G16.4427 Practical MRI 1 - PowerPoint Presentation

Slide1

G16.4427 Practical MRI 1

Basic pulse sequencesSlide2

Gradient Echo (GRE)

A class of pulse sequences that is primarily used for fast scanning

3D volume imaging

Cardiac imaging

Gradient reversal on the frequency-encoded axis forms the echo

A readout

prephasing

gradient lobe

dephases

the spins, then they are

rephased

with a readout gradient with opposite polarity

Can be fast because the flip angle is less than 90°

Why does that allows GRE to be fast?Slide3

Gradient Echo (GRE)

A class of pulse sequences that is primarily used for fast scanning

3D volume imaging

Cardiac imaging

Gradient reversal on the frequency-encoded axis forms the echo

A readout

prephasing

gradient lobe

dephases

the spins, then they are

rephased

with a readout gradient with opposite polarity

Can be fast because the flip angle is less than 90°

Fast T

1

recovery



short TR can be used (e.g. 2-50 ms)Slide4

Small Flip-Angle RF Pulse

What property of the small flip angle RF pulse is evident from this illustration?

Bernstein et al

. (2004) Handbook

of MRI

Pulse SequencesSlide5

Example of a GRE Pulse Sequence

The peak of the GRE occurs when the area under the two gradient lobes is equal.

Bernstein et al

. (2004) Handbook

of MRI

Pulse SequencesSlide6

T2 and T

2

*

Dephasing

T

2

dephasing

:

Inherent to tissue typeMolecular environment

Magnetic fields constantly changing in timeT2

* dephasingImperfect static magnetic fieldAir pockets (e.g. lungs) in bodyMetal parts in body (e.g. stents, clips)Magnetic fields that are constant in time

All of this PLUS T2 dephasingSlide7

Transverse Relaxation

T

2

* is always shorter than T

2

Bernstein et al

. (2004) Handbook

of MRI

Pulse SequencesSlide8

Response to a Series of RF Excitations

The excitation pulse is the only RF pulse in each TR (unless preparation pulses are used)

With a sufficient number of excitation pulses,

M

z

reaches a steady state

GRE sequences can be classified by the response of the transverse magnetization

M

xy

Spoiled: if ~0 just before each excitation

Steady-state free precession (SSFP): if reaches a nonzero steady stateSlide9

Spoiling

Spoiling can be accomplished in different ways

The simplest method is to use TR ~ 5T

2

Practical only with interleaved multi-slice acquisitions

End-of-sequence gradient spoiler

Not effective at spoiling the transverse steady state

Spatially non uniform because gradients produce spatially varying fields

RF spoiling

Phase-cycle the RF excitation pulses according to a predetermined schedule (i.e. flip the magnetization down in a different direction each time)Slide10

RF Spoiling

Stripe pattern artifact due to the spatially varying field produced by the gradients. (e.g. when the phase-encoding gradient is used as a spoiler, so no phase rewinding lobe is used)

(Bright stripes are unspoiled regions)

Bernstein et al

. (2004) Handbook

of MRI

Pulse SequencesSlide11

RF Spoiling

Stripe pattern artifact due to the spatially varying field produced by the gradients. (e.g. when the phase-encoding gradient is used as a spoiler, so no phase rewinding lobe is used)

RF spoiling:

phase of the B

1

field for the

j

th

RF pulse in the rotating frame:

(equivalent to the phase twist imparted by the phase-encoding gradient)

The recommended value for the starting phase increment is ϕ

0

= 117

°

During each TR the received MR signal must be shifted by the same phase, so

that

k

-space data are consistent

(Bright stripes are unspoiled regions)

Bernstein et al

. (2004) Handbook

of MRI

Pulse SequencesSlide12

Steady State Mz

for Spoiled GRE

If the longitudinal magnetization at point A is

M

zA

, after the excitation pulse

M

zB

=

M

zAcosθIn the TR between points B and C, T1 relaxation occurs, so:

When a steady state is reached MzA =

M

zC

Bernstein et al

. (2004) Handbook

of MRI

Pulse SequencesSlide13

The signal

S

spoil

is caused by the gradient

rephasing

the FID at an echo time TE, so it is given by:

Which is equal to:

Ernst Angle

Richard Ernst

August 14, 1933

1991

Nobel Prize in Chemistry

“Ernst angle”

The flip angle that maximize the signal is:Slide14

SSFP-FID (FISP) And SSFP-Echo

Standard GRE with greater signal than spoiled pulse sequences

Often at the cost of less contrast

SSFP-Echo less used

Conditions for SSFP:

phase coherent (RF pulses have the

same phase, or sign alternation, in

the rotating frame)

TR < T

2

Accumulated phase is the same in

each TR (

 same gradient area)

If met, than steady states for both

M

z

and

M

xy

will be established

(A FID-like signal just after the RF and a

time-reversed just before each pulse)

Bernstein et al

. (2004) Handbook

of MRI

Pulse SequencesSlide15

SSFP-FID (FISP) And SSFP-Echo

Chavhan

GB et

al

. (2008)

Radiographics

vol. 28(4)

Phase-coherent RF pulses with same flip-angle and constant TR < T

2

 steady state

Post-excitation signal (S+), FID arising from most recent RF pulse

Echo reformation signal (S-) when residual echo is refocused at the time of the subsequent RF pulseSlide16

SSFP-FID And SSFP-Echo Signals

If TR >> T

2



Slide17

SSFP-FID And SSFP-Echo Signals

If TR >> T

2



If

θ

<< 1



(PD-weighting at low flip angles)Slide18

Balanced SSFP (True FISP)

For SSFP the gradient area on any axis must not vary among TR intervals

For Balanced SSFP the gradient area on any axis is zero during each TR

Peaks of SSFP-FID and SSFP-Echo combine at TE (coherent sum of two signals

The magnitude of the signal changes for sign alternated pulses

If the balanced SSFP signal is

rephased

in the center of the TR interval (i.e. TE = TR/2), the decay is governed by T

2

rather than T

2

*

decreasing TE can increase susceptibility weighting in balanced SSFP

(the contrary happens for spoiled GRE and SSFP-FID)

Used in practice because of greater signalSlide19

Balanced SSFP

Scheffler

K and

Lehnhardt

S (2003)

Eur

Radiol

vol. 13 Slide20

Artifacts of Balanced SSFP

In regions where a phase shift removes the sign alternation there is a signal loss

Banding artifact

Unwanted phase shifts are always present

Short TR (e.g. less than 7 ms) are needed

Question: are balanced SSFP easier or more difficult to implement at higher field strength?Slide21

Banding Artifacts in Balanced SSFP

Scheffler

K and

Lehnhardt

S (2003)

Eur

Radiol

vol. 13 Slide22

Examples of Banding ArtifactsSlide23

Artifacts of Balanced SSFP

In regions where a phase shift removes the sign alternation there is a signal loss

Banding artifact

Question:

for example what could cause a phase shift?

Unwanted phase shifts are always present

Short TR (e.g. less than 7 ms) are needed

Question: are balanced SSFP easier or more difficult to implement at higher field strength?Slide24

Artifacts of Balanced SSFP

In regions where a phase shift removes the sign alternation there is a signal loss

Banding artifact

Unwanted phase shifts are always present

Short TR (e.g. less than 7 ms) are needed

Question:

are balanced SSFP easier or more difficult to implement at higher field strength?Slide25

Artifacts of Balanced SSFP

In regions where a phase shift removes the sign alternation there is a signal loss

Banding artifact

Unwanted phase shifts are always present

Short TR (e.g. less than 7 ms) are needed

More difficult to implement at high field

Increased susceptibility variations

SAR associated with very short TRSlide26

Particular Cases of Balanced SSFP

For short TR (TR << T

2

< T

1

) the signal formula becomes:

Question:

what does the formula tells you about the signal from fluids in balanced SSFP images?Slide27

Particular Cases of Balanced SSFP

For short TR (TR << T

2

< T

1

) the signal formula becomes:

The signal is maximized for:

At flip angles ~ 90° becomes more highly T

2

/ T

1

weighted:

Max of nearly M

0

/2 when T

2

= T

1



extremely strong signal for a short TR pulse sequenceSlide28

Example

SSFP-FID and Spoiled GRE:

TR = 14 ms

TE = 6 ms

Balanced SSFP:

TR = 6 ms

TE = 3 msSlide29

Inversion Recovery (IR)

Pulse sequences with an inversion pulse followed by a time delay prior to an RF excitation

Produce images with T

1

-weighted contrast.

Why?Slide30

Inversion Recovery (IR)

Pulse sequences with an inversion pulse followed by a time delay prior to an RF excitation

Produce images with T

1

-weighted contrast.

Time delay is know as the inversion time (TI)

Consists of two parts:

Inversion pulse, spoiler gradient (optional), slice selection gradient (if selective inversion pulse)

A self-contained pulse sequence (e.g. GRE) after TI

Require long TR (2-11

s

) to preserve the contrast2D IR sequences more frequently usedBenefits from real rather than magnitude reconstructionWhy?Slide31

Inversion Recovery (IR)

Pulse sequences with an inversion pulse followed by a time delay prior to an RF excitation

Produce images with T

1

-weighted contrast.

Why?

Time delay is know as the inversion time (TI)

Consists of two parts:

Inversion pulse, spoiler gradient (optional), slice selection gradient (if selective inversion pulse)

A self-contained pulse sequence (e.g. GRE) after TI

Require long TR (2-11

s) to preserve the contrast2D IR sequences more frequently usedBenefits from real rather than magnitude reconstructionMz

ranges from –M0 and +M0 

increased tissue contrastSlide32

Diagram of IR Pulse Sequence

Besides T1-weighted images, what is another application of IR pulse sequences that we mentioned during a previous lecture?

Bernstein et al

. (2004) Handbook

of MRI

Pulse SequencesSlide33

Principles of IR

Immediately after the inversion pulse:

During the time interval TI

(for long TR)

If

θ

inv

= 180°:

If

θ

inv

= 90°:

Saturation Recovery

(SR)Slide34

IR and SR Curves

The TI value that nulls the longitudinal magnetization is called the “

nulling

time” or “zero-crossing point”

SR

IR

nulling

time

Bernstein et al

. (2004) Handbook

of MRI

Pulse SequencesSlide35

Examples of IR Applications

T

1

mapping

A series of IR images are acquired from the same location with different TI (everything else the same)

Long TR used to avoid signal saturation

Non-linear fitting (for magnitude IR, first need to obtain the zero-crossing and negate signals before it)

Lipid suppression (STIR)

Improves contrast for lesions embedded in fat (e.g. edema in bone marrow), as lipids appear bright like many lesions in post-contrast

Water signal loss (any tissue with T

1

similar to fat)Long acquisition timeSlide36

Radiofrequency Spin Echo (SE)

Formed by an excitation pulse and one or more refocusing pulse

Usually a

90° pulse followed by 180° pulse

Typically 2D mode using interleaved

multislice

Allows to obtain a specific contrast weighting

Greater immunity to off-resonance artifacts

Why?Slide37

Radiofrequency Spin Echo (SE)

Formed by an excitation pulse and one (or more in multi-echo SE) refocusing pulse

Usually a

90° pulse followed by 180° pulse

Typically 2D mode using interleaved

multislice

Allows to obtain a specific contrast weighting

Greater immunity to off-resonance artifacts because of the 180

°

refocusing pulse

As T

2 > T2*  heavily T

2-weighted images possible with long TE without much signal loss (dephasing)Only a single phase-encoding step in any TR intervalSlide38

Single-Echo SE

Bernstein et al

. (2004) Handbook

of MRI

Pulse SequencesSlide39

Determination of TE

The gradient area on the frequency-encoding axis determines the temporal location of the peak of the echo (when the area under readout gradient balances the area of the

prephasing

gradient lobe)

Sometimes

Δ

is nonzero due to systems imperfections (e.g. eddy currents that shift gradient lobes)

What is the effect?

Bernstein et al

. (2004) Handbook

of MRI

Pulse SequencesSlide40

Determination of TE

The gradient area on the frequency-encoding axis determines the temporal location of the peak of the echo (when the area under readout gradient balances the area of the

prephasing

gradient lobe)

Sometimes

Δ

is nonzero due to systems imperfections (e.g. eddy currents that shift gradient lobes)



The signal will have some T

2

* weighting

Note: some specialized sequences use

nonzero

Δ

intentionally

Bernstein et al

. (2004) Handbook

of MRI

Pulse SequencesSlide41

Partial-Echo SE

What differences do you notice?

Bernstein et al

. (2004) Handbook

of MRI

Pulse SequencesSlide42

Partial-Echo SE

The peak of the echo (not the center of the readout) occurs when the RF spin would have refocused in the absence of imaging gradients

Used to avoid T

2

* weighting of the signal and reduce minimum TE

Achieved by reducing the area of the

prephasing

lobe

Image reconstruction with partial Fourier methods

Bernstein et al

. (2004) Handbook

of MRI

Pulse SequencesSlide43

Signal Formula for SE

M

xy

negligible

(TR >> T2, or

spoiler gradient)

= 90°

= 180°

M

zA

short pulse (no T

1

relaxation

between A and B, or C and D)

Bernstein et al

. (2004) Handbook

of MRI

Pulse SequencesSlide44

Multi-Echo SE

The transverse magnetization can be repeatedly refocused into subsequent

SEs

by playing additional RF refocusing pulse

The series of echoes is called an echo train

Each echo number fits its own independent

k

-space

The length of the echo train is limited by T

2

decayIn most cases we are interested in 2 echoes (an early and a late one).

Question: if TR is long, what contrast will have the 2 resulting images?Slide45

Multi-Echo SE

The transverse magnetization can be repeatedly refocused into subsequent

SEs

by playing additional RF refocusing pulse

The series of echoes is called an echo train

Each echo number fits its own independent

k

-space

The length of the echo train is limited by T

2

decayIn most cases we are interested in 2 echoes (an early and a late one).

 if TR is long, the two images will be PD- and T2-weighted, respectivelySlide46

Example of Dual-Echo SE Acquisition

Proton density-weighted

TE/TR = 17/2200 ms

T

2

-weighted

TE/TR = 80/2200 msSlide47

Dual-Echo SE

Bernstein et al

. (2004) Handbook

of MRI

Pulse SequencesSlide48

T2-Mapping

It is a common application of acquiring longer echo trains (otherwise more than two echoes per TR are rarely acquired in MRI)

In theory we can acquire long echo train of

SEs

and fit the signal intensity at each pixel to calculate T

2

In practice there are systematic errors that make it difficult to fit a

monoexponential

decay curve

Variable flip angle across slice profile

Stimulated echoes can introduce unwanted T

1-weighting variations into the echo-train signalsIf magnitude reconstruction is used, the noise floor has nonzero mean leading to incorrectly larger T2 valuesSlide49

Any questions?Slide50

See you on Thursday!