-
Notifications
You must be signed in to change notification settings - Fork 21
Niak Releases
Pierre Bellec edited this page Jun 19, 2015
·
3 revisions
This release corresponds to NIAK version 1887 and PSOM version 760 (release 1.0.2), as numbered by the subversion repository. This is an update on the 0.7.0 release where the version of PSOM (1.0.1) had a bug in the update feature. The default of FLAG_NU_CORRECT in the slice timing brick was also set to off, because for a reason that is unknown at this time this operation behaves differently on different computational platforms. With default parameters, the three pipelines released as part of NIAK 0.7.1 have been successfully tested for reproducibility on two different computational platforms (Ubuntu 12.04 vs CentOS 5.5, both with Octave 3.6.2 and minc-toolkit 0.3.18). The tests were performed both within and between platforms.
This release corresponds to NIAK version 1874 and PSOM version 755 (release 1.0.1), as numbered by the subversion repositories. New features are:
- New pipeline to generate graph properties on connectomes and functional connectivity maps (niak_pipeline_connectome). The graph measures are based on the brain connectivity toolbox https://sites.google.com/site/bctnet/
- New pipeline for region growing
- a command for region growing in 3D (niak_region_growing)
- a new brick for region growing on a single subject's fMRI (niak_brick_region_growing)
- a new pipeline for region growing with multiple fMRI datasets (niak_pipeline_region_growing)
- New tools for linear model analysis:
- read a model in CSV format (niak_read_model)
- least-square estimation (niak_lse)
- general linear model analysis (niak_glm)
- "normalization" of a model, i.e. orthogonalize, correct for mean/variance, create interaction terms, select subsects of the data, etc. (niak_normalize_model).
- White's test for heteroskedasticity (niak_white_test_hetero)
- New statistical tools
- bootstrap of time series (niak_bootstrap_tseries)
- false-discovery rate in a family of p values (niak_fdr)
- New tools for clustering:
- hierarchical clustering (niak_hierarchical_clustering)
- threshold a hierarchy (niak_threshold_hierarchy)
- re-order points based on a hierarchy (niak_hier2order)
- re-order points based on a similarity matrix and a partition (niak_part2order)
- visualize a hierarchy (niak_dendrogram)
- visualize a partition (niak_visu_part).
- An improved version of k-means clustering (with bisecting approach, k-means++ initialization, optimized code, and others)
- Surface analysis:
- A new reader for surface data (niak_read_surf)
- A new tool to visualize surface data (niak_visu_surf)
- New surface template
- A new brick to interpolate volumetric data on the surface (niak_brick_vol2surf)
- fMRI preprocessing:
- Added the scrubbing method of Power 2012
- Preprocessed file now have a xxxx_extra.mat companion, including all confounds that have been regressed out as well as a list of scrubbed volumes.
- Added some regression of confounds in the pipeline (niak_brick_regress_confounds)
- Added a correction of the global average "à la Carbonell" in niak_brick_regress_confounds
- Added the COMPCOR method to correct structured noise in fMRI
- the CORSICA method is now off by default
- The default resolution for the normalized fMRI datasets resampled in stereotaxic space is now 3 mm isotropic.
- It is now possible to choose between the symmetric (default) and asymmetric MNI brain templates for coregistration of T1 scans in stereotaxic space.
- lots of new measures for quality control
- A new tool to estimate partial volume effects on a T1 volume (thanks to Jussi Tohka, niak_brick_pve)
- A new version of the AAL parcellation, tweaked to fit the non-linear symmetric MNI template (40 iterations).
- A new "grabber", i.e. a function to parse the outputs of the fMRI preprocessing pipeline, select a subset of QCed subjects, and generate a structure that can be fed in other pipelines such as the region growing pipeline (niak_grab_fmri_preprocess).
- Many bug fixes and minor improvements.
This release corresponds to NIAK version 1119 and PSOM version 491 (release 0.8.8), as numbered by the subversion repositories. New features are :
- Some progress made on the fMRIstat pipeline. Not functional yet.
- A new brick to convert 4D volumes into an AVI "montage-style" movie.
- Bug fix in the fMRI preprocessing pipeline : the option to get rid of volumes in the motion correction corrupted the data (opt.motion_correction.suppress_vol).
- Bug fix in the fMRI preprocessing pipeline : with multi-session data, the average fMRI volumes are off. As a result, the fMRI to T1 coregistration was likely to fail.
- The pipeline manager (PSOM) is now much faster to initialize and submit jobs.
- Minor bug fixes and improvements.
This release corresponds to NIAK version 1087 and PSOM version 477, as numbered by the subversion repositories. New features are :
- NIAK is now fully compatible with Octave.
- There is a new user guide in pdf - http://www.nitrc.org/frs/downloadlink.php/2726 - and a new user site with tutorials - http://www.nitrc.org/plugins/mwiki/index.php/niak:MainPage -
- A completely revised fMRI preprocessing pipeline. Main changes :
- There are no "styles" anymore. Most steps of the pipeline can instead be skipped using dedicated flags.
- The pipeline depends only on the minc tools.
- There are lots of new steps for quality control purposes.
- the motion correction operates at the run level if possible, and is much faster. The strategy is also now fully hierarchical : within-run, between-run within-session, between-sessions.
- The CORSICA pipeline has been largely revised. It is now possible to specify the masks that will be used to identify components. There is now a separate brick NIAK_BRICK_MASK_CORSICA to generate the masks in individual space based on the canonical masks bundled in NIAK. Those masks are now combined with individual tissue classification masks.
- The T1-fMRI coregistration brick now follows a progressive iterative fit approach, with a mutual information cost function at every step. The volumes are prepared in a different way than before. See the manual for details.
- A new brick to perform quality control of the coregistration between brain volumes (NIAK_BRICK_QC_COREGISTER).
- NIAK_BRICK_SLICE_TIMING has now simplified options available. It is still possible to specify the slice times, but the recommended method is now to indicate the slice timing scheme (sequential, interleaved, ascending, descending).
- support of zipped NIFTI files in NIAK_BRICK_NII2MNC.
- New commands to convert coordinates from voxel to world space and from world space to voxel space (NIAK_COORD_XXX).
- A couple of new functions to estimate connectivity measures (correlation, partial correlation, integration, etc).
- A bunch of small functions to convert symmetric matrices into vectorial form, and back to square form.
- A new command to read/write text files containing a table coded by comma-separated values.
- New bricks for quality control of motion correction of fMRI data at the individual levels (NIAK_BRICK_QC_MOTION_CORRECTION_IND, NIAK_BRICK_QC_MOTION_CORRECTION_GROUP).
- New option in the SMOOTH_VOL brick : correction of edges effect in a binary mask
- Many bug fixes and memory optimization. See the [http://code.google.com/p/niak/source/list logs] for the full story !
This release corresponds to NIAK version 949 and PSOM version 434, as numbered by the subversion repositories. New features are :
- The way temporary file names/folders are generated have been completely revised. They are generated randomly and a test is done to check that the name is not already used (that was the case before) but this is not sufficient for two jobs that would start exactly at the same time, which happened in practice. Now, temporary file names systematically use the name of the job when executed through PSOM, which offers an almost ultimate protection against conflicts of temporary file names.
- Deleting the "PIPE.lock" file in the logs folder will now force the pipeline manager to stop, even if it is running in the background. If a bunch of jobs are being submitted, the manager will quit upon completion of submissions.
- A new option in the fMRI preprocessing template : it is now possible to specify a volume that defines the bounding box and resolution for resampling into stereotaxic space.
- A new function for writing tables in "comma separated values" (CSV) format. This will gradually replace the convention-less text tables generated by NIAK.
- Memory optimization for the resampling brick.
- A new brick for T1 preprocessing (includes non-uniformity correction, brain extraction, tissue segmentation, linear and non-linear coregistration in stereotaxic space). At this stage it has not (yet) been integrated to the fMRI preprocessing pipeline.
- A new brick for T1-EPI coregistration. It is essentially an improved version of the old one, which does not require to specifiy partial-volume effects estimates. At this stage it has not (yet) been integrated to the fMRI preprocessing pipeline.
- New bricks for linear and non-linear multi-step coregistration of a T1 brain volume in stereotaxic space. At this stage it has not (yet) been integrated to the fMRI preprocessing pipeline. This brick is based on PERL scripts from the CIVET pipeline and some brain templates from the MNI that are part of the release (these are distributed under a modified BSD license).
- A bunch of new bricks that are just overlays of MINC tools : NU_CORRECT, MINCMORPH, SEGMENT.
- a new brick for brain extraction in T1 volume. It has a function counterpart, and is based on a new function for spatial clustering using spatial density.
- a new function for multidimensional scaling representation of time series, based on the "high-throughput MDS" code by Marc Strickert, from the Leibniz-Institute of Crop Plant Research, IPK-Gatersleben.
- The port of fMRIstat functions to NIAK is now complete. The fMRIstat pipeline is currently being worked on.
- a bunch of minor bug fix and improvements. See the [http://code.google.com/p/niak/source/list logs] for the full story !
This release corresponds to NIAK version 847 and PSOM version 402, as numbered by the subversion repositories. Main changes with 0.6.1 are :
- The slice timing is now done before the motion correction in the fMRI preprocessing pipeline. The rationale is the following. The best order between st and mc is unclear for single-session data, but for multi-session ones the slice timing has to be done before the motion correction, because the slice timing information is meaningless after large between-sessions motion correction has been applied.
- Bug fix in the minc reader : warning issued by the shell will not crash the reader anymore.
- There is no dependence on the sica toolbox anymore.
- MINC2 is now supported as an output format (before all minc outputs were in MINC1).
- The default for initial transformation in the T1-T2 coregistration brick is now 'identity'. It used to be 'center' which is more robust for large transformation but can sometimes totally fail.
- Added cubic spline scheme in the slice timing correction. It is now the default.
- bricks for level 1 & 2 GLM analysis
- many minor bug fixes
- The latest version of PSOM was used, which includes many bug fix, a faster initialization stage and support of octave (which is not the case yet for NIAK).
A slight change in the roadmap of the project explains the absence of version 0.5.
- the pipeline system has been redesigned and turned into an independent project called [http://code.google.com/p/psom/ psom]. Running pipelines in NIAK is now much more straightforward than it was in version 0.4. PSOM is distributed inside the NIAK package and the user's guide has been updated.
- The main new feature is the [http://www.sciencedirect.com/science?_ob=ArticleURL&_udi=B6T9D-4MG1NXP-5&_user=458507&_coverDate=01%2F31%2F2007&_alid=833955894&_rdoc=2&_fmt=high&_orig=search&_cdi=5112&_sort=d&_docanchor=&view=c&_ct=2&_acct=C000022002&_version=1&_urlVersion=0&_userid=458507&md5=8b982d8caa343b8c98d664f9ab36929c CORSICA] method to correct for physiological noise that have been added to the pre-processing of fMRI data.
- New bricks include individual spatial ICA (and PCA), selection of ICA components based on spatial priors, and suppression of ICA components
- A specific template for physiological noise correction.
- The main new feature is a versatile pipeline for fMRI pre-processing
- Multiple pipeline styles available, optimized for linear modeling or connectivity analysis.
- Available bricks include : slice-timing and motion correction, spatial and temporal filtering, coregistration between T1 and T2 images, linear and non-linear fit in the MNI stereotaxic space with optional resampling of fMRI data in this space.
- reader/writer for MINC1, MINC2, NIFTI and ANALYZE files
- the pipeline system
- A few bricks for fMRI data preprocessing (slice timing, spatial smoothing, temporal filtering).
Brought to you by the SIMEXP lab