QC and fix MRIs converted from other formats
There are 4 common medical image file formats: Analyze, Neuroimaging Informatics Technology Initiative (Nifti), Digital Imaging and Communication in Medicine (Dicom), Medical Image NetCDF (MINC). Medical image processing may start with some form of image conversion from Dicom to a format compatible with available pipelines. For bpipe, CIVET, or MAGeT brain, it is necessary that the files are in MINC format.
Additional information on file formats can be found here: Larobina, Michele, and Loredana Murino. “Medical Image File Formats.” Journal of Digital Imaging 27.2 (2014): 200–206. PMC. Web. 28 Feb. 2017.
Converting from different data types to MINC: http://www.bic.mni.mcgill.ca/ServicesSoftware/ConvertingOtherFileTypesToMINC
Dicom [.dcm] -> MINC: dcm2mnc
dcm2mnc [options] -usecoordinates <dicomdirectoryname> <output>
Analyze [.img, .hdr]-> MINC: ana2mnc
ana2mnc [options] <filename>.img <output>.mnc
Nifti [.nii, or separate image and header file]-> MINC: nii2mnc
nii2mnc [options] <filename>.nii <output>.mnc
All of the options for the commands are listed in the help files. To view the help files use the -help command.
ana2mnc -help
World coordinate systems for the different file formats:
Analyze: +x Left, +y Anterior, +z Superior (LAS)
MINC: +x Right, +y Anterior, +z Superior (RAS)
Nifti: RAS or LAS
In the case of RAS coordinate system:
Positive x: increases from left to right, positive y: increases from posterior to anterior, positive z: increases from inferior to superior.
Dicom file format is the most complete, containing information about the pixels, orientation, procedure and the patient. Nifti files storage is not standardised and can be stored either in the RAS or LAS coordinate system. Nifti header files may contain transformation matrices, qform and sform, that can be used to bring the data coordinates into real world coordinates, but sometimes this information is missing, or dependent on the software used in the conversion. Analyze file format does not store enough information to establish the image orientation.
For more information about Nifti orientation see: https://nifti.nimh.nih.gov/nifti-1/documentation/faq#Q14
Useful atlases to compare conversion against are located in the following folder:
/opt/quarantine/resources/
Symmetric atlas:
/opt/quarantine/resources/mni_icbm152_nlin_sym_09c_minc2/mni_icbm152_t1_tal_nlin_sym_09c.mnc
Asymmetric atlas:
/opt/quarantine/resources/mni_icbm152_nlin_asym_09c_minc2/mni_icbm152_t1_tal_nlin_asym_09c.mnc
To view the atlases use the MINC visualisation tools. The Display and Register commands are the most useful for this application. To view a single MRI use the Display command. To view two MRIs side by side as well as a superposition of the two use the Register command.
Load the appropriate modules
module load minc-toolkit minc-toolkit-extras
The Display command is used as follows:
Display minc_volume.mnc
The Register command is used as follows:
register minc_volume1.mnc minc_volume2.mnc
More information can be found at: http://www.bic.mni.mcgill.ca/ServicesSoftwareVisualization/HomePage
Below is an example of how to use the Display command to view one of the atlases and the register command to register two atlases together.
module load minc-toolkit minc-toolkit-extras
Display /opt/quarantine/resources/mni_icbm152_nlin_sym_09c_minc2/mni_icbm152_t1_tal_nlin_sym_09c.mnc
register /opt/quarantine/resources/mni_icbm152_nlin_sym_09c_minc2/mni_icbm152_t1_tal_nlin_sym_09c.mnc /opt/quarantine/resources/mni_icbm152_nlin_asym_09c_minc2/mni_icbm152_t1_tal_nlin_asym_09c.mnc
The output should look as follows:
Fig. 1
If you are checking multiple images against the atlas, it might be useful to set the atlas location as a variable.
export var_name=location/of/atlas/atlas.mnc
For example:
export MNIMODEL=/opt/quarantine/resources/mni_icbm152_nlin_sym_09c_minc2/mni_icbm152_t1_tal_nlin_sym_09c.mnc
Then you can use the Register command with this variable.
register myfile.mnc $MNIMODEL
Following image file conversion, the orientation of the volume may not be identical to the atlas image. In this case the processing pipelines will not function properly. Below are two example images of instances where the converted image does not match the atlas.
Fig. 2 shows an instance where the axial, sagittal, and coronal orientations match but need to be rotated.
Fig. 2
Fig. 3 shows an instance where the axial, sagittal, coronal orientations do not match.
Fig. 3
Before attempting to fix any image, contact the provider and inquire if you can get the original, dicom format, MRIs. If the original files are not available, information about which converter was used to transform the images to Nifti or ANALYZE would be useful in determining the orientation. If none of this information is available from the provider, continue to possible fixes.
Two commands can be used to rotate the MR volumes: volflip, volmash. For a full explanation, as well as a list of options, see the help files. Volflip can be used if the order of the orientations is correct but some of the slice orientation need to be rotated, as is seen in Fig. 2. Volmash is useful when the orientations are not in the proper order, as exemplified in Fig. 3.
The following will make the axial orientation of Fig. 2 match the atlas.
volflip -y input.mnc output.mnc
Fig. 4
The following will swap the orientations of Fig.3 such that they match the atlas.
volmash -swap xz input.mnc temp1.mnc
volmash -swap xy temp1.mnc output.mnc
Fig. 5
The general rules for volmash are:
volmash -swap xy -OR- volmash -swap yx : swap bottom two orientations
volmash -swap xz -OR- volmash -swap zx : swap top two orientations
volmash -swap yz -OR- volmash -swap zy : swap top and bottom orientations
Edited images can be checked against the atlas using the Register command. To perform a thorough comparison, and ensure that all of the changes are correct, register your MRI to MNI space before comparing it to the atlas. This can be done using the bestlinreg command.
The form of the command is:
bestlinreg_g -nmi -lsq12 -target_mask $MNI_MASK source.mnc $MNIMODEL transform.xfm [optional transformed source.mnc]
The MNI mask is located here in the following folder:
/opt/quarantine/resources/mni_icbm152_nlin_sym_09c_minc2/mni_icbm152_t1_tal_nlin_sym_09c_mask.mnc
And can be set as a variable using the following command:
export MNI_MASK=/opt/quarantine/resources/mni_icbm152_nlin_sym_09c_minc2/mni_icbm152_t1_tal_nlin_sym_09c_mask.mnc
The following command will register the image in Fig. 5 to MNI space:
bestlinreg_g -nmi -lsq12 -target_mask $MNI_MASK input.mnc $MNIMODEL output_reg.xfm output_reg.mnc
It should now be easier to make comparisons to the atlas.
register output_reg.mnc $MNIMODEL
Fig. 6
One way to ensure that the orientation is correct is to check for brain asymmetries. In particular, look at the frontal and occipital regions of the brain. In a correctly oriented image, the anterior part of the right hemisphere should protrude across the mid-line. In the posterior part of the brain the left hemisphere should protrude past the mid-line. The right frontal hemisphere is wider than the left and the left occipital hemisphere is wider than the right.
Fig. 7
Additional information on Petalia and Yakovlevian torque can be found here: Toga, Arthur W., and Paul M. Thompson. "Mapping brain asymmetry." Nature Reviews Neuroscience 4, no. 1 (2003): 37-48.
When comparing against the atlas, first make sure that the locations are synced. This can be done by toggling the "Not Synced" button in the Register window to show "Synced". Then scroll through the slices and check for brain asymmetries. An example is shown in Fig. 8.
Fig. 8
