Pulseq #283

paquiteau · 2025-07-03T10:16:53Z

Closes Pulseq Integration ? #102 Add support for pulseq sequences #262

Changes proposed in this pull request:

Add a reader for pulseq sequences to extract kspace trajectories (formatted to MRI-nufft conventions) from pulseq files
- That would make for a nice functional test traj -> write_seq -> read_seq -> compare.
~~( Add a helper function to create gradients block for pypulseq)~~
- there is now a prepare_trajectory_for_seq that creates gradients waveforms with pre/re-winders. Generating the gradients event directly does not make much sense I feel.
Add a simple generation of a GRE sequence with pypulseq.
More sequence type support ?

chaithyagr · 2025-07-04T07:51:08Z

src/mrinufft/io/pulseq.py

+    # TODO ensure amplitudes are correct
+
+    # TODO deduplicate gradients (reduces the size of the file, speed up the sequence loading).
+    # TODO Are rotation of trajectories supported in pulseq?


yes :) https://github.com/imr-framework/pypulseq/blob/master/src/pypulseq/rotate.py
But i wonder why you need it. The rotation of trajectory to match the GOV rotation is done by the sequence at scanner.

chaithyagr · 2025-07-04T07:53:18Z

src/mrinufft/io/pulseq.py

+    # TODO add prewind/postwind stuff from #276
+    # TODO ensure amplitudes are correct
+
+    # TODO deduplicate gradients (reduces the size of the file, speed up the sequence loading).


i dont know what does this mean? No matter what, i think you need to give all the gradients anyway. thats the downside of pulseq really.. u cant exactly tell run this trajectory and rotate it and run again etc.

yes that was the idea behind this see for more detail: imr-framework/pypulseq#184

chaithyagr · 2025-07-04T07:54:21Z

src/mrinufft/io/pulseq.py

+    for grad_shot_id in grad_shot_ids:
+        pp.add_block() # RF pulse
+        pp.add_block() # pause for tuning echo time
+        pp.add_block() # Gradient block with ADC


Suggested change

pp.add_block() # Gradient block with ADC

pp.add_block() # Gradient block with ADC

pp.add_block() # Gradient spoiler block

pp.add_block() # pause for tuning for TR

paquiteau · 2025-07-04T19:33:56Z

So the example is running well for the 3D GRE case, I think we could cover more cases (A 2D slice selective GRE first).

working on top of #276 was the right choice to get very nice prewind/prephasor and rewind gradients.

Currently the sequence is generated for a given FOV and resolution, I don't know if Pulseq can provide some flexibility on this or not (e.g can you parametrized it at the console for instance, to change TE/TR/FA, or the resolution/fov)

@Daval-G do you have any comments ?

examples/example_pulseq.py

chaithyagr

Great work. Some major questions and minor reviews. I will need another round of review.

examples/example_pulseq.py

src/mrinufft/io/utils.py

chaithyagr · 2025-07-09T14:23:00Z

tests/test_io.py

+    seq = pulseq_gre_3D(
+        trajectory=kspace_loc, fov=FOV, img_size=img_size, TR=1000, TE=100, FA=90
+    )


Maybe you can also add checks to see if TE, TR and FA are satisfied?

chaithyagr · 2025-07-09T14:23:52Z

src/mrinufft/io/pulseq.py

+def _check_timings(seq):
+    # Check whether the timing of the sequence is correct
+    ok, error_report = seq.check_timing()
+    if ok:
+        return None
+    else:
+        warnings.warn("Timing check failed. Error listing follows:" + str(error_report))


Why a separate function really?

Cause I am planning to add other sequences (like a 2D GRE for instance) this would be common to all of them.

chaithyagr · 2025-07-09T14:24:28Z

src/mrinufft/io/pulseq.py

+def pulseq_gre_3D(
+    trajectory: NDArray,


Why not have a generic one for both 2D and 3D?

2D GRE have some specificities (specific pulse and slice selection that changes over time), it makes more sense to me to have it separately (some core will be refactor if needed)

chaithyagr · 2025-07-09T14:26:06Z

src/mrinufft/io/pulseq.py

+    delay_before_grad = pp.make_delay(
+        TE
+        - pp.calc_rf_center(rf_pulse)[0]
+        - rf_pulse.delay
+        - TE_pos * shot_duration
+        - skip_start * system.grad_raster_time
+    )
+    delay_end_TR = pp.make_delay(
+        TR
+        - pp.calc_duration(rf_pulse)
+        - delay_before_grad.delay
+        - full_grads.shape[1] * system.grad_raster_time
+    )


You might need to check for min_TE and min_TR. Can you add a util function to get it?
Basically for a fixed trajectory and spoil levels, give the minimum TE and TR possible.

this is handled by the check_timing function of pulseq (there is check for delay that can't be negative for instance)

src/mrinufft/io/pulseq.py

chaithyagr · 2025-09-22T14:17:04Z

This will ideally depend on #285 eventually?

Co-authored-by: alexis <[email protected]> Update examples/example_pulseq.py Co-authored-by: Chaithya G R <[email protected]>

More updates Fixed some bugs style fixes and [docs] [WIP] style fixes

Co-authored-by: Pierre-Antoine Comby <[email protected]>

fix: remove type annonation for optional deps. Otherwise, its a mess. Acquisition Configuration (#285) * feat!: acquisition Config. This is a breaking change! * fmt: black * feat(test): use acquisition fixture for tests. * feat: docstrings * refactor!: use acq in display functions as well * fmt:black * fix: correct field name. * feat: move adc setup to acquisition level. * fix: example use the acquisition object * feat: example use acquisition. * fix: style * use correct units * fix: enum meta has evolved in python 3.12 * feat: use better default for grad raster time. * feat: more SI units and acq use * fix: don't modify API for write/read_trajectory * fmt * fix examples * update * fix: acq and display * feat: add context manager for the acquisition * update context manager * fmt: PEP604 type union fixes UP045 * [docs] update docstring for acquisition config.

paquiteau added the trajectories Issues concerning non-Cartesian trajectories label Jul 3, 2025

paquiteau force-pushed the pulseq branch from 488988b to 73c0933 Compare July 3, 2025 13:41

chaithyagr reviewed Jul 4, 2025

View reviewed changes

paquiteau force-pushed the pulseq branch from 73c0933 to 8319832 Compare July 4, 2025 16:36

paquiteau marked this pull request as ready for review July 4, 2025 19:28

paquiteau requested a review from Daval-G July 4, 2025 19:34

paquiteau changed the base branch from master to add_prephasors July 4, 2025 19:36

alexisthual reviewed Jul 6, 2025

View reviewed changes

examples/example_pulseq.py Outdated Show resolved Hide resolved

chaithyagr reviewed Jul 9, 2025

View reviewed changes

examples/example_pulseq.py Outdated Show resolved Hide resolved

chaithyagr requested changes Jul 9, 2025

View reviewed changes

paquiteau mentioned this pull request Aug 19, 2025

Add prephasors, rewinders etc. #276

Merged

paquiteau linked an issue Aug 19, 2025 that may be closed by this pull request

Add support for pulseq sequences #262

Open

paquiteau changed the base branch from add_prephasors to master August 20, 2025 11:50

paquiteau mentioned this pull request Aug 20, 2025

Acquisition Configuration #285

Merged

3 tasks

paquiteau added 5 commits September 24, 2025 19:00

feat: add pypulseq reader

4a5b98e

setup GRE sequence from arbitrary trajectories.

bd03d1a

feat: add example for pulseq sequence

60874de

fix: format

e54dddd

feat: tweaks and add a test

55d99f3

paquiteau force-pushed the pulseq branch from e6cdf85 to 971c6a8 Compare September 24, 2025 17:26

paquiteau and others added 6 commits September 24, 2025 19:31

Update examples/example_pulseq.py

49575a6

Co-authored-by: alexis <[email protected]> Update examples/example_pulseq.py Co-authored-by: Chaithya G R <[email protected]>

fix: cleanup example.

69f7b64

use prepare_trajectory_for_seq for nsp files as well.

47a225d

cleanup

216815f

More updates Fixed some bugs style fixes and [docs] [WIP] style fixes

Apply suggestions from code review

2553b08

Co-authored-by: Pierre-Antoine Comby <[email protected]>

Undo changes

e61c4e0

paquiteau added 4 commits September 24, 2025 19:32

feat: add support for 2D GRE

dbe687f

fix: correct function name.

a7d49e9

feat: fmt

4408e6e

paquiteau force-pushed the pulseq branch from 971c6a8 to 2d3942d Compare September 24, 2025 17:33

fix: wrong rebase

b4c1413

paquiteau force-pushed the pulseq branch from 6537749 to b4c1413 Compare September 25, 2025 07:23

paquiteau force-pushed the master branch 3 times, most recently from 0ea2bc4 to d338014 Compare September 28, 2025 18:07

paquiteau added this to the v1.4 milestone Oct 9, 2025

paquiteau force-pushed the master branch from 532e177 to 912ba63 Compare November 14, 2025 13:48

Pulseq #283

Are you sure you want to change the base?

Pulseq #283

Uh oh!

Conversation

paquiteau commented Jul 3, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

paquiteau commented Jul 4, 2025

Uh oh!

Uh oh!

Uh oh!

chaithyagr left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

chaithyagr commented Sep 22, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

paquiteau commented Jul 3, 2025 •

edited

Loading