 mm3

– High bandwidth

– EchoPlanarImaging:

• 128x128 (64x64) matrix

• TE ~ 50 ms (~ 30 ms @ 3T)

• up to 50 slices

• TR ~ 2 to 3 s

– Parallel imaging techniques

EPI Imaging

Important features of EPI

encoding:

• Adjacent points along kx are

acquired with a short Dt (= 5 μs).

(high bandwidth),

•Adjacent points along ky are

taken with a long Dt (= 500 μs)

(low bandwidth)

EPI: Practical issues

• Sensitive to field gradient artifacts

(eddy currents, susceptibility interfaces)

• geometric distortion (anatomical localization of activation

foci)

• signal drop-out (frontal and temporal poles)

5 mT/m 250 ms 10 mT/m 130 ms 16 mT/m 80 ms

Pre-requisities for fMRI analysis

• Probability and Statistics

• Computer programming: ATHLAB/python/UNIX

shell scripting

• Linear Algebra: GLM/image processing

• MRI: data acquisition/artifacts

• Neurophysiology & biophysics: Neuron activities

& blood flow/hemodynamic response

• Signal & Image processing: Fourier analysis based

processing

Thank You

M A Oghabian


introduction to mri - oghabian to neuroimaging... · 2015. 3. 13. · – can resolve brief single...

Documents