Welcome to the "Basic GRE" tutorial! It was initially developed for the Pulseq software demonstration and hands-on session at the ISMRM 2019 in Montreal.
This tutorial demonstrates how to expand a basic Gradient-Echo (GRE) sequence to an advanced GRE sequence with a desired steady-state magnetisation evolution needed to generate T1 contrast. It contains six steps from s01_GRE_tutorial_step0 to s06_GRE_tutorial_step5. We recommend using Meld software to highlight the changes at each step. The slide deck entitled 02_basic_gradient_echo.pdf shows the sequence diagrams of all steps and visualises the changes at each step.
The transverse magnetisation may persist from cycle to cycle in a GRE sequence with short repetition times (e.g. shorter than ~3*T2). Spoiling tries to approximate a situation that the steady-state magnetisation has no transverse components immediately before each RF pulse. Three methods may be used alone or in combination to spoil transverse magnetisation.
-
Long TR spoiling. When TR>>T2*, the transverse magnetisation will naturally decay to zero by the end of the cycle. Thus, any GRE sequence using TR values of several hundred milliseconds or longer will be "naturally" spoiled.
-
Gradient spoiling. In this method, spoiling is performed by applying the slice-selective (and sometimes readout) gradients at the end of each cycle just before the next RF pulse. Several gradient spoiling concepts have been tried in the MR history, with the strength of the spoiler gradient remaining constant or varied linearly or semi-randomly from TR to TR. In the present example a constant spoiler is used, and an appropriate phantom with a sufficiently long T2 demonstrates that gradient spoiling does not work unless the spoiler is so strong that the intrinsic diffusion weighting would kill the signal.
-
It is also demonstrated that it is necessary to refocus the phase encoding moment at the end of the TR cycle to avoid artefacts.
-
RF-spoiling. Here the phase of the RF carrier is changed according to a predefined formula from TR to TR. Using a completely randomised pattern of phase changes is not ideal because unintended spin clustering may occur, and the degree of spoiling may change from one interval to the next. A superior method is to increment the phase quadratically using a recursive formula. RF spoiling is always combined with a moderate constant spoiling in read and slice directions and phase-encoding refocusing. For further details about spoiling, please go to https://mriquestions.com/spoiling---what-and-how.html.
s01_GRE_tutorial_step0
s01 is a basic 2D slice-selective GRE sequence. Each TR of this sequence contains 5 blocks. The corresponding k-space is shown in Figure 1.
Figure 1 K-space of s01_GRE_tutorial_step0 sequence (6*6 encodes).
s02_GRE_tutorial_step1
s02 is a 2D slice-selective GRE sequence with three spoiler gradients added in slice-selective, readout and phase-encoding direction. Its k-space is shown in Figure 2.
Figure 2 K-space of s02_GRE_tutorial_step1 sequence (6*6 encodes).
s03_GRE_tutorial_step2
s03 is a 2D slice-selective GRE sequence with two spoiler gradients in slice-selective and readout directions and one rewinder gradient in the phase-encoding direction. It is built by altering the gyPost gradient in s01 to rephase in the phase-encoding direction for subsequent phase-encoding steps. The k-space is shown in Figure 3.
Figure 3 K-space of s03_GRE_tutorial_step2 sequence (6*6 encodes).
s04_GRE_tutorial_step3
s04 is built by adding RF-spoiling to s03. The RF-spoiling is achieved by quasi-randomly varying the RF phase offset in the i^th^ phase-encoding step, $\text{φ}_{\text{i}}\text{\ =\ mod}\left( \text{117\ \ }\left( \text{i}^{\text{2}}\text{\ +\ }\text{i}\text{\ }\text{+\ 2} \right)\text{,\ 360} \right)\text{}\frac{\text{π}}{\text{180}}$.
s05_GRE_tutorial_step4
In s05, the receiver phase offset is set to follow the transmitter
phase offset,
s06_GRE_tutorial_step5
s06 adds some dummy scans (i.e. plays out multiple cycles of the sequence without recording a signal) to s05 to establish (near) steady-state magnetisation in the GRE sequence prior to the beginning of the ADC recording. For more details about dummy scans, please go to https://mriquestions.com/dummy-cycles.html.
Pulseq Matlab repository: https://github.com/pulseq/pulseq
The source code of the demo sequences and reconstruction scripts is the core of this repository. Please download the files to your computer and make them available to Matlab (e.g. by saving them in a subdirectory inside your Pulseq-Matlab installation and adding them to the Matlab's path). There are two sub-directories:
- seq : contains example pulse sequences specifically prepared for this demo
- recon : contains the reconstruction scripts tested with the above sequences
- data : contains raw MR data in the Siemens TWIX format and the corresponding pulse sequences in the Pulseq format
We strongly recommend using a text compare tool like meld (see this Wikipedia page and compare sequences from subsequent steps to visualise the respective steps.
Check out the main Pulseq repository at https://github.com/pulseq/pulseq and familarising yourself with the code, example sequences, and reconstruction scripts (see pulseq/matlab/demoSeq and pulseq/matlab/demoRecon). If you already use Pulseq, consider updating to the current version.