Unwrapping of 3D and 4D datasets. Coil combination of 5D datasets.
Ordered from simple to more involved approaches.
Download link for mritools as plain text: https://github.com/korbinian90/CompileMRI.jl/releases
Note: It might not work on every Linux distribution due to shared library incompatibilities (#16)
If a compiled MacOS version is required, see Known-issues/MacOS, but the suggested way is Option 2
Option 2: Julia command line usage (every OS)
Identical usage to the standalone version with better compatibility.
- Install julia
- Run romeo.jl command line script
Option 3: Run in Neurodesk (every OS)
Neurodesk is an analysis environment for reproducible neuroimaging running in a docker container. It comes with a range of useful tools preinstalled, including ROMEO. (https://neurodesk.github.io/)
Option 4: Use the julia package ROMEO.jl (every OS)
All the flexibility, but requires handling phase offsets and B0 calculation yourself (see MriResearchTools.jl)
Follow the steps in https://github.com/korbinian90/CompileMRI.jl
ROMEO: Dymerska, B., Eckstein, K., Bachrata, B., Siow, B., Trattnig, S., Shmueli, K., Robinson, S.D., 2020. Phase Unwrapping with a Rapid Opensource Minimum Spanning TreE AlgOrithm (ROMEO). Magnetic Resonance in Medicine. https://doi.org/10.1002/mrm.28563
MCPC-3D-S Coil Combination: Eckstein, K., Dymerska, B., Bachrata, B., Bogner, W., Poljanc, K., Trattnig, S., Robinson, S.D., 2018. Computationally Efficient Combination of Multi-channel Phase Data From Multi-echo Acquisitions (ASPIRE). Magnetic Resonance in Medicine 79, 2996–3006. https://doi.org/10.1002/mrm.26963
Phase (and optionally Magnitude) images in NIfTI fileformat.
For multi-echo/multi-timepoint data, 4D-NIfTI files are used with echoes/timepoints in the 4th dimension.
Individual 3D files can be merged into 4D files using fslmerge.
If 5D-NIfTI datasets with channels in the 5th dimension are given, coil combination is performed as first step.
ROMEO is a command line application. The binary is in the folder mritools/bin
Example usage for B0-mapping with a 5-echo scan with TE = [3,6,9,12,15] ms:
$ romeo ph.nii -m mag.ii -B -t 3:3:15 -o outputdir
Example usage for single-echo data:
$ romeo ph.nii -m mag.ii -k nomask -o outputdir
Example for multiple time points with identical echo time (fMRI):
$ romeo ph.nii -m mag.ii -k nomask -t epi -o outputdir
Example usage for a 3-echo scan with TE = [3,6,9] ms:
$ romeo ph.nii -m mag.ii -t [3,6,9] -o outputdir
Note that echo times are required for unwrapping multi-echo data.
If multi-echo data is available, supplying ROMEO with multi-echo information should improve the unwrapping accuracy. The same is true for magnitude information.
Coil combination will be automatically performed for 5D datasets using MCPC-3D-S. The echoes have to be in the 4th dimension and the channels in the 5th dimension. For bipolar datasets use --phase-offset-correction bipolar
as additional argument (bipolar correction requires >= 3 echoes).
4D data with an equal echo time for all volumes should be unwrapped as 4D for best accuracy and temporal stability. The echo times can be set to -t epi
.
In certain cases, the phase of the first echo/time-point looks differently than the rest of the acquisition, which can occur due to flow compensation of only the first echo or not having reached the steady state in fMRI. This might cause template unwrapping to fail, as the first echo is chosen as the template by default.
With the optional argument --template 2
, this can be changed to the second (or any other) echo/time-point.
If the multi-echo data contains large phase offsets (phase at echo time zero), default template unwrapping might fail. Setting the --individual-unwrapping
flag is a solution, as it performs spatial unwrapping for each echo instead. The computed B0 map is not corrected for remaining phase offsets.
For proper handling, the phase offsest can be removed using MCPC-3D-S with the option --phase-offset-correction
. This works for monopolar and bipolar data, already combined or uncombined channels. However, this requires "linear phase evolution". If the phase is already "corrupted" by other coil combination algorithms, it might not be possible to estimate and remove the phase offsets.
For datasets with disconnected regions, the --max-seeds
, --correct-regions
and --merge-regions
options might be of interest.
The option --max-seeds
creates multiple regions, which are unwrapped independently. The detected regions are written to the output file regions.nii
.
For acquisitions, in which fat and water signals mix, the assumption of linear phase evolution might be broken. Both temporal unwrapping and MCPC-3D-S phase offset removal depend on linear phase evolution and might produce incorrect results, leading to multi-echo unwrapping problems and remaining phase offsets in the resulting B0 map.
Using the --individual-unwrapping
flag might improve the unwrapping performance by disabling temporal unwrapping. Still, a calculated B0 map might contain unwanted phase offsets.
The input data is automatically rescaled to [-π;π]. For example, the input data can be given from [0;4095], which is rescaled to the range [-π;π]. In case the data is already in radians, this can be deactivated using the flag --no-rescale
. This might be necessary for phase difference data, where calculating the phase difference increases the range to [-2π;2π], but no rescaling to [-π;π] should occur.
$ ./bin/romeo
usage: <PROGRAM> [-p PHASE] [-m MAGNITUDE] [-o OUTPUT]
[-t ECHO-TIMES [ECHO-TIMES...]] [-k MASK [MASK...]]
[-u] [-e UNWRAP-ECHOES [UNWRAP-ECHOES...]]
[-w WEIGHTS] [-B]
[--phase-offset-correction [PHASE-OFFSET-CORRECTION]]
[--phase-offset-smoothing-sigma-mm PHASE-OFFSET-SMOOTHING-SIGMA-MM [PHASE-OFFSET-SMOOTHING-SIGMA-MM...]]
[--write-phase-offsets] [-i] [--template TEMPLATE]
[-N] [--no-rescale] [--threshold THRESHOLD] [-v] [-g]
[-q] [-Q] [-s MAX-SEEDS] [--merge-regions]
[--correct-regions] [--wrap-addition WRAP-ADDITION]
[--temporal-uncertain-unwrapping] [--version] [-h]
optional arguments:
-p, --phase PHASE The phase image that should be unwrapped
-m, --magnitude MAGNITUDE
The magnitude image (better unwrapping if
specified)
-o, --output OUTPUT The output path or filename (default:
"unwrapped.nii")
-t, --echo-times ECHO-TIMES [ECHO-TIMES...]
The echo times required for temporal
unwrapping specified in array or range syntax
(eg. "[1.5,3.0]" or "3.5:3.5:14"). For
identical echo times, "-t epi" can be used
with the possibility to specify the echo time
as e.g. "-t epi 5.3" (for B0 calculation).
-k, --mask MASK [MASK...]
nomask | qualitymask <threshold> | robustmask
| <mask_file>. <threshold>=0.1 for qualitymask
in [0;1] (default: ["robustmask"])
-u, --mask-unwrapped Apply the mask on the unwrapped result. If
mask is "nomask", sets it to "robustmask".
-e, --unwrap-echoes UNWRAP-ECHOES [UNWRAP-ECHOES...]
Load only the specified echoes from disk
(default: [":"])
-w, --weights WEIGHTS
romeo | romeo2 | romeo3 | romeo4 | romeo6 |
bestpath | <4d-weights-file> | <flags>.
<flags> are up to 6 bits to activate
individual weights (eg. "1010"). The weights
are (1)phasecoherence
(2)phasegradientcoherence (3)phaselinearity
(4)magcoherence (5)magweight (6)magweight2
(default: "romeo")
-B, --compute-B0 Calculate combined B0 map in [Hz]. This
activates MCPC3Ds phase offset correction
(monopolar) for multi-echo data.
--phase-offset-correction [PHASE-OFFSET-CORRECTION]
on | off | bipolar. Applies the MCPC3Ds method
to perform phase offset determination and
removal (for multi-echo). "bipolar" removes
eddy current artefacts (requires >= 3 echoes).
(default: "off", without arg: "on")
--phase-offset-smoothing-sigma-mm PHASE-OFFSET-SMOOTHING-SIGMA-MM [PHASE-OFFSET-SMOOTHING-SIGMA-MM...]
default: [7,7,7] Only applied if
phase-offset-correction is activated. The
given sigma size is divided by the voxel size
from the nifti phase file to obtain a
smoothing size in voxels. A value of [0,0,0]
deactivates phase offset smoothing (not
recommended).
--write-phase-offsets
Saves the estimated phase offsets to the
output folder
-i, --individual-unwrapping
Unwraps the echoes individually (not
temporal). This might be necessary if there is
large movement (timeseries) or
phase-offset-correction is not applicable.
--template TEMPLATE Template echo that is spatially unwrapped and
used for temporal unwrapping (type: Int64,
default: 1)
-N, --no-mmap Deactivate memory mapping. Memory mapping
might cause problems on network storage
--no-rescale Deactivate rescaling of input images. By
default the input phase is rescaled to the
range [-π;π]. This option allows inputting
already unwrapped phase images without
manually wrapping them first.
--threshold THRESHOLD
<maximum number of wraps>. Threshold the
unwrapped phase to the maximum number of wraps
and sets exceeding values to 0 (type: Float64,
default: Inf)
-v, --verbose verbose output messages
-g, --correct-global Phase is corrected to remove global n2π phase
offset. The median of phase values (inside
mask if given) is used to calculate the
correction term
-q, --write-quality Writes out the ROMEO quality map as a 3D image
with one value per voxel
-Q, --write-quality-all
Writes out an individual quality map for each
of the ROMEO weights.
-s, --max-seeds MAX-SEEDS
EXPERIMENTAL! Sets the maximum number of seeds
for unwrapping. Higher values allow more
seperated regions. (type: Int64, default: 1)
--merge-regions EXPERIMENTAL! Spatially merges neighboring
regions after unwrapping.
--correct-regions EXPERIMENTAL! Performed after merging. Brings
the median of each region closest to 0 (mod
2π).
--wrap-addition WRAP-ADDITION
[0;π] EXPERIMENTAL! Usually the true phase
difference of neighboring voxels cannot exceed
π to be able to unwrap them. This setting
increases the limit and uses 'linear
unwrapping' of 3 voxels in a line. Neighbors
can have (π + wrap-addition) phase difference.
(type: Float64, default: 0.0)
--temporal-uncertain-unwrapping
EXPERIMENTAL! Uses spatial unwrapping on
voxels that have high uncertainty values after
temporal unwrapping.
--version show version information and exit
-h, --help show this help message and exit
The sourcecode is available under ROMEO.jl.
The binaries are a standalone compiled version of RomeoApp.jl. The compilation scripts and recent releases are in CompileMRI.jl.
- single echo unwrapping with magnitude and default mask leads to ReadOnlyMemory error. Fixed in v1.4.1
- binary data type for mask not supported. Fixed in v2.0.1
- input scaling issue (with INT16). Fixed in v2.0.2
- quality map output is corrupted. Fixed in v3.1.1
To run romeo executables on MacOS, multiple files have to be flagged as save to execute. Additionally, the executable might only run on specific OS versions. You can try the newest MacOS executable, v3.2.2 and v3.1. This problem is still unsolved and no clear way how to improve the compatibility. Please look at the different ways to run it on MacOS above.
libstdc++
error (libstdc++.so.6: version `GLIBCXX_3.4.29' not found)- Segmentation fault (error code 139)
Problem: ROMEO crashes with version conflicts of shared dependencies when called within MATLAB via unix()
or system()
.
Reason: MATLAB runs other programs with a modified environment variable
Solution: clean LD_LIBRARY_PATH before execution and restore it afterwards
Example:
% remove matlab specific LD_LIBRARY_PATH
if isunix; paths = getenv('LD_LIBRARY_PATH'); setenv('LD_LIBRARY_PATH'); end
success = system(<romeo call>);
% restore paths
if isunix; setenv('LD_LIBRARY_PATH', paths); end
Feature requests and bug reports are welcome!