No technical changes to the pipeline here, but citations and methods boilerplate have been updated to reflect the changes in 0.18.0alpha0.
First release moving towards 1.0! Please open bug reports if anything suspicious comes up. This release changes the anatomical workflow significantly, synthstrip and synthseg are used. The recon workflow "dsi_studio_autotrack" has also been added.
## What's Changed * Bump sentry-sdk from 0.13.1 to 1.14.0 by @dependabot in PennLINC#539 * [ENH] Update FreeSurfer to 7.3.1, dmri-amico to 1.5.4 by @mattcieslak in PennLINC#537 * WIP: ENH: Make pyAFQ tests faster, add export all by @36000 in PennLINC#534 * [ENH] move biascorrect so it runs on resampled data by default by @mattcieslak in PennLINC#527 * [Fix] Fix threading on DRBUDDI interface by @mattcieslak in PennLINC#540 * [ENH] add CNR to the imageqc.csv by @mattcieslak in PennLINC#541 * [FIX] pin pandas version to < 2.0.0 by @mattcieslak in PennLINC#543 * ENH: Replace avscale with non-fsl tools by @jbh1091 in PennLINC#542 * ENH: Replace fsl applymask by @jbh1091 in PennLINC#544 * Replace fsl split by @jbh1091 in PennLINC#548 * [FIX] Update distortion_group_merge.py by @smeisler in PennLINC#555 * [ENH] Redo anatomical workflow by @mattcieslak in PennLINC#553 * [FIX] remove pre bids-filter acq type argument by @octomike in PennLINC#557 * FIX: Replace deprecated np.int instances by @smeisler in PennLINC#558 * [WIP] ENH: 482 remove fsl dependency by @jbh1091 in PennLINC#498 * [ENH] Update TORTOISE for improved T2w registration by @mattcieslak in PennLINC#564 * [FIX] T2w anat-modality issues by @mattcieslak in PennLINC#565 * [FIX] update boost in tortoise by @mattcieslak in PennLINC#569 * [FIX] connections on multi-anat workflow by @mattcieslak in PennLINC#572 * [ENH] Update DSI Studio to the latest commit by @mattcieslak in PennLINC#573 * [ENH] Add DSI Studio AutoTrack recon workflow by @mattcieslak in PennLINC#576
## New Contributors * @dependabot made their first contribution in PennLINC#539 * @jbh1091 made their first contribution in PennLINC#542 * @smeisler made their first contribution in PennLINC#555
Full Changelog: https://github.com/PennLINC/qsiprep/compare/0.17.0...0.18.0alpha0
Adds a critical fix for ABCD-style acquisitions (described in #449). This change forces TOPUP to use the raw, unprocessed b=0 images from the DWI series and the epi fieldmaps to estimate distortion. Previously, the most-denoised version of each image was used in TOPUP. To disable this change and return to the previous behavior, use the --denoised-image-sdc flag.
Note, this is a change in the default behavior of QSIPrep!!
Upgrades
- Update ITK to 5.3, update ANTs #449
- Add --denoised-image-sdc #465
Bug fixes
- Use safe_load instead of load for yaml #443
- Add fugue and prelude back to the qsiprep image #463
- Adds multithreading to connectome2tck #429
Fixes a naming error in the schaefer 400 atlas #428
Major additions to the reconstruction workflows! Most notably PyAFQ is available as a reconstruction workflow. The default atlases included in QSIPrep have been updated to include subcortical regions if they weren't already present in the original atlas.
- Add PyAFQ reconstruction workflows #398 Credit: @36000
- Make sure all recon workflows respect omp_nthreads #368
- Add DKI derivatives #371
- Properly transform 4D CNR images from Eddy #393
- Update amico to version 22.4.1 #394
- Fix concatenation bug #403 credit: @cookpa
- Prevent divide by zero error #405 credit: @cookpa
- Critical Fix, use correct transform to get atlases into T1w space #417
- Add resampled atlases back into derivatives #418
- Add connectome2tck exemplar streamlines for mrtrix connectivity workflows #420
- Update the atlases to include subcortical regions #426 [details here](https://github.com/PennLINC/qsiprep-atlas/blob/main/QSIRecon%20atlases.ipynb)
WARNING There is an bug in the connectome pipelines that makes the connectivity matrices unreliable. Do not use this version for connectome estimation.
Due to persistent difficulties with crashing ODF plots in the reconstruction workflows, there is now a --skip-odf-reports option that will disable the ODF and peak plots in the html reports. This should only be used once you've run some test workflows with the reports still enabled, so you know that your ODFs are correctly oriented.
- Make ODF Plots optional (#364)
- Bugfix: ABCD gradient data for extrapolation (#363)
- Adds dipy_dki reconstruction workflow (#366)
WARNING There is an bug in the connectome pipelines that makes the connectivity matrices unreliable. Do not use this version for connectome estimation.
A lot of changes in QSIPrep. The big-picture changes are
- The build system was redone so a multistage build is used in a different repository (https://github.com/PennLINC/qsiprep_build). The container should be about half as big as the last release.
- The way anatomical masks are handled in reconstruction workflows has been changed so that FreeSurfer data can be incorporated.
- FAST-based anatomically-constrained tractography is now deprecated in QSIPrep. If you're going to use anatomical constraints, they should be very accurate. The hybrid surface-volume segmentation (HSVS) is amazing and should be considered the default way to use the MRtrix3/3Tissue workflows. The [documentation](https://qsiprep.readthedocs.io/en/latest/reconstruction.html) describes the new built-in workflow names.
- The reconstruction workflows have been totally refactored. This won't affect the outputs of the reconstruction workflows, but will affect anyone who is using intermediate files from the working directory. The working directories no longer have those unfortunate ..'s in their names.
- FSL is updated to 6.0.5.1!
Since these are a lot of changes, please be vigilant and check your results! The QSIPrep preprocessing workflows have not changed with this release, but the dependencies have been upgraded for almost everything.
- Update FSL to 6.0.5.1 (#334)
- Move ODF plotting to a cli tool so xvfb is handled more robustly (#357)
- Better FreeSurfer license documentation (#355)
- Edited libQt5Core.so.5 so it's loadable in singularity on CentOS (#336)
- Fixed typo in patch2self (#338)
- Inaccurate bids-validator errors were removed (#340)
- Bug in --recon-input fixed #286
- Correct streamline count is reported in the mrtrix connectivity matrices (#330)
- Add option to ingress freesurfer data (#287)
- Add Nature Methods citation to dataset_description.json
- Refactor build system (#341)
- SHORELine bugfixes (#301)
- Bugfix: handle cases where there is only one b=0 (#279)
Change in behavior in Patch2Self:
- Updates Patch2Self with optimal parameters (use OLS instead of ridge)
Bugfixes and documentation
- Updates documentation for containers (#270)
- Fixes a bug when reading fieldmap metadata from datalad inputs (#271)
- Change incorrect option in the documentation (#272)
Adds a new reconstruction workflow for the NODDI model.
- Adds NODDI reconstruction workflow (#257). Thanks @cookpa!
- Fixes issue with unequal aspect ratios in q-space plots (#266)
- Adds a flag for a BIDS filter file #256
- Fixes a bug where --dwi-only is selected along with --intramodal-template
Many bugfixes
- Fix bug that produced flipped scalar images (#251)
- Added a default working directory to prevent uninterpretable error message (#250)
- Fix a bug in the dipy_3dshore reconstruction workflow (#249)
- Remove hardlinking from DSI Studio interfaces (#214)
- Add an option to use a BIDS database directory (#247)
- Fix bug in interactive reports for HCP-style acquisitions (#238)
- Update defaults for Patch2Self (#230, #239)
- Remove cmake installer from docker image after compiling ANTS (#229)
This version introduces major changes to the TOPUP/eddy workflow. Feedback would be greatly appreciated!
- Added new algorithm for selecting b=0 images for distortion corretion (#202)
- Added the Patch2Self denoising method (#203, credit to @ShreyasFadnavis)
- Documentation has been expanded significantly (#212)
- Boilerplate for DWI preprocessing is greatly expanded (#200)
Adds options for processing infant dMRI data. Also enables running without a T1w image.
- Adds
--dwi-onlyand--infantoptions to QSIPrep. (#177)
NEW: Workflow defaults have changed. T1w-based spatial normalization is done by
default (disabled by --skip-t1-based-spatial-normalization) and dwi scans
are merged before motion correction by default (disabled by --separate-all-dwis).
- Deprecate some commandline arguments, change defaults (#168)
- Fix typo in workflow names (#162)
- Fix bug from 0.10.0 where ODFs were not appearing in plots (#160)
- Adds support for oblique acquisitions (#156)
- Adds support for HCP lifespan sequences
- Introduces --distortion-group-merge option for combining paired scans
- Removes oblique angles from T1w headers to fix N4 bug (#103)
- Fixed a bug in b=0 masking when images have high signal intensity in ventricles (#99)
- Image QC summary data is produced for each output (#95)
- Update DSI Studio (#88)
- Update ANTs (#80)
- Include workflows for ss3t (#82)
- Add some boilerplate to the FSL workflow (#38)
- Reduce the number of calls to N4 (#74, #89)
- Add CUDA capability in the containers (#75)
- Add mrdegibbs and accompanying reports (#58)
- Fix reports graphics (#64)
- Rework the DWI grouping algorithm (#92)
- This release adds some rather big updates to QSIPrep.
- FSL is updated to version 6.0.3
- CUDA v9.1 support is added to the image (works with GPUS in Docker and Singularity)
- A new robust b=0 masking algorith is introduced.
- Improved handling of Freesurfer path (#50)
- Better logic in commandline argument checking (#50, #62)
- More robust brain masking for b=0 reference images (#73)
- Bugfix for reverse phase encoding directon dwi series (#68)
- Bugfix for warping eddy's CNR output (#72)
- IMPORTANT: commandline call changed to use official BIDS App
- eddy will use multiple cores if available
- Fixed bug in sentry interaction
- Bugfix: masking was not working on eddy.
- Bugfix: static versioning was not workign in the container.
- New graphics in the documentation.
- Use BSpline Interpolation if --output-resolution is higher than the input resolution.
An issue was discovered in how voxel orientation interacts with TOPUP/eddy and outside fieldmaps. Unless everything is in LAS+ prior to going into TOPUP/eddy, the warps are incorrectly applied at the end of eddy. This resulted in fieldmap unwarping reports that looked good but a final output that is bizarrely warped. Additionally, GRE fieldmaps would result in outputs being under-unwarped. To fix all of these, TOPUP (if PEPOLAR fieldmaps are being used) and eddy occur in LAS+, then their outputs are converted to LPS+ for GRE fieldmaps, SyN. The rest of the pipeline happens in LPS+, like the SHORELine version.
- Update installation method to match fMRIPrep
- Add CI tests for reconstruction workflows
- Make the
--sloppyoption affect the reconstruction workflows- Fixes bug in 3dSHORE reconstruction (incorrect scaling)
- CRITICAL bug fix: convert everything to LAS+ if eddy is going to be used
- Added built-in reconstruction workflows
- Added Brainnetome, AICHA and the remaining Schaefer atlases
- Address issues in Nipype causing random crashes
- Use antsMultiVariateTemplateConstruction2.sh to make a b=0 template across scan groups
- Control the number of template iterations and deformation model with
--intramodal_template_itersand--intramodal_template_transform.
- More documentation updates
- MSD calculated for MAPMRI
- Scalar outputs from MAPMRI
- Default eddy configuation changed to not use CUDA by default.
- Valerie added content to documentation
FSL tools are used to match SHORELine motion parameters to those from eddy.
- Fieldcoefs are calculated from PEPOLAR and GRE fieldmaps and sent to TOPUP
- Motion estimates from SHORELine match eddy
Add workflows for eddy and TOPUP.
- Adds eddy tests on CircleCI.