Added the feature to read DICOM images into a 4D Nifti image #99

jph6366 · 2025-04-20T15:30:00Z

Allow reading of DICOM images

Spent a few days getting acclimated to DICOM concepts (shoutout DicomIsEasy) and investigated the best options for this feature request.

I initially wanted to use ITK like it is here, but it seems perhaps a configure-first philosophy might suit this codebase by introducing a new feature with less complexity added to future development. I opted to use dcm2niix, which converts diffusion vectors andit is written in C, so it is very portable, and It attempts to preserve diffusion gradient information. dcm2nii is the predecessor to dcm2niix. It is deprecated for modern images, but does handle image formats that predate DICOM.

There are many good NeuroImaging Tools & Resource Collaboratory to choose from but dicom2niix is a reliable option, following a configure-first policy, we utilize the highly portable C code and simply expose the configurations relevant to the research and development of perfusion imaging. dicom2niix documentation is good and the repository is open source, also they list many alternatives and other tools that exploit dcm2niix.

EDIT:

Refactored code to access both dcm2niix ~~and dcm2niibatch~~ commands with access to some configurations without any need for local code modifications. Handles multiple input directories and will run nifti_wrapper cli if desired ~~(only if user has selected to convert using dcm2niix [I did not add the functionality to call nifti_wrapper on every output from batch call] ).~~

There is a now a separate dockerfile because if a user has no DICOM data then we do not want to install the dicom2niix packages. The README has been updated to instruct users on how to use the different builds.

Link this PR to an issue [optional]

Fixes #68

Checklist

Test dicom2niix_wrapper.py CLI
Update Dockerfile to install dicom2niix latest release
Update README

Oversights

Grassroots DICOM, written in C++, has higher performance executables
- 3D Slicer uses a GDCM port, presumably so does ITK and SimpleITK, I attempted to utilize python-gdcm, an unofficial GDCM package for Python owned by InVesalius developer.
Reading in DICOMDIR records
Maintaining latest updates in future developments
Running via SSH

Happy Testing

UnHappy Testing

vigorously tested using flags and seeing what fails based on missing images or mismatched images
providing invalid inputs for CLI runs
building dockerfiles

Simulating IVIM signal as DICOM data

ITK bridges with numpy and ITK::image is robust, see documentation here and actively maintained.
We crudely configure the header to read for valid conversion
Adding the diffusion tags are different for each Vendor
- we simply just configure the basic tags for encoding gradient directionality and diffusion b-values; although this should be refactored with more direction of what will provide value in the future to the maintainers/end-users/researchers.

For further improvments:

We should utilize nibabel for its dicom reader, nicom, and its Wrapper classes for different vendor-specific format data.
We can also compile for the Web using JS+WebAssembly library @niivue/dcm2niix

…g dicom2niix updated Dockerfile, README, and nifti_wrapper.py

etpeterson

Good start! I have a few comments in specific places.

WrapImage/nifti_wrapper.py

etpeterson · 2025-04-20T20:20:31Z

WrapImage/nifti_wrapper.py

+        bval_files = list(vol_dir.glob("*.bval"))
+        bvec_files = list(vol_dir.glob("*.bvec"))
+        bval_path = str(bval_files[0]) if bval_files else None
+        bvec_path = str(bvec_files[0]) if bvec_files else None


This is all making the assumption that we only want one. What if they're different, which do we choose then? I'm not exactly sure what dcm2niix outputs, but probably best is to require just one output and error if there are more.

For handling batch conversion for many files we could leverage the yaml config for batch processing, as specfied in the dicom2niix repo, so we can provide that option and updated instructions in the README.

I'm not exactly sure what dcm2niix outputs, but probably best is to require just one output and error if there are more.

dicom2niix handles it nicely really by just adding characters in an alphabetical order

e.g.

len(nifti_files) = 0 *.bval, *.bvec *.nii.gz

len(nifti_files) = 1 *a.bval, *a.bvec *a.nii.gz

len(nifti_files) = 2 *b.bval, *b.bvec *b.nii.gz

len(nifti_files) = 3 *c.bval, *c.bvec *c.nii.gz

etpeterson · 2025-04-20T20:27:57Z

Docker/Dockerfile

@@ -5,6 +5,13 @@ WORKDIR /usr/src/app
 RUN apt-get update && apt-get install -y --no-install-recommends \
    build-essential \
    libssl-dev \
+    wget \
+    pigz \
+    && wget https://github.com/rordenlab/dcm2niix/archive/refs/tags/v1.0.20241211.tar.gz \


This could be a separate build stage in the dockerfile. That simplifies the cleanup.

I'm also wondering about the version. Is there a way we can select the latest stable?

it seems the easiest way to do it is just pin the latest version.

We could query the GitHub API but then our docker environment will get rate limited without a personal access token for authorization. Ideally, we dont want to increase our dependency and just lean towards manual updates based on alerting users to make an issue to update the dockerfile.

~~there is an update check option to add to our dicom2niix subprocess call, this command will cause the software report if there are any newer versions of dcm2niix available online. (UNIX ONLY)~~

Refactored previous commits to decouple dicom2niix build from the base Dockerfile for the testing the main code collection. - we now have a separate dockerfile, one for nifti files and one for converting dicom to nifti - we now also handle multiple inputs seamless and the researcher/end-user has access more configuration - we made a separate file to inquire user input and then just inject the nifti wrapper through a subprocess call like we do with our dcm2niix and dcm2niibatch

…I_CodeCollection into dicom2niix

missed this unused import

etpeterson

Thanks for keeping at this, it's converging. Several more comments though.

WrapImage/inquire_dicom2niix.py

Docker/dicom2nifti/Dockerfile

Updating Dockerfile and refactored inquire to function more like a wrapper to the cli and writing a one-call script for easy streamlined use, but also preserving the interactives for debugging Removed dcm2niibatch as its easier to focus on one command that is better maintained. Also updated the readmee

updates to the CLI and conversion error handling

…I_CodeCollection into dicom2niix

Fixed up readme and enabled the output to buffered line by line for our subprocess output

Fixed the --series_number validation

Squashed some bugs and tried some more use case testing. Wrote up a function to save simulated DWI as DICOM images with headers configured for all in x-direction. provided pyproject.toml for separation of dependenccies and setuptools configured for the different root packages.

…I_CodeCollection into dicom2niix

pyproject.toml

Docker/dicom2nifti/DICOM2NiFTI.png

Docker/dicom2nifti/Dockerfile

Rewrote the save dicom function still using ITK, but instead of reading the from the Array, we just convert the nifti file generated into dicom slices and reuse the bval and bvec files. add github workflow test for dicom2niix

…I_CodeCollection into dicom2niix

etpeterson · 2025-04-30T04:58:31Z

The new test fails: https://github.com/OSIPI/TF2.4_IVIM-MRI_CodeCollection/actions/runs/14710507373/job/41395983046?pr=99

.github/workflows/docker-build-and-run.yml

etpeterson · 2025-04-30T04:59:23Z

Getting close. Just smaller comments now.

Hopefully fixed the github action job And updated pyproject.toml comments

…I_CodeCollection into dicom2niix

etpeterson · 2025-05-09T04:20:35Z

Sorry, looks like the test still fails.

etpeterson

Just because the test fails

…I_CodeCollection into dicom2niix

etpeterson

Sorry for the long delay. I was at a conference. Still just that test failure.

etpeterson · 2025-05-23T04:41:20Z

.github/workflows/docker-build-and-run.yml

@@ -98,9 +98,10 @@ jobs:

      - name: Verify input files
        run: |
+          IVIM_SIM_DIR = "${{ github.workspace }}/ivim_simulation"


This test fails. I think because of the whitespace here.

removed whitespace so expression is read instead of command.

.github/workflows/docker-build-and-run.yml

etpeterson

Tests are passing! Just took another look across the code and came up with one final question.

etpeterson · 2025-05-28T05:17:01Z

pyproject.toml

+description = "IVIM MRI signal processing and NIfTI/DICOM processing tools"
+requires-python = ">=3.9"
+readme = "README.md"
+dependencies = [


I'm noticing these dependencies don't completely match with those in requirements.txt. Is that intentional?

Added the feature to read DICOM images into a 4D Nifti image utilizin…

4bc40f2

…g dicom2niix updated Dockerfile, README, and nifti_wrapper.py

etpeterson requested changes Apr 20, 2025

View reviewed changes

jph6366 and others added 5 commits April 21, 2025 10:39

Update README.md

1e060cd

Merge branch 'dicom2niix' of https://github.com/jph6366/TF2.4_IVIM-MR…

e331dfc

…I_CodeCollection into dicom2niix

Update README.md

1247c2c

[Feature] <Allow reading of DICOM images> OSIPI#68

a5f63e6

missed this unused import

jph6366 requested a review from etpeterson April 21, 2025 17:17

etpeterson requested changes Apr 22, 2025

View reviewed changes

jph6366 and others added 9 commits April 22, 2025 09:18

Update README.md

ae56831

[Feature] <Allow reading of DICOM images> OSIPI#68

317d72d

updates to the CLI and conversion error handling

Merge branch 'dicom2niix' of https://github.com/jph6366/TF2.4_IVIM-MR…

783e2f3

…I_CodeCollection into dicom2niix

[Feature] <Allow reading of DICOM images> OSIPI#68

5347a31

Fixed up readme and enabled the output to buffered line by line for our subprocess output

[Feature] <Allow reading of DICOM images> OSIPI#68

22df73a

Fixed the --series_number validation

Update README.md

d18c2e6

Merge branch 'dicom2niix' of https://github.com/jph6366/TF2.4_IVIM-MR…

606b667

…I_CodeCollection into dicom2niix

jph6366 requested a review from etpeterson April 25, 2025 17:00

Update README.md

9a17fb1

etpeterson requested changes Apr 27, 2025

View reviewed changes

jph6366 added 2 commits April 28, 2025 08:31

[Feature] <Allow reading of DICOM images> OSIPI#68

614c8f8

Rewrote the save dicom function still using ITK, but instead of reading the from the Array, we just convert the nifti file generated into dicom slices and reuse the bval and bvec files. add github workflow test for dicom2niix

Merge branch 'dicom2niix' of https://github.com/jph6366/TF2.4_IVIM-MR…

dda0e88

…I_CodeCollection into dicom2niix

jph6366 requested a review from etpeterson April 28, 2025 12:34

jph6366 added 2 commits April 28, 2025 10:32

Update pyproject.toml

ac752cb

Update generate_signal_docker_test.py

92abf87

etpeterson requested changes Apr 30, 2025

View reviewed changes

.github/workflows/docker-build-and-run.yml Show resolved Hide resolved

[Feature] <Allow reading of DICOM images> OSIPI#68

201f495

Hopefully fixed the github action job And updated pyproject.toml comments

Merge branch 'dicom2niix' of https://github.com/jph6366/TF2.4_IVIM-MR…

d32460b

…I_CodeCollection into dicom2niix

jph6366 requested a review from etpeterson April 30, 2025 12:07

Update pyproject.toml

9cd090f

etpeterson requested changes May 9, 2025

View reviewed changes

jph6366 and others added 3 commits May 9, 2025 10:07

Update README.md

30a0603

updated job to be fixed or otherwise log error to help fix

64dc1bc

Merge branch 'dicom2niix' of https://github.com/jph6366/TF2.4_IVIM-MR…

682d655

…I_CodeCollection into dicom2niix

jph6366 requested a review from etpeterson May 9, 2025 16:02

etpeterson requested changes May 23, 2025

View reviewed changes

OSIPI#68

814cdc9

removed whitespace so expression is read instead of command.

etpeterson requested changes May 27, 2025

View reviewed changes

.github/workflows/docker-build-and-run.yml Show resolved Hide resolved

tested locally successfully

015f7dc

etpeterson requested changes May 28, 2025

View reviewed changes

Added the feature to read DICOM images into a 4D Nifti image #99

Are you sure you want to change the base?

Added the feature to read DICOM images into a 4D Nifti image #99

Uh oh!

Conversation

jph6366 commented Apr 20, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Allow reading of DICOM images

Link this PR to an issue [optional]

Checklist

Oversights

Happy Testing

UnHappy Testing

Simulating IVIM signal as DICOM data

Uh oh!

etpeterson left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

etpeterson Apr 20, 2025

Choose a reason for hiding this comment

Uh oh!

jph6366 Apr 20, 2025

Choose a reason for hiding this comment

Uh oh!

jph6366 Apr 21, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

etpeterson Apr 20, 2025

Choose a reason for hiding this comment

Uh oh!

jph6366 Apr 20, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

etpeterson left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

etpeterson commented Apr 30, 2025

Uh oh!

Uh oh!

etpeterson commented Apr 30, 2025

Uh oh!

etpeterson commented May 9, 2025

Uh oh!

etpeterson left a comment

Choose a reason for hiding this comment

Uh oh!

etpeterson left a comment

Choose a reason for hiding this comment

Uh oh!

etpeterson May 23, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

etpeterson left a comment

Choose a reason for hiding this comment

Uh oh!

etpeterson May 28, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

jph6366 commented Apr 20, 2025 •

edited

Loading

jph6366 Apr 21, 2025 •

edited

Loading

jph6366 Apr 20, 2025 •

edited

Loading