Purpose:
NIfTI is a commonly used neuroimaging data format. Many of the analysis software packages require input in NIfTI format. The raw data from most scanners, however, are in the industry standard DICOM format. Therefore, they should first be converted to NIfTI format. This document explains how to do this conversion. The procedure explained here produces a BIDS-compliant folder structure using sample data from our group. The filenames are kept simple for this specific example for compactness. For more proper naming see Storing & Sharing MR Studies, part of our group’s MR analysis pipeline. (Also see https://bids.neuroimaging.io/)
Description:
On your computer, create a folder ‘fMRI-HandsOn’. Under this folder, create the following three folders: ‘DICOM’, ‘NIFTI’, ‘results’
Download this sample dataset: https://drive.google.com/file/d/19TtgIT9OmeIThzhVFwpvbQYF7RnnmA4h/view?usp=sharing
Extract its content under the DICOM folder. (You may have a hierarchy with several subfolders after extracting the contents. In the end make sure that your folder structure looks like the one below)
Under the NIFTI folder, create a new folder ‘sub-01’. Under sub-01, create three new folders: ‘func’, ‘anat’, ‘fmap’
Download and open MRICroGL (accept the License agreement and download the zip file; then extract its content under a folder. Double click on the MRICroGL program inside that folder)
On MRICroGL, click on import and convert DICOM to NIfTI
On the new “dcm2nixx DICOM to NIfTI conversion” window, edit the output file name as sub-01_T1w
For Output Directory, choose the NIFTI/sub-01/anat folder
Next, select a folder to convert. This button is at the bottom of the left column
On the browser window, choose the T1 source folder under DICOM
Click Open. This should convert all the DICOM files under the T1 folder to NIfTI format, and place the NIFTI file under the anat folder. Go back to the MRICroGL main window. Go to File – Open, and open the newly created nifti file. Inspect the images.
Note that the conversion creates a JSON file, too. Open and inspect that file with a text editor.
The NIFTI file created this way will have the facial features visible. You MUST de-face this image. Using software (e.g. pydeface) de-face the image. Replace the original (with face) image with the de-faced one. The original (with face) image MUST be deleted, and never shared with the world.
For the functional data, follow the same procedure. Open MRIcroGL, click on import and convert DICOM to NIFTI
For output filename enter: sub-01_task-motor_bold
For Output Directory choose NIFTI/sub-01/func
Select the motor scan folder to convert
Click Open. This should convert all the DICOM files under the MOCOSERIES_0004 folder to NIfTI format, and place the NIfTI file under the func folder. Open and inspect the NIFTI file with MRICroGL. Click on various voxels and inspect the time course of the BOLD response shown at the bottom panel.
Note that the conversion creates a JSON file, too
Open and edit the JSON file to include the following line
"TaskName": "Motor Task",
Apply the same procedures to convert the vision run DICOM files into NIFTI:
Output file name: sub-01_task-vision_bold
Output Directory: NIFTI/sub-01/func
folder to convert: DICOM/MOCOSERIES_0006
After the conversion is complete edit the sub-01_task-vision_bold.json file to include
"TaskName": "Vision Task",
Finally, the fieldmaps: Open MRIcroGL, click on import and convert DICOM to NIFTI
Output filename: sub-01_dir-AP_epi
Output Directory: NIFTI/sub-01/fmap
Select folder to convert: DICOM/CMRR_MBEP2D_SE_TOPUP_A2P_MB4_0007
Repeat the same procedure for the other direction
Output filename: sub-01_dir-PA_epi
Output Directory: NIFTI/sub-01/fmap
Select folder to convert: DICOM/CMRR_MBEP2D_SE_TOPUP_P2A_MB4_0008
This completes the DICOM to NIfTI conversion. To make the folder BIDS-compliant, and make the files ready for further analyses a few more steps are needed.
First, read and get familiar with the information here: https://github.com/bids-standard/bids-starter-kit. Especially this one: https://bids-specification.readthedocs.io/en/stable/01-introduction.html
Further, read our lab’s procedures for Storing & Sharing MR Studies.
BIDS will require a dataset_description.json file in the NIfTI folder. A minimal file can be something like
{
"Name":"fMRI-Demo",
"BIDSVersion":"1.0.2",
"Authors":["Beyza Akkoyunlu","Huseyin Boyaci"]
}
“Name” and “BIDSVersion” are required fields by BIDS. A more detailed description file would be better (see this example). For example, how is your dataset licensed? Can everyone use it? Under what conditions? If there is one, what is the DOI?
It is optional, but you can have plain text, human-readable README file describing the data under the NIfTI folder. For example
NSC672, Fall 2021 NIFTI files for the vision and motor experimental scans under the Block design, GLM study for the NSC672 class (project). These data are collected over two weeks for the Module 4 component of the course. With the intention of using them for level-1, level-2 and level-3 analysis. Two participants' data were collected.
BIDS does not require but recommends events files for corresponding task fMRI scans. There are two such files for each task scan. One is a tsv file showing the event onsets and duration, second one is json file describing the columns in the tsv file. Here is the events file for the motor task, sub-01_task-motor_events.tsv
onset duration task
0 14 rest
14 12 right
26 12 left
38 12 rest
50 12 both
62 12 left
74 12 right
98 12 both
110 12 right
122 12 left
134 12 rest
146 12 both
158 12 left
170 12 right
182 12 rest
194 12 both
206 12 right
218 12 left
230 12 rest
242 12 both
254 12 left
266 12 right
278 12 rest
290 12 both
302 12 right
314 12 left
326 12 rest
338 12 both
350 12 left
362 12 right
378 12 rest
386 12 both
398 12 right
410 12 left
422 12 rest
434 12 both
446 12 left
458 12 right
470 12 rest
482 12 both
494 12 right
506 12 left
518 12 rest
530 12 both
542 12 left
554 12 right
566 12 rest
578 12 both
and here is the corresponding json file, sub-01_task-motor_events.json
{
"task": {
"LongName": "Task side",
"Description": "Which hand moved",
"Levels": {
"left": "Left hand squeeze",
"right": "Right hand squeeze",
"both": "Both hand squeeze",
"rest": "No hand squeeze"
}
}
}
Note that the onset and duration columns are REQUIRED. Any number of additional columns MAY be added.
For later analyses, you need to tell fsl what the fieldmap files will be used for. This is done by adding the following at the beginning of sub-01_dir-AP_epi.json and sub-01_dir-PA_epi.json files
"IntendedFor":["func/sub-01_task-motor_bold.nii.gz","func/sub-01_task-vision_bold.nii.gz"]
Finally, check whether your NIfTI folder is BIDS-compliant using the online BIDS validator: https://bids-standard.github.io/bids-validator/
Note:
I have outlined the procedure using MRIcroGL above. You SHOULD streamline your conversion using scripts. For basics of scripting see http://reproducibility.stanford.edu/bids-tutorial-series-part-1b. When I have time, I will prepare and upload a sample script for the pipeline we use in CBVG.