CLI Tutorial

The software has two command line interfaces: dcmstack and nitool. The dcmstack command is used for converting DICOM data to Nifti files with the optional DcmMeta extension. The nitool command is used to work with these exteneded Nifti files.

Advanced Conversion

While the dcmstack command has many options, the defaults should do the right thing in most scenarios. To see a complete list of the command line options (with brief descriptions) use the -h option.

Embedding Meta Data

If the –embed option is used, all of the meta data in the source DICOM files will be extracted and summarized into a DcmMeta extension, which is then embedded into the Nifti header. The meta data keys are the keywords from the DICOM standard. For details on the DcmMeta extension see DcmMeta Extension.

The meta data is filtered using regular expressions to reduce the chance of including PHI (Private Health Information). There are two types of regular expressions used for filtering: ‘exclude’ and ‘include’ expressions. Any meta data where the key matches an exclude expression will be excluded, unless it also matches an include expression. That is to say that the include expressions override the exclude expressions.

To see the list of the default regular expressions use the –default-regexes option. To add an additional exclude expression use –exclude-regex (-e) option and to add an additional include expression use the –include-regex (-i) option.

By default, any private DICOM elements are ignored unless there is a “translator” for that element. To see a list of available translators use the –list-translators (-l) option. To disable a specific translator use the –disable-translator option. To include private elements that don’t have a translator use the –extract-private option.

IT IS YOUR RESPONSABILITY TO KNOW IF THERE IS PRIVATE HEALTH INFORMATION IN THE RESULTING FILE AND TREAT SUCH FILES APPROPRIATELY.

Output Names and Grouping

All DICOM files from the same series will grouped into a stack together. The output file name is determined by a Python format string that is formatted with the meta data. This can be specified with the –output-format option. By default the program will try to figure out an appropriate format string for the available meta data. Generally this will be the ‘SeriesNumber’ followed by the ‘ProtocolName’ or ‘SeriesDescription’ (or just the word “series”).

Ordering Time and Vector Dimensions

In addition to the three spatial dimensions, Nifti images can have time and (less commonly) vector dimensions. By default, the software will try to guess the appropriate meta data key for sorting the time dimension. If you would like to specify the meta data key, or stack along the vector dimension, you can do so with the –time-var (-t) and –vector-var (-v) options. Both options take a meta data key as an argument.

If there isn’t an attribute that can be used with a simple ascending order to sort along these dimensions, the –time-order or –vector-order options can be used. The argument to the option should be a text file with one value per line corresponding to the sorted order to use.

Creating Uncompressed Niftis

By default the output Nifti files will be compressed, and thus have the extension ‘.nii.gz’. Almost every program that can read Nifti files will still read them if they are compressed. To override this behavior you can use the –output-ext option.

Handling Bad Data

Valid DICOM files should have a specific preamble (an initial byte pattern) to identify them as a DICOM file. It is not uncommon to come across files that are missing this preamble but are otherwise valid (generally due to bad software). You can force dcmstack to try to read these files using the –force-read option.

With some data sets (generally EPI) slices can be missing their pixel data due to an error in the reconstruction. Using the –allow-dummies option will allow these files and fill the corresponding slice with the maximum possible value (i.e. 65535 for uint16).

Voxel Order

While the affine transform stored in the Nifti allows a mapping from voxel indices to patient space, some programs do not make use of the affine information. To provide a similar orientation in these programs we reorder voxels in the same manner as dcm2nii. This results in the positive row, column, and slice directions pointing toward the left, anterior, and superior (LAS) patient directions. This can be overridden with the –voxel-order option.

Working with Extended Nifti Files

The nitool command can be used to perform various tasks with the extended Nifti files (that is files with the the DcmMeta extension embedded). The nitool command exposes functionality through a number of sub commands. To see a list of sub commands with brief explanations use the -h option. To see detailed help for a specific subcommand use:

$ nitool <sub_command> -h

Looking Up Meta Data

To lookup meta data in an extended Nifti, use the lookup sub command. If you don’t specify a voxel index (using –index) then only constant meta data will be considered.

$ nitool lookup InversionTime 032-MPRAGE_AX_TI900_Pre.nii.gz
900.0
$ nitool lookup InstanceNumber 032-MPRAGE_AX_TI900_Pre.nii.gz
$ nitool lookup InstanceNumber --index 0,0,0 032-MPRAGE_AX_TI900_Pre.nii.gz
1
$ nitool lookup InstanceNumber --index 0,0,1 032-MPRAGE_AX_TI900_Pre.nii.gz
2

In the above example ‘InversionTime’ is contant across the Nifti and so an index is not required. The ‘InstanceNumber’ is not constant (it varies over slices) and thus only returns a result if an index is provided.

Merging and Splitting

To merge or split extended Nifti files use the merge and split sub commands. This will automatically create appropriate DcmMeta extensions for the output Nifti file(s). Both sub commands take a –dimension (-d) option to specify the index (zero based) of the dimension to split or merge along.

If the dimension is not specified to the split command, it will use the last dimension (vector, time, or slice). By default each output will have the same name as the input only with the index prepended (zero padded to three spaces). A format string can be passed with the option –output-format (-o) to override this behavior.

If the dimension is not specified for the merge command, it will use the last singular or missing dimension (slice, time, or vector). By default the inputs will be merged in the order they are provided on the command line. To instead sort the inputs using some meta data key use the –sort (-s) option.

Dumping and Embedding

The DcmMeta extension can be dumped using the dump sub command. If no destination path is given the result will print to stdout. A DcmMeta extension can be embedded into a Nifti file using the embed sub command. If no input file is given it will be read from stdin. For details about the DcmMeta extension see DcmMeta Extension.

Injecting Meta Data

If you want to inject some new meta data into the header extension you can use the inject command. You need to specify the meta data classification, key, and values. For example, to set a globally constant element with the key ‘PatientID’ and the value ‘Subject_001’:

$ nitool inject 032-MPRAGE_AX_TI900_Pre.nii.gz global const PatientID Subject_001