Skip to content

Commit 0259974

Browse files
committed
Updated sequence simulations material for 2020
1 parent 1fc5872 commit 0259974

16 files changed

+86
-38
lines changed

images/GRE_approach_equilibrium.gif

7.99 MB
Loading

images/SPGR_approach_equilibrium.gif

9.13 MB
Loading

images/SSFPanimation_-150.gif

576 KB
Loading

images/SSFPanimation_-90.gif

477 KB
Loading

images/SSFPanimation_0.gif

417 KB
Loading

images/SSFPanimation_45.gif

460 KB
Loading

images/approach_SS_gre.png

73.9 KB
Loading

images/approach_SS_spgr.png

76.8 KB
Loading

images/approach_ss_configs_v2.gif

1.84 MB
Loading

images/iso_sim_3d_100_isochroms.gif

-6.17 MB
Loading
8.03 MB
Loading

images/iso_sim_3d_single_isochrom.gif

4 MB
Loading

images/mpsi_vs_mk_v2.gif

739 KB
Loading

images/ssfp_plot.png

126 KB
Loading

teaching-sequence-sims.html

Lines changed: 43 additions & 19 deletions
Original file line numberDiff line numberDiff line change
@@ -4,60 +4,84 @@
44
<xmp theme=spacelab style=display:none;>
55
# RF pulse design and simulations: supplementary animations
66

7+
### Steady state calculations
8+
In lectures we derived a steady-state relationship for an SSFP sequence, which can be illustrated as below for a particular set of tissue and sequence parameters:
9+
10+
<img src="images/ssfp_plot.png" width="70%">
11+
12+
We can pick out some of these and animate the motion of a magnetization isochromat depending on the off-resonance dephasing angle (per TR) \\(\psi\\):
13+
14+
<img src="images/SSFPanimation_0.gif" width="20%">
15+
<img src="images/SSFPanimation_-90.gif" width="20%">
16+
<img src="images/SSFPanimation_45.gif" width="20%">
17+
<img src="images/SSFPanimation_-150.gif" width="20%">
18+
19+
20+
721
### Isochromat ensembles
822

923
A simple sequence of RF and gradient pulses, just one isochromat:
1024

11-
<img src="images/iso_sim_3d_1_isochroms.gif" width="70%">
25+
<img src="images/iso_sim_3d_single_isochrom.gif" width="70%">
1226

1327
A single isochromat can't characterise the system because in reality we have a
1428
continuous distribution of magnetization at different frequencies. Here's the
1529
same sequence with 100 isochromats:
1630

17-
<img src="images/iso_sim_3d_100_isochroms.gif" width="70%">
31+
<img src="images/iso_sim_3d_100_isochroms_STAGGERED.gif" width="100%">
32+
33+
In the diagram above the isochromats are spead out in the direction of the applied gradient - this is a physically realistic picture. We may also draw the vectors on top of one another so that we can directly compare their directions:
34+
35+
<img src="images/iso_sim_3d_100_isochroms.gif" width="100%">
36+
37+
1838

1939
### Simulation of the effect of a single gradient
2040

2141
Simulate transverse magnetization when a gradient is applied.
2242
First using 5 isochromats:
2343

24-
<img src="images/iso_sim_5_isochroms.gif" width="70%">
44+
<img src="images/iso_sim_5_isochroms.gif" width="60%">
2545

2646
You see that there are **spurious echoes** when all the magnetisation vectors are aligned. This is unphysical, and these disappear when the number of isochromats is increased. For example, using 100 isochromats:
2747

28-
<img src="images/iso_sim_100_isochroms.gif" width="70%">
48+
<img src="images/iso_sim_100_isochroms.gif" width="60%">
2949

3050
### Isochromat simulations including relaxation
3151

3252
Below is the behaviour of 100 isochromats during a gradient echo sequence
3353
without RF spoiling (i.e. spoiling phase = 0) for flip angle 30 degrees, TR=20ms, T1=1000ms and T2=100ms:
3454

35-
<img src="images/spgr_iso_sim_spoil0.gif" width="50%">
55+
<img src="images/GRE_approach_equilibrium.gif" width="100%">
56+
57+
We see that the magnetization slowly reaches the steady state (as a function of \\(\psi\\)) that we predicted analytically.
58+
59+
The steady state doesn't have the properties
60+
that would be predicted by the Ernst equation. This can be seen by comparing the integrated signal over the voxel with the Ernst equation:
3661

37-
The magnetisation reaches a steady state but note that the transverse magnetization
38-
doesn't disappear after each pulse. The steady state doesn't have the properties
39-
that would be predicted by the Ernst equation. Below is the same sequence but with
40-
a shorter TR (5ms) - the dynamics are clearer here because relaxation is less (i.e. the
41-
lengths of the magnetization vectors don't get so small):
62+
<img src="images/approach_SS_gre.png" width="80%">
4263

43-
<img src="images/spgr_iso_sim_spoil0_tr5.gif" width="50%">
64+
The reason is that the TR is not short compared with the T2 so we have imperfect spoiling. This can be improved by using RF phase cycling, as below:
4465

45-
We can reduce the effect of coherent addition of transverse magnetization by
46-
using quadratic phase cycling, here shown for 117 degrees. Firstly with the longer TR=20ms:
66+
<img src="images/SPGR_approach_equilibrium.gif" width="100%">
4767

48-
<img src="images/spgr_iso_sim_spoil117_tr20.gif" width="50%">
68+
which has the following steady state behaviour:
4969

50-
and now with the shorter TR=5ms:
70+
<img src="images/approach_SS_spgr.png" width="80%">
5171

52-
<img src="images/spgr_iso_sim_spoil117_tr5.gif" width="50%">
5372

5473
### EPG representation
5574

56-
Magnetisation as a function of space (gradient induced phase offset) and then also as a function of k - the Fourier conjugate variable. This is a simple sequence of RF pulses and gradients
75+
Magnetisation as a function of space (gradient induced phase offset) and then also as a function of k - the Fourier conjugate variable. This is a simple sequence of RF pulses and gradients. We can see that RF pulses have the effect of rotating all isochromats, while gradients add a rotation that varies over the voxel.
76+
77+
<img src="images/approach_ss_configs_v2.gif" width="100%">
78+
79+
This can be looked at in the k-space domain:
80+
81+
<img src="images/mpsi_vs_mk_v2.gif" width="100%">
5782

58-
<img src="images/mpsi_vs_mk.gif" width="100%">
5983

60-
As time goes by, the number of non-zero Fourier terms needed to describe the magnetisation increases. This is because each successive gradient period adds new k-space terms to the mix.
84+
Here again we see that the RF pulse mixes Mx, My, Mz but doesn't change the spatial modulation while the gradients do. In the Fourier domain we see that RF pulses mix up coefficients *of a given spatial order*, whereas gradients shift to higher order coefficients.
6185
</xmp>
6286
<script src=http://strapdownjs.com/v/0.2/strapdown.js></script>
6387
</html>

teaching-sequence-sims.md

Lines changed: 43 additions & 19 deletions
Original file line numberDiff line numberDiff line change
@@ -1,56 +1,80 @@
11
# RF pulse design and simulations: supplementary animations
22

3+
### Steady state calculations
4+
In lectures we derived a steady-state relationship for an SSFP sequence, which can be illustrated as below for a particular set of tissue and sequence parameters:
5+
6+
<img src="images/ssfp_plot.png" width="70%">
7+
8+
We can pick out some of these and animate the motion of a magnetization isochromat depending on the off-resonance dephasing angle (per TR) \\(\psi\\):
9+
10+
<img src="images/SSFPanimation_0.gif" width="20%">
11+
<img src="images/SSFPanimation_-90.gif" width="20%">
12+
<img src="images/SSFPanimation_45.gif" width="20%">
13+
<img src="images/SSFPanimation_-150.gif" width="20%">
14+
15+
16+
317
### Isochromat ensembles
418

519
A simple sequence of RF and gradient pulses, just one isochromat:
620

7-
<img src="images/iso_sim_3d_1_isochroms.gif" width="70%">
21+
<img src="images/iso_sim_3d_single_isochrom.gif" width="70%">
822

923
A single isochromat can't characterise the system because in reality we have a
1024
continuous distribution of magnetization at different frequencies. Here's the
1125
same sequence with 100 isochromats:
1226

13-
<img src="images/iso_sim_3d_100_isochroms.gif" width="70%">
27+
<img src="images/iso_sim_3d_100_isochroms_STAGGERED.gif" width="100%">
28+
29+
In the diagram above the isochromats are spead out in the direction of the applied gradient - this is a physically realistic picture. We may also draw the vectors on top of one another so that we can directly compare their directions:
30+
31+
<img src="images/iso_sim_3d_100_isochroms.gif" width="100%">
32+
33+
1434

1535
### Simulation of the effect of a single gradient
1636

1737
Simulate transverse magnetization when a gradient is applied.
1838
First using 5 isochromats:
1939

20-
<img src="images/iso_sim_5_isochroms.gif" width="70%">
40+
<img src="images/iso_sim_5_isochroms.gif" width="60%">
2141

2242
You see that there are **spurious echoes** when all the magnetisation vectors are aligned. This is unphysical, and these disappear when the number of isochromats is increased. For example, using 100 isochromats:
2343

24-
<img src="images/iso_sim_100_isochroms.gif" width="70%">
44+
<img src="images/iso_sim_100_isochroms.gif" width="60%">
2545

2646
### Isochromat simulations including relaxation
2747

2848
Below is the behaviour of 100 isochromats during a gradient echo sequence
2949
without RF spoiling (i.e. spoiling phase = 0) for flip angle 30 degrees, TR=20ms, T1=1000ms and T2=100ms:
3050

31-
<img src="images/spgr_iso_sim_spoil0.gif" width="50%">
51+
<img src="images/GRE_approach_equilibrium.gif" width="100%">
52+
53+
We see that the magnetization slowly reaches the steady state (as a function of \\(\psi\\)) that we predicted analytically.
54+
55+
The steady state doesn't have the properties
56+
that would be predicted by the Ernst equation. This can be seen by comparing the integrated signal over the voxel with the Ernst equation:
3257

33-
The magnetisation reaches a steady state but note that the transverse magnetization
34-
doesn't disappear after each pulse. The steady state doesn't have the properties
35-
that would be predicted by the Ernst equation. Below is the same sequence but with
36-
a shorter TR (5ms) - the dynamics are clearer here because relaxation is less (i.e. the
37-
lengths of the magnetization vectors don't get so small):
58+
<img src="images/approach_SS_gre.png" width="80%">
3859

39-
<img src="images/spgr_iso_sim_spoil0_tr5.gif" width="50%">
60+
The reason is that the TR is not short compared with the T2 so we have imperfect spoiling. This can be improved by using RF phase cycling, as below:
4061

41-
We can reduce the effect of coherent addition of transverse magnetization by
42-
using quadratic phase cycling, here shown for 117 degrees. Firstly with the longer TR=20ms:
62+
<img src="images/SPGR_approach_equilibrium.gif" width="100%">
4363

44-
<img src="images/spgr_iso_sim_spoil117_tr20.gif" width="50%">
64+
which has the following steady state behaviour:
4565

46-
and now with the shorter TR=5ms:
66+
<img src="images/approach_SS_spgr.png" width="80%">
4767

48-
<img src="images/spgr_iso_sim_spoil117_tr5.gif" width="50%">
4968

5069
### EPG representation
5170

52-
Magnetisation as a function of space (gradient induced phase offset) and then also as a function of k - the Fourier conjugate variable. This is a simple sequence of RF pulses and gradients
71+
Magnetisation as a function of space (gradient induced phase offset) and then also as a function of k - the Fourier conjugate variable. This is a simple sequence of RF pulses and gradients. We can see that RF pulses have the effect of rotating all isochromats, while gradients add a rotation that varies over the voxel.
72+
73+
<img src="images/approach_ss_configs_v2.gif" width="100%">
74+
75+
This can be looked at in the k-space domain:
76+
77+
<img src="images/mpsi_vs_mk_v2.gif" width="100%">
5378

54-
<img src="images/mpsi_vs_mk.gif" width="100%">
5579

56-
As time goes by, the number of non-zero Fourier terms needed to describe the magnetisation increases. This is because each successive gradient period adds new k-space terms to the mix.
80+
Here again we see that the RF pulse mixes Mx, My, Mz but doesn't change the spatial modulation while the gradients do. In the Fourier domain we see that RF pulses mix up coefficients *of a given spatial order*, whereas gradients shift to higher order coefficients.

0 commit comments

Comments
 (0)