What’s new

1.0.3 (3rd of January 2018)

  • [FIX] Roll back run/task groupings after BIDS query (#918). Groupings for the multi-echo extension will be reenabled soon.

1.0.2 (2nd of January 2018)

  • [FIX] Grouping runs broke FMRIPREP on some datasets (#916) Thanks to @emdupre

1.0.1 (1st of January 2018)

With thanks to @emdupre for contributions.

  • [PIN] Update required niworkflows version to 0.2.3
  • [FIX] Refine antsBrainExtraction if recon-all is run (#912) With thanks to Arno Klein for his helpful comments here (https://github.com/poldracklab/fmriprep/issues/431#issuecomment-299583391)
  • [FIX] Use thinner contours in reportlets (#910)
  • [FIX] Robuster EPI mask (#911)
  • [FIX] Set workflow return value before potential error (#887)
  • [DOC] Documentation about FreeSurfer and --fs-no-reconall (#894)
  • [DOC] Fix example in installation ants-nthreads -> omp-nthreads (#885) With thanks to @mvdoc.
  • [ENH] Allow for multiecho data (#875)

1.0.0 (6th of December 2017)

  • [ENH] Add --resource-monitor flag (#883)
  • [FIX] Collision between Multi-T1w and --no-freesurfer (#880)
  • [FIX] Setting use_compression on resampling workflows (#882)
  • [ENH] Estimate motion parameters before STC (#876)
  • [ENH] Add --stop-on-first-crash option (#865)
  • [FIX] Correctly handling xforms (#874)
  • [FIX] Combined ROI reportlets (#872)
  • [ENH] Strip reportlets out of full report (#867)

1.0.0-rc13 (1st of December 2017)

  • [FIX] Broken --fs-license-file argument (#869)

1.0.0-rc12 (29th of November 2017)

  • [ENH] Use Nipype MultiProc even for sequential execution (#856)
  • [REF] More memory annotations and considerations (#816)
  • [FIX] Controlling memory explosion (#854)
  • [WRAPPER] Mount nipype repositories as niworkflows submodule (#834)
  • [FIX] Reduce image loads in local memory (#839)
  • [ENH] Always sync qforms, refactor error messaging (#851)

1.0.0-rc11 (24th of November 2017)

  • [ENH] Check for invalid qforms in validation (#847)
  • [FIX] Update pybids to include latest bugfixes (#838)
  • [FIX] MultiApplyTransforms failed with nthreads=1 (#835)

1.0.0-rc10 (9th of November 2017)

  • [FIX] Adopt new FreeSurfer (v6.0.1) license mechanism (#787)
  • [ENH] Output affine transforms from original T1w images to preprocessed anatomical (#726)
  • [FIX] Correct headers in AFNI-generated NIfTI files (#818)
  • [FIX] Normalize T1w image qform/sform matrices (#820)

1.0.0-rc9 (2nd of November 2017)

  • [FIX] Fixed #776 (aCompCor - numpy.linalg.linalg.LinAlgError: SVD did not converge) via #807.
  • [ENH] Added CSF column to _confounds.tsv (included in #807)
  • [DOC] Add more details on the outputs of FMRIPREP and minor fixes (#811)
  • [ENH] Processing confounds in BOLD space (#807)
  • [ENH] Updated niworkflows and nipype, including the new feature to close all file descriptors (#810)
  • [REF] Refactored BOLD workflows module (#805)
  • [ENH] Improved memory annotations (#803, #807)

1.0.0-rc8 (27th of October 2017)

  • [FIX] Allow missing magnitude2 in phasediff-type fieldmaps (#802)
  • [FIX] Lower tolerance deciding t1_merge shapes (#798)
  • [FIX] Be robust to 4D T1w images (#797)
  • [ENH] Resource annotations (#746)
  • [ENH] Use indexed_gzip with nibabel (#788)
  • [FIX] Reduce FoV of outputs in T1w space (#785)

1.0.0-rc7 (20th of October 2017)

  • [ENH] Update pinned version of nipype to latest master
  • [ENH] Added rX permissions to make life easier on Singularity users (#757)
  • [DOC] Citation boilerplate (#779)
  • [FIX] Patch to remove long filenames after mri_concatenate_lta (#778)
  • [FIX] Only use unbiased template with --longitudinal (#771)
  • [FIX] Use t1_2_fsnative registration when sampling to surface (#762)
  • [ENH] Remove --skull_strip_ants option (#761)
  • [DOC] Add reference to beginners guide (#763)

1.0.0-rc6 (11th of October 2017)

  • [ENH] Add inverse normalization transform (MNI -> T1w) to derivatives (#754)
  • [ENH] Fall back to initial registration if BBR fails (#694)
  • [FIX] Header and affine transform updates to resolve intermittent misalignments in reports (#743)
  • [FIX] Register FreeSurfer template to FMRIPREP template, handling pre-run FreeSurfer subjects more robustly, saving affine to derivatives (#733)
  • [ENH] Add OpenFMRI participant sampler command-line tool (#704)
  • [ENH] For SyN-SDC, assume phase-encoding direction of A-P unless specified L-R (#740, #744)
  • [ENH] Permit skull-stripping with NKI ANTs template (#729)
  • [ENH] Erode aCompCor masks to target volume proportions, instead of fixed distances (#731, #732)
  • [DOC] Documentation updates (#748)

1.0.0-rc5 (25th of September 2017)

  • [FIX] Skip slice time correction on BOLD series < 5 volumes (#711)
  • [FIX] Skip AFNI check for new versions (#723)
  • [DOC] Documentation clarification and updates (#698, #711)

1.0.0-rc4 (12th of September 2017)

With thanks to Mathias Goncalves for contributions.

  • [ENH] Collapse ITK transforms of head-motion correction in only one file (#695)
  • [FIX] Raise error when run.py is called directly (#692)
  • [FIX] Parse crash files when they are stored as text (#690)
  • [ENH] Replace medial wall values with NaNs (#687)

1.0.0-rc3 (28th of August 2017)

With thanks to Anibal Sólon for contributions.

  • [ENH] Add --low-mem option to reduce memory usage for large BOLD series (#663)
  • [ENH] Parallelize anatomical conformation step (#666)
  • [FIX] Handle missing functional data in SubjectSummary node (#670)
  • [FIX] Disable --no-skull-strip-ants (AFNI skull-stripping) (#674)
  • [FIX] Initialize SyN SDC more robustly (#680)
  • [DOC] Add comprehensive documentation of workflow API (#638)

1.0.0-rc2 (12th of August 2017)

  • [ENH] Increased support for partial field-of-view BOLD datasets (#659)
  • [FIX] Slice time correction is now being applied to output data (not only to intermediate file used for motion estimation - #662)
  • [FIX] Fieldmap unwarping is now being applied to MNI space outputs (not only to T1w space outputs - #662)

1.0.0-rc1 (8th of August 2017)

  • [ENH] Include ICA-AROMA confounds in report (#646)
  • [ENH] Save non-aggressively denoised BOLD series (#648)
  • [ENH] Improved logging messages (#621)
  • [ENH] Improved resource management (#622, #629, #640, #641)
  • [ENH] Improved confound header names (#634)
  • [FIX] Ensure multi-T1w image datasets have RAS-oriented template (#637)
  • [FIX] More informative errors for conflicting options (#632)
  • [DOC] Improved report summaries (#647)

0.6.0 (31st of July 2017)

With thanks to Yaroslav Halchenko and Ilkay Isik for contributions.

  • [ENH] Set threshold on up-sampling ratio in conformation, report results (#601)
  • [ENH] Censor non-steady-state volumes prior to CompCor (#603)
  • [FIX] Conformation failure in thick-slice, oblique T1w datasets (#601)
  • [FIX] Crash/report failure of phase-difference SDC pipeline (#602, #604)
  • [FIX] Prevent AFNI NIfTI extensions from crashing reference EPI estimation (#619)
  • [DOC] Save logs to output directory (#605)
  • [ENH] Upgrade to ICA-AROMA 0.4.1-beta (#611)

0.5.4 (20th of July 2017)

  • [DOC] Improved report summaries describing steps taken (#584)
  • [ENH] Uniformize command-line argument style (#592)

0.5.3 (18th of July 2017)

With thanks to Yaroslav Halchenko for contributions.

  • [ENH] High-pass filter time series prior to CompCor (#577)
  • [ENH] Validate and minimally conform BOLD images (#581)
  • [FIX] Bug that prevented PE direction estimation (#586)
  • [DOC] Log version/time in report (#587)

0.5.2 (30th of June 2017)

With thanks to James Kent for contributions.

  • [ENH] Calculate noise components in functional data with ICA-AROMA (#539)
  • [FIX] Remove unused parameters from function node, resolving crash (#576)

0.5.1 (24th of June 2017)

  • [FIX] Invalid parameter in bbreg_wf (#572)

0.5.0 (21st of June 2017)

With thanks to James Kent for contributions.

  • [ENH] EXPERIMENTAL: Fieldmap-less susceptibility correction with --use-syn-sdc option (#544)
  • [FIX] Reduce interpolation artifacts in ConformSeries (#564)
  • [FIX] Improve consistency of handling of fieldmaps (#565)
  • [FIX] Apply T2w pial surface refinement at correct stage of FreeSurfer pipeline (#568)
  • [ENH] Add --anat-only workflow option (#560)
  • [FIX] Output all tissue class/probability maps (#569)
  • [ENH] Upgrade to ANTs 2.2.0 (#561)

0.4.6 (14th of June 2017)

  • [ENH] Conform and minimally resample multiple T1w images (#545)
  • [FIX] Return non-zero exit code on all errors (#554)
  • [ENH] Improve error reporting for missing subjects (#558)

0.4.5 (12th of June 2017)

With thanks to Marcel Falkiewicz for contributions.

  • [FIX] Correctly display help in fmriprep-docker (#533)
  • [FIX] Avoid invalid symlinks when running FreeSurfer (#536)
  • [ENH] Improve dependency management for users unable to use Docker/Singularity containers (#549)
  • [FIX] Return correct exit code when a Function node fails (#554)

0.4.4 (20th of May 2017)

With thanks to Feilong Ma for contributions.

  • [ENH] Option to provide a custom reference grid image (--output-grid-reference) for determining the field of view and resolution of output images (#480)
  • [ENH] Improved EPI skull stripping and tissue contrast enhancements (#519)
  • [ENH] Improve resource use estimates in FreeSurfer workflow (#506)
  • [ENH] Moved missing values in the DVARS* and FramewiseDisplacement columns of the _confounds.tsv from last row to the first row (#523)
  • [ENH] More robust initialization of the normalization procedure (#529)

0.4.3 (10th of May 2017)

  • [ENH] --output-space template targets template specified by --template flag (MNI152NLin2009cAsym supported) (#498)
  • [FIX] Fix a bug causing small numerical discrepancies in input data voxel size to lead to different FOV of the output files (#513)

0.4.2 (3rd of May 2017)

  • [ENH] Use robust template generation for multiple T1w images (#481)
  • [ENH] Anatomical MNI outputs respect --output-space selection (#490)
  • [ENH] Added support for distortion correction using opposite phase encoding direction EPI images (#493)
  • [ENH] Switched to FSL BET for skullstripping of EPI images (#493)
  • [ENH] --omp-nthreads controls maximum per-process thread count; replaces --ants-nthreads (#500)

0.4.1 (20th of April 2017)

  • Hotfix release (dependencies and deployment system)

0.4.0 (20th of April 2017)

  • [ENH] Added an option to choose the degrees of freedom used when doing BOLD to T1w coregistration (--bold2t1w_dof). Set default to 9 to account for field inhomogeneities and coils heating up (#448)
  • [ENH] Added support for phase difference and GE style fieldmaps (#448)
  • [ENH] Generate GrayWhite, Pial, MidThickness and inflated surfaces (#398)
  • [ENH] Memory and performance improvements for calculating the EPI reference (#436)
  • [ENH] Sample functional series to subject and fsaverage surfaces (#391)
  • [ENH] Output spaces for functional data may be selected with --output-space option (#447)
  • [DEP] --skip-native functionality replaced by --output-space (#447)
  • [ENH] fmriprep-docker wrapper script simplifies running in a Docker environment (#317)

0.3.2 (7th of April 2017)

With thanks to Asier Erramuzpe for contributions.

  • [ENH] Added optional slice time correction (#415)
  • [ENH] Removed redundant motion parameter conversion step using avscale (#415)
  • [ENH] FreeSurfer submillimeter reconstruction may be disabled with --no-submm-recon (#422)
  • [ENH] Switch bbregister init from fsl to coreg (FreeSurfer native #423)
  • [ENH] Motion estimation now uses a smart reference image that takes advantage of T1 saturation (#421)
  • [FIX] Fix report generation with --reports-only (#427)

0.3.1 (24th of March 2017)

  • [ENH] Perform bias field correction of EPI images prior to coregistration (#409)
  • [FIX] Fix an orientation issue affecting some datasets when bbregister was used (#408)
  • [ENH] Minor improvements to the reports aesthetics (#428)

0.3.0 (20th of March 2017)

  • [FIX] Affine and warp MNI transforms are now applied in the correct order
  • [ENH] Added preliminary support for reconstruction of cortical surfaces using FreeSurfer
  • [ENH] Switched to bbregister for BOLD to T1 coregistration
  • [ENH] Switched to sinc interpolation of preprocessed BOLD and T1w outputs
  • [ENH] Preprocessed BOLD volumes are now saved in the T1w space instead of mean BOLD
  • [FIX] Fixed a bug with MCFLIRT interpolation inducing slow drift
  • [ENH] All files are now saved in Float32 instead of Float64 to save space

0.2.0 (13th of January 2017)

  • Initial public release

0.1.2 (3rd of October 2016)

  • [FIX] Downloads from OSF, remove data downloader (now in niworkflows)
  • [FIX] pybids was missing in the install_requires
  • [DEP] Deprecated -S/--subject-id tag
  • [ENH] Accept subjects with several T1w images (#114)
  • [ENH] Documentation updates (#130, #131)
  • [TST] Re-enabled CircleCI tests on one subject from ds054 of OpenfMRI
  • [ENH] Add C3D to docker image, updated poldracklab hub (#128, #119)
  • [ENH] CLI is now BIDS-Apps compliant (#123)

0.1.1 (30th of July 2016)

  • [ENH] Grabbit integration (#113)
  • [ENH] More outputs in MNI space (#99)
  • [ENH] Implementation of phase-difference fieldmap estimation (#91)
  • [ENH] Fixed bug using non-RAS EPI
  • [ENH] Works on ds005 (datasets without fieldmap nor sbref)
  • [ENH] Outputs start to follow BIDS-derivatives (WIP)


  • [ENH] Added Docker images
  • [DOC] Added base code for automatic publication to RTD.
  • Set up CircleCI with a first smoke test on one subject.
  • BIDS tree scrubbing and subject-session-run selection.
  • Refactored big workflow into consistent pieces.
  • Migrated Craig’s original code