dsv2pulseq

Create Pulseq sequence files from Siemens dsv simulation files. The converter was only tested with the VE line dsv file layout. VB/VD files are not supported, as the simulator can not output RF phase data.
RF values might be differing slightly compared to the original sequence, as the dsv files contain RF values only on a 5us raster.

Dependencies & Installation

A Python environment with the dependencies can be installed with the provided yml file: conda env create -f dsv2pulseq.yml. Afterwards run: pip install ..
Unittests can be run with python -m unittest discover test/. This includes testing, whether the conversion is successful for each dsv dataset in the folder "test/test_data".

This package only depends on numpy and a slightly modified version of PyPulseq [1], which is added as a submodule.

Merging data to the original Siemens raw data file for retrospective reconstruction (see "Reconstruction of Pulseq data") requires the twixtools package (https://github.com/pehses/twixtools).

Sequence simulation

The sequence should be simulated in transversal orientation with phase-encode direction A->P and no FOV shift (which is the default). It has to be simulated with RF phase output (sim /RFP+). Mandatory dsv files are "_INF", "_RFD", "_RFP", "_GRX", "_GRY" and "_GRZ". Example data "gre_XXX.dsv" can be found in the test/test_data folder. These simulation files are from the Siemens MiniFLASH demo sequence.

Create Pulseq output

The conversion can be started by running:

dsv_to_pulseq.py -o OUT_FILE -r REF_VOLT IN_FILE_PREFIX

from the shell.

The IN_FILE_PREFIX is the prefix of the dsv files, e.g. "gre" for "gre_XXX.dsv". The OUT_FILE is the Pulseq output sequence file (default: "external.seq"). The reference voltage is the voltage the sequence was simulated with (default: 223.529007 V) The Pulseq sequence has the same orientation as the original sequence, when running in "XYZ in TRA" mode. This can fail, if a sequence block in the original Siemens sequence uses a different rotation matrix.

The conversion can also be done in Python by running:

from dsv2pulseq import read_dsv
seq = read_dsv('/path/to/dsv/dsv_prefix')
seq.write_pulseq('external.seq')

There is an experimental function to check the shapes of RF waveforms and gradients that plots the difference between the original and converted waveforms:

from dsv2pulseq import check_dsv
check_dsv('/path/to/dsv_original/dsv_prefix', '/path/to/dsv_pulseq/dsv_prefix')

Note that the RF and gradient waveforms might be slightly different due to fixed rotation matrices in the original sequence for some event blocks.

Reconstruction of Pulseq data

Data acquired with the converted Pulseq sequence can not be easily reconstructed, as the scan headers containing the meta information are missing. The data from the Pulseq sequence (IN_FILE_2) can be inserted into a raw data file acquired with the original sequence (IN_FILE_1) using twixtools (see dependencies). If twixtools is installed, the following command will insert the data:

dsv_to_pulseq.py IN_FILE_1 IN_FILE_2 -o OUT_FILE

The merged output file (OUT_FILE) can be used for rectrospective reconstruction at the scanner.

References

[1] Ravi, Keerthi, Sairam Geethanath, and John Vaughan. "PyPulseq: A Python Package for MRI Pulse Sequence Design." Journal of Open Source Software 4.42 (2019): 1725., https://github.com/imr-framework/pypulseq