Python utilities for converting from DICOM to BIDS and NDAR-compliant neuroimaging formats.
Python script which takes a directory tree containing imaging series from one or more subjects (eg T1w MPRAGE, BOLD EPI, Fieldmaps), converts the imaging data to Nifti-1 format with JSON metadata files (sidecars) and populates a Brain Imaging Data Structure (BIDS) which should pass the online BIDS validation tool (http://incf.github.io/bids-validator).
There are two options available for installation and running the BIDS conversion:
(This assumes you have Docker installed on your system (https://docs.docker.com/engine/installation/)
Simply pull the docker image of this app from Docker Hub and point to your DICOM folders as below:
docker pull rnair07/bidskit(This downloads the bidskit docker image to your system)
You could also skip the above step and directly run the command below instead and it will automatically pull the image for you + run the conversion.
docker run -it -v /PATH_TO_YOUR_RAW_DICOM_FOLDER/:/mnt rnair07/bidskit --indir=/mnt/DICOM --outdir=/mnt/BIDS
where PATH_TO_YOUR_RAW_DICOM is the root directory containing the mydicom folder as shown in the file structure below.
Clone the repository, add the resulting directory to your path and install dependencies mentioned below (Will upgrade this to a python setup soon).
% git clone https://github.com/jmtyszka/bidskit.git
Dependencies This release was developed under Python 3.5 (os, sys, argparse, subprocess, shutil, json, glob). Other dependencies include:
- pydicom 0.9.9 (latest version in PyPi)
- Chris Rorden's dcm2niix - the latest version at the time of writing is v1.0.20171103 (source or precompiled binaries)
Organize the source DICOM images into separate subject and subject-session directories within a root directory (mydicom in the example below, but the script default is simply dicom). The organization of DICOM files within each subject directory can follow a session-series heirarchy or by a simple flat directory containing all subject files. The conversion to Nifti-1 format and JSON sidecar generation is handled by dcm2niix, so whatever works for dcm2niix will hopefully work for dcm2bids.py. A typical DICOM directory tree might look something like the following (where "Ra0950" and "Ra0951" are subject IDs and "first", "second" are session names for each subject):
mydicom
└── Ra0950
└── first
├── IM-0001-0001.dcm
├── IM-0001-0002.dcm
...
└── second
...
└── Ra0951
└── first
├── IM-0001-0001.dcm
├── IM-0001-0002.dcm
...
└── second
...
The required command line arguments and defaults for dcm2bids.py can be displayed using:
% dcm2bids.py -h
usage: dcm2bids.py [-h] [-i INDIR] [-o OUTDIR] [--no-sessions]
Convert DICOM files to BIDS-compliant Nifty structure
optional arguments:
-h, --help show this help message and exit
-i INDIR, --indir INDIR
DICOM input directory with Subject/Session/Image
organization [dicom]
-o OUTDIR, --outdir OUTDIR
Output BIDS source directory [source]
--no-sessions Do not use session sub-directories
Note that the defaults for the input DICOM and output BIDS directories are dicom and source respectively. So the simplest possible setup would be to place subject DICOM folders within a directory called dicom and run dcm2bids.py from the parent directory of dicom. This would generate a BIDS source directory called source, a BIDS derivatives directory called derivatives with a conversion subdirectory containing a protocol translator JSON file and a working directory called work.
If you're using the Docker image, run the following:
docker run -it -v /PATH_TO_YOUR_RAW_DICOM_FOLDER/:/mnt rnair07/bidskit --indir=/mnt/dicom --outdir=/mnt/source
If you're running dcm2bids.py locally from source, you can use any of the following:
% dcm2bids.py % dcm2bids.py -i mydicom % dcm2bids.py -i mydicom -o mysource
The first pass conversion will create new translator dictionary (Protocol_Translator.json) in the root DICOM folder. This has been prefilled with the protocol series names from the DICOM header of all unique series detected in the original DICOM files. The command will also create the new BIDS directory containing a single temporary conversion directory containing Nifti images and JSON sidecars for all series in the source DICOM folder:
derivatives/
└── conversion/
├── Protocol_Translator.json
dicom/
└── Ra0950
└── first/
└── ...
└── second/
└── ...
└── Ra0951
└── first/
└── ...
└── second/
└── ...
source/
work/
└── conversion/
└── sub-Ra0950/
└── ses-first/
├── sub-Ra0950_ses-first_....nii.gz
├── sub-Ra0950_ses-first_....json
You can omit the use of session subdirectories if you only have one session per subject. Use the --no-sessions command line flag to achieve this (this feature is switched off by default):
% dcm2bids.py --no-sessions -i mydicom -o mybids
dcm2bids.py creates a JSON series name translator in the derivatives/conversion folder. You'll use this file to specific how you want individual series data to be renamed into the output BIDS source directory. Open the Protocol_Translator.json file in a text editor. Initially it will look something like the following, with the BIDS directory, filename suffix and IntendedFor fields set to their default values of "EXCLUDE_BIDS_Name", "EXCLUDE_BIDS_Directory" and "UNASSIGNED" (the double quotes are a JSON requirement):
{
"Localizer":[
"EXCLUDE_BIDS_Directory"
"EXCLUDE_BIDS_Name",
"UNASSSIGNED"
],
"rsBOLD_MB_1":[
"EXCLUDE_BIDS_Directory"
"EXCLUDE_BIDS_Name",
"UNASSSIGNED"
],
"T1_2":[
"EXCLUDE_BIDS_Directory"
"EXCLUDE_BIDS_Name",
"UNASSSIGNED"
],
"Fieldmap_rsBOLD":[
"EXCLUDE_BIDS_Directory"
"EXCLUDE_BIDS_Name",
"UNASSSIGNED"
],
...
}
The IntendedFor field is only relevant for fieldmap series and links the fieldmap to one or more EPI series for distortion correction.
Edit the BIDS directory and filename suffix entries for each series with the BIDS-compliant filename suffix (excluding the sub-xxxx_ses-xxxx_ prefix and any file extensions) and the BIDS purpose directory name (anat, func, fmap, etc). In the example above, this might look something like the following:
{
"Localizer":[
"EXCLUDE_BIDS_Directory",
"EXCLUDE_BIDS_Name",
"UNASSIGNED"
],
"rsBOLD_MB_1":[
"func",
"task-rest_acq-MB_run-01_bold",
"UNASSIGNED"
],
"T1_2":[
"anat",
"run-02_T1w",
"UNASSIGNED"
],
"Fieldmap_rsBOLD":[
"fmap",
"acq-rest",
["task-rest_acq-MB_run-01_bold", "task-rest_acq-MB_run-02_bold"]
],
...
}
Complete documentation for the BIDS standard, including appropriate filenaming conventions, can be found at http://bids.neuroimaging.io
The bidskit now has enough information to correctly organize the converted Nifti images and JSON sidecars into a BIDS directory tree. Any protocol series with a BIDS name or directory begining with "EXCLUDE" will be skipped (useful for excluding localizers, teleradiology acquisitions, etc from the final BIDS directory). Rerun the docker command or dcm2bids.py (use the same command as in the first pass):
If your using the Docker image, run the following:
% docker run -it -v /PATH_TO_YOUR_RAW_DICOM_FOLDER/:/mnt rnair07/bidskit --indir=/mnt/dicom --outdir=/mnt/source
If you're running the script locally, run something similar to the following depending on the command that was run for Phase 1:
% dcm2bids.py -i mydicom -o mysource
This will populate the BIDS source directory from the working conversion directory:
source
├── dataset_description.json
├── participants.tsv
└── sub-Ra0950
└── ses-first
├── anat
│ ├── sub-Ra0950_run-01_T1w.json
│ ├── sub-Ra0950_run-01_T1w.nii.gz
│ ├── sub-Ra0950_run-02_T1w.json
│ └── sub-Ra0950_run-02_T1w.nii.gz
├── fmap
│ ├── sub-Ra0950_acq-fmap_magnitude1.nii.gz
│ ├── sub-Ra0950_acq-fmap_phasediff.json
│ └── sub-Ra0950_acq-fmap_phasediff.nii.gz
└── func
├── sub-Ra0950_task-rest_acq-MB_run-01_bold.json
├── sub-Ra0950_task-rest_acq-MB_run-01_bold.nii.gz
├── sub-Ra0950_task-rest_acq-MB_run-01_events.tsv
├── sub-Ra0950_task-rest_acq-MB_run-02_bold.json
├── sub-Ra0950_task-rest_acq-MB_run-02_bold.nii.gz
└── sub-Ra0950_task-rest_acq-MB_run-02_events.tsv
bidskit attempts to sort the fieldmap data appropriately into magnitude and phase images (for multi-echo GRE fieldmaps), or phase-encoding reversed pairs (for SE-EPI fieldmapping). The resulting dataset_description.json and functional event timing files (func/*_events.tsv) will need to be edited by the user, since the DICOM data contains no information about the design or purpose of the experiment.
Please use the GitHub Issues feature to raise issues with the bidskit repository (https://github.com/jmtyszka/bidskit/issues)