• PreviousReleaseNotes

Stable v5.3.0

15 May 2013

Stable release version 5.3.0 is a bug fix release to correct the problem with pial and white surface creation affecting thickness and area measures.

What's New

• New features:
  • 64bit version of freeview (MacOSX Lion). Note: upon opening freeview for the first time, it may display a message prompting to Allow or Deny network connections. This originates from a Qt library which manages File Open functionality. You may select Deny without affecting the functionality of freeview, except perhaps in regards to some network drive accessibility.
  • freeview contains features intended to transition to replacing tkmedit and tksurfer. Also slice scrolling speed is improved.
  • added -first_wm_peak flag to mris_make_surfaces, which may aid more accurate white matter surfaces especially in heavily myelinated areas like visual cortex on data that is higher res than 1mm. enable the flag via an expert.opts file.
• Bug fixes:
  • fix for mris_make_surfaces to correct pial/white surface bug.
  • fix for mri_mask problem affecting edits.
  • fix for qdec problem affecting threshold entry box.
  • fix for Fedora 18, Ubuntu 13, and Gentoo 64 glibc crashes (seg-faults on any binary).
  • fix for recon-all tal-check failure-threshold sensitivity.

Known Issues

• On Mac OSX platforms you need to install XQuartz. Pre-Yosemite versions of OS X require XQuartz 2.7.5, Yosemite requires XQuartz 2.7.6 or higher.

• Mac OSX platforms may get a mris_resample: Command not found error. If so, download the mris_resample, copy it into your $FREESURFER_HOME/bin directory, and try again. Make sure it is executable (chmod a+x mris_resample).

• Opensuse platforms versions greater than 12.3 are not compatible with the FreeSurfer licensing scheme. There are two potential solutions. The first is to copy /lib64/libcrypt-2.14.1.so from an Opensuse 12.1 installation to $FREESURFER_HOME/lib. Then make a symbolic link to it in that directory (ln -s libcrypt-2.14.1.so libcrypt.so.1). Or 2) just download and install the Development Version of FreeSurfer.

• The Talairach transform step will fail at the mri_nu_correct.mni command on newer Linux distributions that come with perl version 5.22 (and higher) due to an incompatibility with the mni tools shipped with FreeSurfer and newer versions of perl. The issue has been fixed in the Development Version of FreeSurfer.

• freeview & Ubuntu: When you run freeview, you get the error "freeview.bin: error while loading shared libraries: libjpeg.so.62: cannot open shared object file: No suchfile or directory." Freeview will work fine if you install libjpeg62-dev and run:

sudo apt-get install libjpeg62-dev

• freeview: on the Fedora 18 platform, when loading a surface overlaid on a volume, the surface might be misaligned. The message "Warning: MatrixInverse failed." may also be displayed in the terminal. A fix or workaround is not yet available.

• fsfast: on the Mac, preproc-sess might fail looking for 3dvolreg.afni. 3dvolreg.afni is just a redistributed copy of AFNIs 3dvolreg, so the workaround is to install AFNI on your own, then: sudo cp 3dvolreg /Applications/freesurfer/bin/3dvolreg.afni Or you can download 3dvolreg.afni from here. (ERROR: cannot find AFNI command 3dvolreg.afni)

• The -dti option in freeview will not display eigenvectors correctly if the orientation of the volumes is anything other than RAS. This will be fixed for the next release.

• Creation of the aseg.stats table fails when only running -autorecon1 or -autorecon2 because ribbon.mgz is not present. This has been fixed. A patch can be downloaded here (ftp://surfer.nmr.mgh.harvard.edu/pub/dist/freesurfer/5.3.0-patch/). Just download the relevant mri_segstats binary and overwrite your existing one.

• mris_preproc --qdec and --qdec-long only works with unix input text files (\n newline). If you have difficulties (i.e. old mac \r or windows \r\n newline files), ask for a fixed version.

• flirt.fsl and other fsl binaries will be missing from the MacOS snow leopard build. Download the the binaries from here. Extract the contents of the tar file into the $FREESURFER_HOME/bin directory. This should result in 6 new files with a .fsl extension in the $FREESURFER_HOME/bin directory.

• recon-all/mri_segstats: when the flags -autorecon1 -autorecon2 are used with recon-all, then a recon-all error exists with message about missing ribbon.mgz file. There are two workarounds: 1. include the flag -autorecon3 (or just use -all, which is the recommendation), or 2. get mri_segstats from here, then:cp mri_segstats $FREESURFER_HOME/bin

• LGI: issue when using Matlab 2013. the fix is to edit the file $FREESURFER_HOME/matlab/SearchProjectionOnPial.m and make these changes:

find this line:

< for t=1:step:size(perim,2)

and change to:

> si=max(size(perim));
> for t=1:step:si

• fslregister (and so bbregister with --init-fsl) requires that FSL be installed. This has been fixed. The fix will be in the next version. Ask for a patch if you need it before then.

• Unpacking DICOM files from the Siemens Skyra scanner fails on MacOS. This has been fixed. Download the new version of mri_convert and copy it to $FREESURFER_HOME/bin/mri_convert

• Longitudinal pipeline, adding time points: there is a bug in recon-all that prevents adding time points to an existing base (usually you should not do this anyway), make these changes to fix it:

in recon-all line 6148 change:

  if ( ! $found_tpNid  ) then

to

  if ( ! $found_tpNid  && ! $DoAddTp ) then

• selxavg3-sess can use a huge amount of memory when running mni305. This is mostly due to poor memory management in matlab. For the next release, selxavg3-sess has been modified to reduce the amount of memory used by a factor of 2 or more.

• mni152reg exits with a "Permission denied" error. Use ftp://surfer.nmr.mgh.harvard.edu/pub/dist/freesurfer/5.3.0-patch/mni152reg instead. Copy it into $FREESURFER_HOME/bin

• recon-all: two issues with using T2 to refine the pial: (1) statistics were computed before the pial surface was refined, and (2) no way to recover from registration errors. This version: ftp://surfer.nmr.mgh.harvard.edu/pub/dist/freesurfer/5.3.0-patch/recon-all fixes these problems. Copy it into $FREESURFER_HOME/bin. If the registration fails, then try using -bbr-init-header.

• Poorly aligned data that can not be fixed by the primary talairach alignment utility will result in a "Can't locate shellwords.pl in @INC error on system running perl5.16 or greater. One possible solution is to use a version of perl pre-v5.16, if one is installed on your system. This is done by modifying the 1st line of file $FREESURFER_HOME/mni/bin/mritotal. For example:

< #!/usr/bin/perl
> #!/usr/bin/perl5.12

Another workaround is to use tkregister to manually align the image, such that mritotal is not run, and the rest of recon-all should run properly.

• Entorhinal and perirhinal labels are incorrect due to a spherical warp that is too rigid. This results in the labels being shifted too medial, so that the pc is often where ec should be. This problem has been fixed for the next release (v6).

• When using the VirtualBox version of freesurfer (freesurfer-Virtualbox-linux-x86-stable-pub-v5.3-full), the "mri_nu_correct.mni" command will fail when executed in the VirtualBox environment on a subject whose data exists on a shared folder between the host and the virtual machine. Currently, the only known solution is to import the subject data folder into the VirtualBox environment, and then run recon-all.

Upcoming Features (v6.0)

• Brain networks (cognitive components) estimated from 10449 Experiments and 83 tasks in the Brainmap database are released in MNI152 and fsaverage space:

  • Wikipage: BrainmapOntology_Yeo2015

  • There are three sets of data (networks+auxiliary information) that are released: (1) Probability that a task would recruit a component (csv files), (2) Probability that a component would activate a voxel/vertex, (3) quantitative measures of functional specificity and flexibility (i.e., whether a voxel/vertex specializes for a specific cognitive component or supports multiple components).
  • The volumetric maps + csv files are found in average/Yeo_Brainmap_MNI152/. See Yeo_Brainmap_MNI152_README in directory for more details.
  • The surface maps are found in the subjects/fsaverage/label directory. See Yeo_Brainmap_fsaverage_README in directory for more details.
• Substructure Segmentation:

  • New hippocampal subfields generation - See HippocampalSubfields

  • New brainstem substructures generation - See BrainstemSubstructures

• Longitudinal:

  • New dedicated longitudinal pipeline for subfield segmentation - See LongitudinalHippocampalSubfields

  • New longitudinal pipeline for tractography
  • Fix for time point addition (see bug in 5.3)
• Matlab Linear Mixed Effects Tools:
  • Updated Matlab LME tools to newer Matlab versions
  • Allow missing Parallel Toolbox (process sequentially)
  • minor improvements to F-test

• recon-all now produces aseg.mgz (subcortical atlas) with Hi-Res data (<1mm). the -hires flag is still necessary to include with recon-all when hi-res data is input. changes to mri_normalize, mri_em_register and mri_watershed were made to support this feature.

• Improved accuracy of ?h.cortex.label
• Improved prevention of surfaces from crossing into the contralateral hemisphere
• Adding the 'fs_update' script which allows users to update binaries or other elements of the freesurfer installation.

• Adding the 'fs_tutorial_data' script which provides a means for downloading the data for completing the Tutorials.

• bbregister now uses the FS mri_coreg program by to initialize BBR. FSL or SPM/matlab are no longer needed. mri_coreg is based on spm_coreg and gives very similar results as to when spmregister is run.
• mri_fdr -- command line program to compute and apply the false discovery rate algorithm
• Bug fix in mri_ca_register that prevented unfolding of lattice and caused many hours of unecessary unfolding and also stopped warp from evolving.
• Add code for Wash. U. HCP to settle white surface near maxima in second directional derivative if -first_wm_peak is specified (off by default).

Stable v5.2.0

2 March 2013

Stable release version 5.2.0 contains new features and bug fixes.

What's New

• New features:

  • For changes in TRACULA, see TRACULA release notes.

  • Mac OS X Lion 64b build - the recon-all binaries are now 64-bit builds, which allows memory usage greater than 4GB, such that memory-intensive programs like mri_cvs_register will complete. Also has CUDA support for mri_em_register and mri_ca_register (-use-gpu flag w/ recon-all).

  • OpenMP parallelization (CentOS 6 distribution only!): mri_em_register and mri_ca_register can now run multi-threaded, decreasing their runtime by about a third. To enable this feature, add the flag -openmp <numthreads> to recon-all, where <numthreads> is the number of threads you would like to run, where 8 is a typical value, assuming you have a four-core system. Note that the results of an openmp run compared to a non-openmp run are identical.

  • T2/FLAIR used for pial surface refinement: if you have T2 or FLAIR scans ~1mm^3, then they can be used by the recon-all stream to refine the pial surfaces. We recommended a T2-space that is bandwidth, geometry and readout matched to the mprage so there is no differential distortion. These scan types allow dura to be distinguished from grey matter, which is generally very difficult to do in a T1 scan, resulting in incorrect pial surfaces containing sections of dura (which otherwise would require manual editing of the brainmask.mgz file to remove). Using this feature requires two flags described below. Note that these two flags can be run on prior-processed data if you want to just refine the pial surfaces, but you should run using the flags -autorecon3 and -T2pial, which will re-run the portions affected by a pial surface change. The -T2pial flag will create the file T2.mgz, which is conformed and registered using bbregister (called FLAIR.mgz for the -FLAIRpial flag).

    • -T2 <path_to_T2_dicom> or -FLAIR <path_to_FLAIR_dicom> to import one or the other (not both!), analogous to the -i <path_to_T1_dicom> flag

    • -T2pial or -FLAIRpial which will refine the pial surfaces making use of the T2 or FLAIR

  • New matlab toolbox for longitudinal analysis: LME is a new Matlab toolbox (found in freesurfer/matlab/lme directory) for the statistical analysis of longitudinal neuroimaging data with longitudinal linear mixed effects models. See LinearMixedEffectsModels.

  • New cortical parcellations and labels:
    • Parcellations estimated by intrinsic functional connectivity from 1000 subjects:

      • Wikipages: CorticalParcellation_Yeo2011, CerebellumParcellation_Buckner2011, StriatumParcellation_Choi2012

      • Cortical Parcellation distributed in subjects/<fsaverage, fsaverage5, fsaverage6>/labels/. Files are ?h.Yeo2011_<7 or 17>Networks_N1000.annot and confidence maps ?h.Yeo2011_<7 or 17>NetworksConfidence_N1000.mgz. Use mri_surf2surf to map these to a subject.

      • Cortical, Cerebellar, Striatum Parcellation in MNI152 space found in average/<Yeo_JNeurophysiol11_MNI152>, average/<Buckner_JNeurophysiol11_MNI152> and average/<Choi_JNeurophysiol12_MNI152>. See README file in each directory for more details. Note that these README files will be updated to reflect minor improvements in parcellations beyond the publication wikipages

    • OASIS-based Cortical Hubs (ROIs). See The Dynamics of Cortical and Hippocampal Atrophy in Alzheimer Disease. Distributed in fsaverage/labels dir, named '?h.oasis.chubs.*'. Use mri_label2label to map these to a subject.

    • perirhinal cortex labels based on exvivo data. The labels are named <subjid>/label/?h.perirhinal.label, and produced by recon-all with the -balabels flag (which is included by default). See Predicting the location of human perirhinal cortex, Brodmann's area 35, from MRI.

    • Mindboggle cortical atlas. If using this atlas in a study, please be sure to cite their Frontiers article. The new DKTatlas40 atlas is automatically used by recon-all during new -cortparc3 and -parcstats3 stages, which will create ?h.aparc.DKTatlas40.annot/.stats files.

  • CVS target atlas in MNI305 space (affecting mri_cvs_register).
  • Cubic B-Spline interpolation was added to mri_convert (-rt cubic), mri_vol2vol and the mri_robust... commands. The mri_robust_register and mri_robust_template use cubic interpolation by default for the final output and the interal Gaussian pyramid.

  • New gm/wm contrast cortical measure: new utility 'pctsurfcon' computes surface-wise gray/white contrast, to file ?h.w-g.pct.mgh. See Age-associated alterations in cortical gray and white matter signal intensity and gray to white matter contrast. This stage can be run separately by using the flag -pctsurfcon with recon-all.

  • recon-all flags:
    • added -3T flag to enable the 3T-specific NU intensity correction parameters (aka -nuintensitycor3t) and usage of the special 3T atlas for Talairach alignment (aka -schwartzya3t-atlas)

    • added -norm2-b and -norm2-n flags to pass -b and -n params to the second instance of mri_normalize. Used to pass parameters found to improve segmentation in the anterior temporal lobes, such as: recon-all -s subjid -all -norm2-b 20 -norm2-n 5 Note that -norm1-b and -norm1-n flags are available to adjust the first instance of mri_normalize if needed.

    • added -cubic flag to allow enabling usage of -rt cubic by mri_convert during conform (on by default in dev, off by default in stable)
  • longitudinal stream:
    • now uses cubic interp by default
    • can run subjects with single time point (by creating an artificial, upright and straight base image). This allows the inclusion of single time point subjects into the statistical analysis to increase power (e.g. in a linear mixed effects model) and reduces potential bias from processing subjects with more time points differently.
  • mri_segstats ICV output name changed to eTIV (Estimated Total Intracranial Volume) to reduce confusion.
  • mri_ca_label: including the -bigventricles recon-all flag no longer affects mri_ca_label (just mri_ca_register)
  • fsaverage: recon-all checks if fsaverage directory in SUBJECTS_DIR corresponds to the 5.2 version which contains several new files. If a previous version of fsaverage is detected, recon-all will delete the old fsaverage and create a new symlink.
• Bug fixes (see Known Issues of last release for bug details):
  • fix for tksurfer "Configure Overlay Display" clipped for low resolution monitors issue
  • fix for mri_robust_register final half way images and maps are incorrect issue
  • fix for preproc-sess with -per-run flag issue on the Mac
  • fix for recon-all producing differents results on the second run

  • fix for mris_calc failure on long filenames (>255)

  • fix for mris_anatomical_stats underestimating the volume because the volume is based on the area of the white surface
  • fix for control points adversely affecting aseg (no longer used in the mri_ca_normalize stage)
  • fix for the bug in the way that FS converts surface-based area (eg, lh.area) and volume (eg, lh.volume) into standard space
  • fix for frequent talairach alignment failures (re-implemented NU correction prior to talairach)

  • fix for incorrect SupraTentorial volume in aseg.stats

  • fix for mri_label2vol excluding the last label
  • fix for mris_expand bug with filename written when multiple surfaces saved
  • fix for mris_make_final_surfaces to fix pial crossing issue

  • fix for '_FindFacePath' issue in mris_topo_fixer

Known Issues

• mris_make_surfaces: where wm lesions exist, even if the lesion is filled in wm.mgz volume, when mris_make_surfaces is run to create the final surfaces, because it uses brain.finalsurfs.mgz for its intensity info, and doesnt consider the fill data in wm.mgz, its possible for surfaces to not follow grey matter on the perimeter of the lesion. A fix will appear in a future release. A potential work-around is to set lesion voxels in brainmask.mgz and norm.mgz to 110 and lesions voxels in aseg.mgz to 77 (lesion mask can be derived manually or from an automated algorithm), then run recon-all -autorecon2-cp. To set a voxel value, in tkmedit, goto Tools->Configure Volume Brush, and enter the value (110) into the New Value box, then press Enter in the New Value box. Voxel edits will get the 'New Value' instead of the default 255 value when button 2 is pressed.

• qdec: in the Display tab, when a value is entered in the Threshold Min entry box, the display is not updated. This will be fixed in the next release.

• Fedora 18: freesurfer does not work on Fedora 18. the problem is rooted in some code that improperly uses the crypt function, which causes most binaries to crash. the only workaround is to use Fedora 17.

• mri_mask: bug that results in empty brain.finalsurfs.mgz when voxels have been edited off. a fix will appear in the next release.
• mris_make_surfaces: bug that results in inaccurate pial and white surfaces affecting thickness and area measures. surface registration is unaffected. a fix will appear in v5.3.

• recon-all: the failure threshold for the tal_QC_AZS test script in the -tal-check stage was set to low, which could cause a false-positive failure, indicating a talairach failure. a fix will appear in v5.3, or edit recon-all and find/replace two instances of "< -36" with "< -60" in the DoTalCheck section.

Upcoming Features (release v5.3 or v6.0)

• Additions to Freeview so that it is a total replacement for tkmedit, tksurfer and tkregister.
  • -freeview flag added to tkmedit-sess to have it (FS-FAST) use freeview instead of tksurfer
• Splitting the code and data in the distribution so that the download is much smaller, but necessitating individual download of data (atlases, templates, training data, etc) when it is needed. In the meantime, if you want to free-up space based on features you dont think you will need (retaining what is needed to run the default stream), then here is a list of directories of substantial size, their feature/function, and size, in total about 3GB:
  • freesurfer/average/mult-comp-cor - Correction for Multiple Comparisons pre-calc - 1.3G
  • freesurfer/subjects/fsaverage? - fsaverage at other iso-levels - 310M
  • freesurfer/subjects/fsaverage_sym - symmetric fsaverage for contra-lateral surface mapping - 197M
  • freesurfer/subjects/cvs_avg* - CVS Combined Volume and Surface registration target - 531M
  • freesurfer/trctrain - Tracula training set - 886M

---

Stable v5.1.0

24 May 2011

Stable release version 5.1.0 contains new features and bug fixes, and is covered under new license agreements.

What's New

• New license:

  • Freesurfer, with the exception of the 'tktools', is now being released under a new 'BSD-like' license: FreeSurferSoftwareLicense. This license text is found in the file named LICENSE in the freesurfer installation directory, and in the agreement text of the license registration page. The 'tktools' (tkmedit, tksurfer and tkregister2) remain under the CorTechs license.

• New features:

  • Tracula: new application handling white matter tracts, constrained by anatomy

  • GEMS: segmentation tools, used to segment hippocampal subfields

  • Longitudinal Features:

    • Longitudinal Stream: the longitudinal runs are now resampled to the base/template space to further reduce variability and allow for better comparison and quality checks. See LongitudinalProcessing51 for details.

    • Longitudinal Editing: manual edits are transferred from [CROSS] to [LONG] and in some cases from [BASE] (e.g. brainmask). This significantly simplifies editing in longitudinal processing. In many cases it is sufficient to edit only the [BASE] instead of all time points. For a detailed description of the individual edits see LongitudinalEdits.

    • Longitudinal Post-Processing: several scripts have been added to post-process longitudinal data. It is now possible to create within subject rate maps of cortical thinning and compare rate of change between groups. See FsTutorial/LongitudinalTutorial for details.

    • Longitudinal Affine [BASE]: in order to remove callibration differences of the scanner across time, the base can now be created with an affine registration instead of rigid (pass -base-affine instead of -base).
    • FSFAST - now masks the volume-based analysis to include only subcortical areas
  • mri_cvs_register updates:
    • besides its pair-wise application, now subject-to-template registration is possible
    • there is an implicit cvs atlas distributed that is the target of cvs_register unless otherwise specified
    • using more compact registration warp files to save the registration results (m3z vs tm3d)
    • allows for single hemi registration
    • ** NOTE ** the new cvs atlas will be released at the end of June due to some unforeseen delays; mri-cvs-register remains functional in the meantime, but the user needs to explicitly name a template subject

    • ** UPDATE ** we found a bug in the mri_cvs_register script after the release. Please, download the new version of applyMorph and mri_cvs_register from one of the below links: Linux Centos4, Linux Centos4 x86 64 and Mac Leopard

  • mris_preproc: now masks out non-cortical areas when the input is a volume (keep them with --no-mask-mask-noncortex). For paired differences, now zeros voxels for all subjects if a voxel is 0 for any time point for any subject.
  • mri_ca_register and mri_ca_label: updated intensity distribution estimation to better handle 3T data; also allow more dramatic warps to properly segment large ventricle cases
  • remove -force flag, because it was incorrectly suggesting that results get overwritten. If an existing subject is re-run using the -i flag, recon-all will now stop. The user has two options: 1. re-run on the same inputs (then simply omit the -i flags) or 2. re-run with new inputs (then first rename or remove this subject's directory and start from scratch)
  • mri_gcut is no longer run by default during -skullstrip, due to a rare tendency to remove gray matter and/or cerebellum. It can still be used (and works quite well in removing dura) by adding -gcut to the end of recon-all. Just be sure to inspect the brainmask.mgz for excessive stripping (which should be done in any case).

  • mri_ca_normalize now makes use of user control points (which a user would add manually to help define white matter). Note: see Known Issues below for an issue concerning this feature.

• Bug fixes (see Known Issues of last release for bug details):
  • fixed mri_convert "near-zero valued voxels in rawavg.mgz" issue.
  • fixed mri_robust_template linear transform for motion correction and longitudinal processing issue.
  • fixed longitudinal base name partial match and long name issues.
  • fixed freeview not running on Snow Leopard issue.
  • fixed mris_make_surfaces incorrectly labelling aseg gm near calcarine issue.
  • fixed mri_nu_correct.mni conversion from float to uchar issue.
  • mris_make_face_parcellation was merging some parcellation units into a single unit. This is now fixed
  • mri_segstats partial volume correction had a bug which is now fixed.
  • mri_segstats had a bug when using partial volume correction with data that did not have a 1mm voxel size
  • preproc-sess (FSFAST) when a subject other than fsaverage was specified it was not used. This has been fixed in the next vesion.
  • mris_make_face_parcellation would sometimes merge units into a single unit resulting in too few parcellations. This has been fixed in the next vesion.
  • mris_make_face_parcellation documentation suggested using the inflated surface. This is no longer allowed in the next vesion. You should use sphere.reg or sphere

Known Issues

• mris_make_surfaces: where wm lesions exist, even if the lesion is filled in wm.mgz volume, when mris_make_surfaces is run to create the final surfaces, because it uses brain.finalsurfs.mgz for its intensity info, and doesnt consider the fill data in wm.mgz, its possible for surfaces to not follow grey matter on the perimeter of the lesion. A fix will appear in a future release. A potential work-around is to set lesion voxels in brainmask.mgz and norm.mgz to 110 and lesions voxels in aseg.mgz to 77 (lesion mask can be derived manually or from an automated algorithm), then run recon-all -autorecon2-cp. To set a voxel value, in tkmedit, goto Tools->Configure Volume Brush, and enter the value (110) into the New Value box, then press Enter in the New Value box. Voxel edits will get the 'New Value' instead of the default 255 value when button 2 is pressed.

• tksurfer: where the "Configure Overlay Display" is clipped for low resolution monitors. The fix is here. Copy this file to $FREESURFER_HOME/lib/tcl/. And this applies only to FreeSurfer v5.1.0. The next FreeSurfer release will include this fix. Image Courtesy: Gonzalo Rojas Costa

• mri_robust_register: the final half way images and maps are incorrect in situations where the registration does not terminate after maxit (default 5) iterations on the highest resolution. A fix will be included in the next release. Workaround: use mri_robust_template with the two inputs.

• preproc-sess with -per-run flag: when you receive the error "no run directories found", otherwise known as the "getrunlist" encountered sometimes on Mac platforms. To fix this, download this. Then run 'cp getrunlist $FREESURFER_HOME/fsfast/bin' to copy getrunlist to the proper FreeSurfer sub-directory.

• if recon-all is run a second time, the results may differ slightly. the reason is that the file talairach_with_skull.lta is created during the skull-strip stage, and created again after the aseg stage, at which point it may differ slightly due to usage of the aseg, thus, if recon-all is run again, during skull-strip, this new talairach_with_skull is used, resulting in a slightly different brainmask.mgz, which affects downstream results. There is no workaround. This issue will be fixed in v5.2 (by naming the first talairach_with_skill.lta something else so that the later one is not reused, despite the possibility of the later one being more accurate).

• mris_calc fails on long filenames (>255). A fix will be included in the next release. The fix is also available here(32b) and here(64b).

• mris_anatomical_stats underestimates the volume because the volume is based on the area of the white surface.
• Control points may adversely affect aseg: beginning with v5.1, control points (added to fix white matter intensity issues, per normal procedures) are now used in the mri_ca_normalize stage which precedes the creation of the aseg. This functionality was added to v5.1 to help the aseg creation. However, we have discovered that it is possible for this normalization, making use of the control points, to adversely affect the subcortical segmentation, such that some structures differ in size when compared to processing without control points being used by mri_ca_normalize. This effect seems to occur only in instances where adding control points to regions where signal drop-off within white matter is dramatic (even while correcting the white matter intensity issue as intended). We did not observe this affect within our own test sets. The advice we give is to be sure that control points are not placed into regions which share or near voxel intensity values close to that of grey matter (including subcortical grey matter structures such as pallidum and thalamus). An option is to disable the running of mri_ca_normalize when re-running the -autorecon2 or -autorecon2-cp stage after adding control points by adding the flag -nocanorm to the end of recon-all. We will continue to investigate a more automated solution to detection of this problem. The more permanent workaround for v5.1 users is to edit their recon-all script making the following change, which will disable usage of control points with ca_norm:

# find these lines:
set cmd = (mri_ca_normalize)
if($UseControlPoints)  set cmd = ($cmd -f $ControlPointsFile)

# and comment-out the second line like this:
set cmd = (mri_ca_normalize)
#if($UseControlPoints)  set cmd = ($cmd -f $ControlPointsFile)

# then re-run your subjects with the flags: -autorecon2 -autorecon3 -clean-aseg

• There is a bug in the way that FS converts surface-based area (eg, lh.area) and volume (eg, lh.volume) into standard space. The files themselves are fine, it is the way they are transfered to standard space that is the problem. You will not need to re-analyze any of your individual subject data. The volumes of the subcortical structures are unaffected by this problem. Global measures of cortical volume are also unaffected.

• recon-all now runs the talairach stage (creation of talairach.xfm) before the nu_correct stage (creation of nu.mgz) because creation of nu.mgz now makes use of the talairach.xfm file (to roughly identify white matter for intensity scaling purposes). the implication of this change is that the talairach stage now takes orig.mgz as input (instead of nu.mgz) which may lead to tal-check failures on some data. See this page for workarounds. In the next release of freesurfer, this will be run automatically if tal-check fails, but more importantly, in the next release, nu_correct will run on orig.mgz to create orig_nu.mgz which will be used as input to the talairach stage (after which orig_nu.mgz will be deleted). It may also be the case that the ICV (eTIV) calculation is inaccurate, due to a poor quality talairach alignment (which might none-the-less be suitable for the recon to work). The workaround is to run these commands which will update the aseg.stats file with a new more accurate ICV value:

cd subjid/mri
talairach_avi --i nu.mgz --xfm transforms/talairach.xfm
recon-all -s subjid -segstats

• The SupraTentorial volume in aseg.stats is incorrect (generally yields volumes that are too high). A fixed linux version can be obtained from ftp://surfer.nmr.mgh.harvard.edu/transfer/outgoing/flat/greve/mri_segstats.linux . Copy this into $FREESURFER_HOME/bin/mri_segstats, then run recon-all -segstats to generate a new aseg.stats file. A fix will be in the next release.

• mri_label2vol excludes the last label. Fixed linux version here ftp://surfer.nmr.mgh.harvard.edu/transfer/outgoing/flat/greve/mri_label2vol Copy to $FREESURFER_HOME/bin. A fix will appear in the next release.

• mris_expand has a bug with filename written when multiple surfaces saved. a fix will be in the next release.
• Do not use DOSS with QDEC
• Longitudinal pipeline, adding time points: there is a bug in recon-all that prevents adding time points to an existing base (usually you should not do this anyway), make these changes to fix it:

in recon-all line 5407 change:

  if ($status) then

to

  if ( $status && ! $DoAddTp ) then

---

Stable v5.0.0

16 Aug 2010

Stable release version 5.0.0 is a major release containing new features and bug fixes.

What's New

• New features:

  • mri_cvs_register: a new pairwise registration tool to produce highly accurate cortical and subcortical alignment; the inputs to the algorithm require two subjects that have been already processed by the recon-all stream. For further reference, see Combined Volumetric and Surface Registration.

  • mri_robust_template: now used to do motion correction (production of the rawavg.mgz file when two input structurals are present), replacing usage of the FSL 'flirt' tool. The recon-all flag -motioncor-flirt can be used to revert back to flirt over robust_template for motion correction stage.

  • mri_gcut: the skull-strip stage includes a new step following mri_watershed, where the new utility 'mri_gcut' is run to refine the skull-strip, resulting in removal of dura and skull in some cases where it might otherwise remain after the mri_watershed step. See Skull Stripping Using Graph Cuts.

  • mri_glmfit-sim and qdec: correction for multiple comparisons by method of Monte Carlo simulation is now instantaneous for most common parameter settings. In qdec, a new menu option allows selection. mri_glmfit-sim --help describes its own usage. Raw data is found in $FREESURFER_HOME/average/mult-comp-cor directory.
  • mris_volmask: completely rewritten, eliminating dependency on the VTK library (a bane to usage in many clusters). Results (?h.ribbon.mgz files) are identical.
  • mri_nu_correct.mni: will use float internally instead of uchar. It will also rescale the output so that the global mean is the same as that of the input. These two changes are linked and can be turned off with --no-float.
  • 32b Linux now produces the same results as 64b Linux. this requires that the processor support SSE2 instructions, which became available in the Pentium 4 series (around 2004).
  • aseg.stats changes:
    • Removed:

      • BrainMask, BrainMaskVol, BrainMaskNVox, BrainSeg, BrainSegVol and BrainSegNVox measures

      • volume-based Cortical Gray and White Matter measures
    • New:
      • lhCortex, lhCortexVol: Left hemisphere cortical gray matter volume
      • rhCortex, rhCortexVol: Right hemisphere cortical gray matter volume

      • Cortex, CortexVol: Total cortical gray matter volume (based on surface-stream)

      • SubCortGray, SubCortGrayVol: Subcortical gray matter volume

      • TotalGray, TotalGrayVol: Total gray matter volume

      • SupraTentorial, SupraTentorialVol: Supratentorial volume

      • lhCorticalWhiteMatter, lhCorticalWhiteMatterVol: Left hemisphere cortical white matter volume
      • rhCorticalWhiteMatter, rhCorticalWhiteMatterVol: Right hemisphere cortical white matter volume

      • CorticalWhiteMatter, CorticalWhiteMatterVol: Total cortical white matter volume

  • GPU support:

    • beta-release of some binaries making use of an NVidia GPU card, reducing runtime by 3x to 20x: mri_em_register, mri_ca_register, mris_inflate and mris_sphere. Adding the flag -use-gpu to recon-all will enable GPU support on CUDA-compatible Nvidia cards for these and future freesurfer executables.

    • currently only released on the linux centos4 and centos4_x86_64 distributions (others soon).
    • be sure to consult the Known Issues section below, as GPU development is on-going and updates will be provided periodically.

  • Support for reading/writing annotation/colortable data in Gifti LabelTable format.

  • -show-edits command will show 'persistent' edits. See Edits.

  • -cw256 works better than -cropsize256 so -cropsize256 has been removed.
  • Reconciled ?h.cortex.label with Desikan and Destrieux parcellations: ?h.cortex.label is used as the template defining cortex, and is used in mris_ca_label to correct any medial wall discrepancies, and is used in mris_anatomical_stats as a mask.
  • A new Desikan atlas, dated 2010-03-25, has a different color for temporalpole (the old color was gray, which looked like the default tksurfer surface color).
  • tksurfer will now load multiple overlays from the command-line.
  • QDEC:
    • Added Nuisance Factor menu in Design tab, allowing selection of any number of continuous variables of which will be treated as 'nuisance variables' in the glm (contrast matrix will have value '0' for these).

    • Added FindClustersAndGotoMax button to run mri_surfcluster on a selected result, which will find clusters based on currently selected display threshold, output a table of results to the terminal, and move the cursor to the max vertex in the #1 cluster. 'Next' and 'Prev' buttons allow cycling movement of the cursor through the found clusters.

    • Added QuickSnapsTIFF option to automatically take snapshots of medial/lateral pial/inflated and save these four pics to an automatically generated filename.
    • Added optional parsing of an 'ignore.factors' file to allow specifying columns in the qdec.table.dat to ignore (as would be case for columns containing alternate subject IDs).
    • Added optional parsing of an 'ordinal.factors' file to allow specifying which continuous factors are ordinals (integers, not floating point). only affects 'Save Data Table' menu option where the qdec.table.dat is written to disk (ordinals are not written as floats).
    • Added usage of NaN (or nan) to indicate missing data point, allowing automatic exclusion of that subject from an analysis.
    • Added correction for multiple comparisons by Monte Carlo cluster-wise simulation making use of pre-run data, making run-time near instantaneous.

  • -nuintensitycor-3T flag to recon-all: optimal parameters for nu_correct for 3T scans, as described in Improvement of brain segmentation accuracy by optimizing non-uniformity correction using N3.

  • new EC labels: recon-all will now by default produce labels for entorhinal cortex based on exvivo data. The labels will be named <subjid>/label/?h.entorhinal_exvivo.label, and can also be produced with the -label-exvivo-ec flag.

  • new V1 labels: recon-all will optionally produce labels for V1 based on a probabilistic atlas (see Accurate prediction of V1 location from cortical folds in a surface coordinate system). The labels will be named <subjid>/label/?h.v1.predict.label, and created via the -label-v1 flag.

  • new PALS-B12-based ROIs: see PALS_B12 wiki page.

  • FS-FAST:
    • New dicom unpacker dcmunpack (can still use unpacksdcmdir)
    • Requires paradigm files to be 3 or 4 columns.
    • Explicit registration template creation
    • Motion correction and function-structure registration performed on a run-by-run basis
    • Function-structure registration performed with BBR
    • Raw time series data are sampled directly into output group space (lh, rh, and mni305)
    • Raw time series can also be sample into subject's individual cortex.
    • Masked smoothing
    • Better retinotopy support
    • Can now use any number of task or nuisance external regressors
    • mkanalysis-sess GUI has been disabled
    • subcortical mask for group mni305 analysis
    • can now be configured to use octave instead of matlab. This continues to be experimental.
• Bug fixes:
  • fixed mri_ca_label bug where this message might appear: "GCA sequential renormalization: label 28 not consistently computed."
  • fixed mris_make_face_parcellations where it created labels with spaces in the name, which causes table parsing in aparcstats2table to fail.

  • fixed aparcstats2table, asegstats2table where output of the parcellations/segmentations appeared in a random order compared to the input aseg.stats/?h.aparc.stats files.

  • fixed mris_volmask: removed VTK library dependency.
  • fixed mri_watershed where this message might appear: "GLOBAL region of the brain empty !".
  • fixed make_average_subject out-of-memory problem.

Known Issues

• recon-all longitudinal, and cross-sectional motion correction stage: mri_robust_template inherited a problem from the transform libraries which breaks the longitudinal stream, and in some rare case, the cross-sectional -motioncor stage might fail. Fixed mri_robust_register and mri_robust_template binaries can be downloaded for your platform from here and should replace the binaries from the 5.0 download. This fix will be present in the next release.

• recon-all longitudinal: if the base name partially matches the name of a timepoint, the name error checking gives a false positive. A fix will appear in the next release. A new recon-all script can be downloaded from here.

• mris_make_surfaces: where wm lesions exist, even if the lesion is filled in wm.mgz volume, when mris_make_surfaces is run to create the final surfaces, because it uses brain.finalsurfs.mgz for its intensity info, and doesnt consider the fill data in wm.mgz, its possible for surfaces to not follow grey matter on the perimeter of the lesion. A fix will appear in a future release. A potential work-around is to set lesion voxels in brainmask.mgz and norm.mgz to 110 and lesions voxels in aseg.mgz to 77 (lesion mask can be derived manually or from an automated algorithm), then run recon-all -autorecon2-cp.
• freeview does not work on Mac OS Snow Leopard.

• _cuda GPU binaries: these will have ongoing fixes. The lastest _cuda binaries can be downloaded from here. If you find problems with those, please post details to the list. These are 'beta', in that they are not fully tested and should not be used in a formal analysis quite yet.

• qdec: if the right hemi surface is loaded manually, and a label is subsequently drawn or loaded, and then the "Map Label to Subjects" is selected, then the mri_label2label incorrectly uses the lh as the hemi. A fixed qdec.bin is available here, and will be present in the next release.

• mri_concatenate_lta (in longitudinal -base run): when using long subjects ids especially for the base, mri_concatenate_lta cannot open the corresponding lta (transform files) in the -base stream. Workaround: use shorter subject names. A fix for this will appear in the 5.1 release.
• mris_make_surfaces: in some cases, the aseg labelled gm near calcarine as ventricle and therefore the surfaces didn't go all the way in. There is no workaround. A fix will appear in the 5.1 release.

• recon-all and mris_make_surface: in some cases it is possible for surfaces to overlap near the ventricles. To fix this, the create surfaces stage (-finalsurfs) has been eliminated and replaced with two stages: -white and -pial, such that pial surface creation can benefit from the cortical parcellation. The -pial stage runs after the cortical parcellation stage (-cortparc) has completed. This means pial surfaces created from prior versions should not be mixed with surfaces created with this version in a group analysis. This change will appear in the 5.1 release.

• mri_convert: near-zero valued voxels in rawavg.mgz where not being set to zero in orig.mgz, so that the background was not completely zero. There is no workaround. A fix will appear in the 5.1 release.
• mri_nu_correct.mni: during the conversion from float to uchar, it is possible for the voxel values outside the brain to skew the intensity histogram in such as a way as to disrupt later stages. A fix for this will appear in the next release. A workaround is to run mri_nu_correct.mni with the --nofloat flag.
• fsaverage Desikan and Destrieux cortical parcellations: these have been rerun making use of the new mris_ca_label which reconciles the ?h.cortex.label with the cortical parcellations. in particular, this affects the medial wall, which now has 'None' where cortex does not exist (eliminating the labels 'unknown', 'medial wall' and 'corpus callosum'.
• Do not use DOSS with QDEC

---

Stable v4.5.0

11 Aug 2009

Stable release version 4.5.0 is a minor release containing bug fixes and new features affecting both the default (cross-sectional) and longitudinal analysis streams.

What's New

• Bug fixes:
  • fixed mri_glmfit-sim, mri_glmfit, and mri_surfcluster: when sign was set to abs, gave the same significance for all clusters.
  • fixed mris_volmask (the -cortribbon stage of recon-all): segfaulted if not enough memory was available. a memory leak was plugged.
• New features:

  • improvements to the longitudinal stream further reducing bias found between cross-sectional and longitudinal runs. See LongitudinalChangeLog.

  • updated C.Destrieux atlas (aparc.a2009s). See DestrieuxAtlasChanges.

  • -norandomness is the the default. The binaries in the recon-all stream which are subject to variability due to usage of a random number generator (mris_smooth, mris_sphere, mris_fix_topology, mris_topo_fixer and mris_ca_label) are now seeded with the same number, with the result being that repeated runs of the recon-all stream on a subject on the same platform will always give identical results. The -randomness flag can be used to allow usage of a random seed.
  • updated versions of aparcstats2table and asegstats2table have:
    • ability to choose common segmentations or all segmentations in asegstats2table
    • different invocations of these scripts produce outputs which have segs/parcs in the same order

    • better debug information ( with the -v or --debug switch )

  • support for reading .vtk surface format, and writing patch files in gifti (.gii) format
  • built against static libtiff, negating the problem of some systems having shared lib libtiff.so.3 and others libtiff.so.4

Known Issues

• mri_ca_label: this message might appear: "GCA sequential renormalization: label 28 not consistently computed." prior to exiting. This error is invalid (an exit should not occur). There is no work-around. A fix will appear in the next release. Or users can download a new mri_ca_label for their platform here.

• mris_make_surfaces: where wm lesions exist, even if the lesion is filled in wm.mgz volume, when mris_make_surfaces is run to create the final surfaces, because it uses brain.finalsurfs.mgz for its intensity info, and doesnt consider the fill data in wm.mgz, its possible for surfaces to not follow grey matter on the perimeter of the lesion. A fix will appear in a future release. There is no work-around.
• make_average_volume: can exit with 'Cannot alloc memory' error if there are a lot of subjects. The work-around is to include the flag '-no-aseg', which will prevent creating the aseg.mgz.
• mris_make_face_parcellations: creates labels with spaces in the name, which causes table parsing in aparcstats2table to fail. a fix (using underscores) will appear in the next release.

• aparcstats2table, asegstats2table: output the parcellations/segmentations in a random order compared to the input aseg.stats/?h.aparc.stats files. A fix will appear in the next release. For 4.5.0 users, they can be downloaded from here. Copy the files asegstats2table, aparcstats2table and datastruct_utils.py to your $FREESURFER_HOME/bin.

• mris_volmask: the VTK library dependency can be problematic on some systems, particularly clusters. The next release will include a new mris_volmask with the VTK library dependency removed.
• mri_watershed: some subjects will fail the skull-strip stage with a message containing "GLOBAL region of the brain empty !". A fix for this will appear in the next release. A work-around is to include the -no-wsgcaatlas flag at the end of recon-all.
• make_average_subject: an out-of-memory error can occur with large subject sets (300+) due to memory leaks in mri_average and mris_make_average_surface. A fix for this will appear in the next release.
• mri_concatenate_lta (in longitudinal -base run): when using long subjects ids especially for the base, mri_concatenate_lta cannot open the corresponding lta (transform files) in the -base stream. Workaround: use shorter subject names. A fix for this will appear in the 5.1 release.

---

Stable v4.4.0

11 June 2009

Stable release version 4.4.0 is a minor release containing bug fixes affecting the longitudinal analysis stream.

What's New

• Bug fixes:

  • longitudinal processing (recon-all -long) was biased toward the timepoint selected as the base. The new longitudinal processing stream is described in the --help output of recon-all, and extensively here: LongitudinalProcessing.

Known Issues

• mris_make_surfaces: where wm lesions exist, even if the lesion is filled in wm.mgz volume, when mris_make_surfaces is run to create the final surfaces, because it uses brain.finalsurfs.mgz for its intensity info, and doesnt consider the fill data in wm.mgz, its possible for surfaces to not follow grey matter on the perimeter of the lesion. A fix will appear in a future release. There is no work-around.
• make_average_volume: can exit with 'Cannot alloc memory' error if there are a lot of subjects. The work-around is to include the flag '-no-aseg', which will prevent creating the aseg.mgz.
• mri_glmfit-sim, mri_glmfit, and mri_surfcluster: when sign is set to abs, gives the same significance for all clusters. This has been fixed for the next version.
• mris_volmask (the -cortribbon stage of recon-all) will segfault if not enough memory is available. A fix will appear in the next release. The workaround is to run a single subject at a time and ensure that 4GB of ram is available.

---

Stable v4.3.1

19 May 2009

Stable release version 4.3.1 is a minor release containing bug fixes (but the changes do not change the results in the default analysis stream relative to v4.3.0), and new features.

What's New

• Bug fixes:

  • fixed FS-FAST GUI bug.

  • fixed mri_nu_correct.mni bug: incorrectly used the -distance flag when --fwhm is passed.
  • fixed mri_normalize bug on Mac OS X: using the -no1d flag caused segfault.
  • some files supporting aparcstats2table and asegstats2table that were missing are now included.
• New features:
  • fsaverage (the default common-space subject) now contains the insula label.

Known Issues

• mris_make_surfaces: where wm lesions exist, even if the lesion is filled in wm.mgz volume, when mris_make_surfaces is run to create the final surfaces, because it uses brain.finalsurfs.mgz for its intensity info, and doesnt consider the fill data in wm.mgz, its possible for surfaces to not follow grey matter on the perimeter of the lesion. A fix will appear in a future release. There is no work-around.
• make_average_volume: can exit with 'Cannot alloc memory' error if there are a lot of subjects. The work-around is to include the flag '-no-aseg', which will prevent creating the aseg.mgz.
• longitudinal processing (recon-all -long) is biased toward the timepoint selected as the base. There is no work-around. A new longitudinal processing scheme will appear in a future release.

---

Stable v4.3.0

26 March 2009

Stable release version 4.3.0 is a minor release containing bug fixes (that change some results in the default analysis stream) and new features.

What's New

• Bug fixes:
  • fixed aparcstats2table bug: if some of the subjects in a group had missing values of a certain label, aparcstats2table failed to designate a 0 for them, leaving the number of columns for that subject short by one (or specifically, by the number of missing labels relative to the other subjects), thereby making the table nearly useless.
  • fixed isxconcat-sess bug: used --projfrac-avg 0 1 0.2 regardless of what is specified, which meant it would average across the cortical ribbon along the surface normal. The new default is to use a projection fraction of 0.5.

  • fixed wmparc.stats - the BrainSeg, !BrainSegNVox and BrainSegNotVent values are now correct.

  • fixed mri_convert bug: when using the -oc flag (which is used during make_average_subject) during make_average_subject, the tal transform and the centering (-oc 0 0 0) were applied in consecutive transforms instead of being composed into one. The first translated it out of the FOV and the 2nd translated it back-in, exposing the cropped FOV (typically posterior) from the first transform.
  • fixed tkmedit bug: seg-faulted on some systems when scrolling and middle-button clicking at the same time over the tkmedit volume image.

  • the estimated Total Intracranial Volume (eTIV) calculation has been found to be more accurate if the talairach.xfm transform is used instead of the talaiarch_with_skull.lta transform. in this release (4.3.0), the new corrected value will appear in the aseg.stats file, and will be output by mri_segstats and mri_label_volume. there is no simple scale factor to correct existing measures.

  • fixed mri_average bug: -sqr option would overflow for short data-type images.
  • fixed recon-all -make all bug: the -finalsurfs stage incorrectly depended on brain.mgz instead of brain.finalsurfs.mgz, so if brain.finalsurfs.mgz is edited, recon-all -make all wont detect the changed file.
• New features:
  • new insula label in cortical parcellation atlas

Known Issues

• aparcstats2table and asegstats2table: are missing four helper scripts which raises a Python import error. These can be downloaded from the following locations and put in $FREESURFER_HOME/bin/. This will be fixed in the next version. ( Note that since these are scripts, they work on all platforms without the need for compilation. )

  • fsutils.py

  • subject_info.py

  • datastruct_utils.py

  • misc.py

• mris_make_surfaces: where wm lesions exist, even if the lesion is filled in wm.mgz volume, when mris_make_surfaces is run to create the final surfaces, because it uses brain.finalsurfs.mgz for its intensity info, and doesnt consider the fill data in wm.mgz, its possible for surfaces to not follow grey matter on the perimeter of the lesion. A fix will appear in a future release. There is no work-around.
• make_average_volume: can exit with 'Cannot alloc memory' error if there are a lot of subjects. The work-around is to include the flag '-no-aseg', which will prevent creating the aseg.mgz.
• mri_nu_correct.mni incorrectly uses the -distance flag when --fwhm is passed. The work-around is to edit the script to correct this copy and paste error. A fix will appear in the next release.
• longitudinal processing (recon-all -long) is biased toward the timepoint selected as the base. There is no work-around. A new longitudinal processing scheme will appear in a future release.

• FS-FAST GUI bug

---

Stable v4.2.0

21 Jan 2009

Stable release version 4.2.0 is a minor release containing bug fixes (that change some results in the default analysis stream) and new features.

What's New

• Bug fixes:
  • aseg.stats/wmparc.stats bug fix: structure volumes, such as hippocampus, were reported as being about 2% smaller in wmparc.stats compared to the same structure in aseg.stats.
  • mri_ca_train bug fix: this message appeared: "GCAupdateNodeCovariance: could not find label".

  • mri_surfcluster bug fix: problem was in the way mri_surfcluster reported the area of a cluster when using a group surface such as fsaverage. fyi, see GroupAverageSurface for info on Freesurfer's group average surfaces.

  • mri_em_register bug fix: a bug prevented the full searching over rotation angles so the solution it found might not be optimal. this bug would manifest itself during mri_watershed, outputing the error message: "GLOBAL region of the brain empty !". this fix may change the results of the analysis stream in some regions by a very small amount.
  • tksurfer minor bug fixes:
    • The text in the colorbar when an overlay is loaded now reflects the Min and Max fields in the Overlay options.
    • A bug in tksurfer was responsible the labels drawn in tkmedit being outside the actual regions ( scattered )
    • A bug in tksurfer gave an "invalid command name" error when switching between overlays.
    • Due to a minor bug, the histogram in "Configure Window Overlay" dialog didn't include an entry for the values at 1.
    • Bug which stopped displaying annotation units after the 3000th was fixed.
• New features:
  • mri_glmfit-sim: utility to more easily perform cluster-based correction for multiple-comparisons
  • tksurfer : a button to write all 4 views ( medial, lateral, inferior and superior ) of the current surface in the window to a file.

Known issues

• aparcstats2table bug: if some of the subjects in a group have missing values of a certain label, aparcstats2table fails to designate a 0 for them, leaving the number of columns for that subject short by one (or specifically, by the number of missing labels relative to the other subjects), thereby making the table nearly useless. there is no workaround. a fix will appear in the next release, or can be downloaded here. ( extract the file to $FREESURFER_HOME/bin/ directory )

• isxconcat-sess: will use --projfrac-avg 0 1 0.2 regardless of what is specified. This means it will average across the cortical ribbon along the surface normal. This has been fixed, and the fix will appear in the next release. The new default will be to use a projection fraction of 0.5. This new version is already in place in both the dev and stable versions in the Martinos Center.

• wmparc.stats - the BrainSeg, !BrainSegNVox and BrainSegNotVent values are incorrect. Use the values from aseg.stats

• mri_convert: a bug in using the -oc flag (which is used during make_average_subject) exists: during make_average_subject, the tal transform and the centering (-oc 0 0 0) were applied in consecutive transforms instead of being composed into one. The first translated it out of the FOV and the 2nd translated it back-in, exposing the cropped FOV (typically posterior) from the first transform. A fix will appear in the next release. There is no work-around.

• tkmedit: seg-faults on some systems when scrolling and middle-button clicking at the same time over the tkmedit volume image. A fix will appear in the next release, or can be downloaded here.

• the estimated Total Intracranial Volume (eTIV) calculation has been found to be more accurate if the talairach.xfm transform is used instead of the talaiarch_with_skull.lta transform. in the next release, the new corrected value will appear in the aseg.stats file, and will be output by mri_segstats and mri_label_volume. there is no simple scale factor to correct existing measures.
• mri_average: -sqr option will overflow for short data-type images. A fix will appear in the next release.
• mris_make_surfaces: where wm lesions exist, even if the lesion is filled in wm.mgz volume, when mris_make_surfaces is run to create the final surfaces, because it uses brain.finalsurfs.mgz for its intensity info, and doesnt consider the fill data in wm.mgz, its possible for surfaces to not follow grey matter on the perimeter of the lesion. A fix will appear in a future release. There is no work-around.
• make_average_volume: can exit with 'Cannot alloc memory' error if there are a lot of subjects. The work-around is to include the flag '-no-aseg', which will prevent creating the aseg.mgz.
• recon-all -make all bug: the -finalsurfs stage incorrectly depends on brain.mgz instead of brain.finalsurfs.mgz, so if brain.finalsurfs.mgz is edited, recon-all -make all wont detect the changed file. a fixed recon-all.makefile will appear in the next release.

Upcoming features

• The next release will contain a new cortical atlas containing the insula label.

---

Stable v4.1.0

14 Oct 2008

Stable release version 4.1.0 is a minor release containing bug fixes (that change some results in the default analysis stream) and new features.

What's New

• Bug fixes:

  • Problems with the longitudinal analysis stream (affecting recon-all, mri_ca_label, mri_ca_register, and FSL's flirt) have been fixed. Refer to the help text output from recon-all -help for new usage instructions of the longitudinal stream.

  • mri_segstats and mris_wm_volume bugs fixed. Users of the ?h-cerebral-white-matter stats in the aseg.stats and wmparc.stats files should re-run recon-all -segstats -wmparc.

  • mris_make_surfaces problem causing a 'spike' feature in the pial surface in the insula has been fixed.
  • mri_mergelabel bug fixed: added -n to tail usage, as Fedora Core 5 needs -n to specify the number of lines to tail. mri_mergelabel is called by make_average_surface (which is called by make_average_subject).
• New features:
  • Qdec is now built against vtk v5.2.

  • Detection of images having a FOV > 256. If detected, recon-all exits. Users must add the -cw256 flag to recon-all, which will crop the image to no greater than 256 voxels.

Known issues

• aparcstats2table bug: if some of the subjects in a group have missing values of a certain label, aparcstats2table fails to designate a 0 for them, leaving the number of columns for that subject short by one (or specifically, by the number of missing labels relative to the other subjects), thereby making the table nearly useless. there is no workaround. a fix will appear in a future release.
• aseg.stats/wmparc.stats problem: structure volumes, such as hippocampus, are reported as being about 2% smaller in wmparc.stats compared to the same structure in aseg.stats. the difference is due to differing partial-voluming compensation methods. a fix will appear in the next release. the values in aseg.stats are the correct values.
• mri_ca_train bug: this message may appear: "GCAupdateNodeCovariance: could not find label". a fix for this bug will appear in the next release. there is no work-around.
• mri_surfcluster bug: problem in the way mri_surfcluster reports the area of a cluster when using a group surface such as fsaverage. The bottom line is that the area reported is about 25% too big. However, the cluster-wise p-values reported by mri_surfcluster based on simulations from mri_glmfit ARE NOT AFFECTED. Volume clustering is unaffected. A fix will appear in the next release.

• See GroupAverageSurface for info on Freesurfers group average surfaces.

• mri_em_register bug: a bug prevented the full searching over rotation angles so the solution it found might not be optimal. A fix will appear in the next release.

---

Stable v4.0.5

15 June 2008

Stable release version 4.0.5 is a new feature and bug fix release.

What's New

• New features:

  • Now supports Gifti v1.0 surface format read and write (.gii file extension) for exchanging surface files with Caret, AFNI, BrainVisa, BrainVoyager and Vistrails.

  • Updated Qdec v1.1: has aseg/aparc stats tables import; miscellaneous user features.
  • New volume and surface viewing app: freeview (beta). Note: freeview is not yet available for the Mac.
• Bug fixes:

  • Qdec now uses VTK v5.0 instead of v5.1, to fix the problem with message: "TclTk error: invalid command name "vtkTemp349"

Known issues

• Same issues as v4.0.4
• Bug in mri_segstats: some input data exposes a problem which results in a huge (GB) recon-all.log file, and incorrect results for ?h-cerebral-white-matter in the wmparc.stats file. this bug will be fixed in the next release.

• It is possible for a 'spike' to appear in the pial surface near the insula, like this image, in some subject data. A fix for mris_make_surfaces will appear in the next release.

• White matter parcellation (mris_wm_volume) bug:
  • isthmus cingulate segmentation appears between thalamus and pallidum.
  • parahippocampal segmentation appears next to ventralDC.
  • white matter extending into cortex (affects 4.0.3 and 4.0.4).
  • ASeg.stats: white matter volume way too low.
  • A fix for mris_wm_volume will appear in the next release.
• A bug in the longitudinal analysis stream has been found. A fix will appear in the next release.

---

Stable v4.0.4

6 May 2008

Stable release version 4.0.4 is a bug fix release.

What's New

• Bug fixes:
  • The tksurfer bug, where the text on the color-scalebar, and the caption text, may not appear, or appear dim, has been fixed.
  • The tksurfer bug, where it can exit with an error 'alloc: invalid block' when an overlay is loaded, has been fixed.
  • The recon-all -localGI bugs have been fixed.

Known issues

• mris_make_surfaces can let the pial surface extend inward toward the insular cortex in cases where the extreme capsule is so thin that no voxels are labeled as white matter (and no significant gm/wm contrast boundary exists). There is no work-around.

• The qdec app produces a 'Bad Match' error when started on some Linux platforms. The work-around is to add this to your /etc/X11/xorg.conf file:

Section "Extensions"
   Option "Composite" "false"
EndSection

• Input data having a FOV greater than 256 will fail in the -autorecon2 stage during mri_fill, because the wm.mgz file is 256³ while the norm.mgz and aseg.mgz files are FOV³ (where FOV > 256). The work-around is to use the --cropsize256 flag with recon-all with -autorecon1.

• The surface files ?h.volume are not produced by recon-all yet, but are mentioned in mris_preproc. Production of the ?h.volume files by recon-all (calculating ?h.area * ?h.thickness) will be available in a future release.

---

Stable v4.0.3

31 March 2008

Stable release version 4.0.3 is a bug fix and new feature release.

What's New

• Bug fixes:
  • The -log option of mris_anatomical_stats was fixed.

  • In qdec, the problem where a popup window appeared containing the message 'FsgdfPlot_IsWindowShowing' has been fixed.

  • The problem where make_average_subject failed with the message 'Cannot allocate memory' has been fixed.
  • The mri_ca_register and mri_ca_label segfault bug on the Mac OS has been fixed.
  • In tkmedit, the problem where the cursor goes to the wrong location when tal coords are entered has been fixed.
  • mris_preproc fixed: can now transform multiple volumes from the same subject.
• New features:
  • Updated qdec, now allows saving/loading a project data file.

  • A new option, -lGI, will calculate a local gyrification index. See localGI.

  • New Brodmann area labels for the 'fsaverage' subject. See BA Maps.

  • talairach_with_skull.lta generation is now robust in presence of increased background noise levels.

Known issues

• mris_make_surfaces can let the pial surface extend inward toward the insular cortex in cases where the extreme capsule is so thin that no voxels are labeled as white matter (and no significant gm/wm contrast boundary exists). There is no work-around.

• The qdec app produces a 'Bad Match' error when started on some Linux platforms. The work-around is to add this to your /etc/X11/xorg.conf file:

Section "Extensions"
   Option "Composite" "false"
EndSection

• Input data having a FOV greater than 256 will fail in the -autorecon2 stage during mri_fill, because the wm.mgz file is 256³ while the norm.mgz and aseg.mgz files are FOV³ (where FOV > 256). The work-around is to use the --cropsize256 flag with recon-all with -autorecon1.

• The surface files ?h.volume are not produced by recon-all yet, but are mentioned in mris_preproc. Production of the ?h.volume files by recon-all (calculating ?h.area * ?h.thickness) will be available in a future release.
• In tksurfer, the text on the color-scalebar, and the caption text, may not appear, or appear dim. A fix for this will appear in the next release. There is no work-around.
• The -localGI feature is broken, in particular, the mris_fill utility. The effect is a shifted filled.mgz relative to the true surfaces. A fix will appear in the next release.
• tksurfer can exit with an error 'alloc: invalid block' when an overlay is loaded. A fix will appear in the next release.

---

Stable v4.0.2

16 December 2007

Stable release version 4.0.2 is a bug fix and new feature release.

What's New

• Bug fixes:
  • Fixed mris_flatten bug where the map did not unfold properly.

  • Fixed unpacksdcmdir (mri_convert) issue, string: "syngo MR B15".

  • Fixed bug where mri_fill might fail with the message "could not find any points where lh and rh wm are nbrs".

  • mris_thickness_diff and mris_surface_stats now properly handle .mgh file io.
• New features:
  • Improved -tal-check option, for checking the Talairach alignment.
  • The FSL package is no longer included in the distribution (except for the binary 'flirt', found in freesurfer/bin).

Known issues

• mris_make_surfaces can let the pial surface extend inward toward the insular cortex in cases where the extreme capsule is so thin that no voxels are labeled as white matter (and no significant gm/wm contrast boundary exists). There is no work-around.

• The qdec app produces a 'Bad Match' error when started on some Linux platforms. The work-around is to setenv XLIB_SKIP_ARGB_VISUALS 1 if using tcsh, or export XLIB_SKIP_ARGB_VISUALS=1 if using bash.

• Input data having a FOV greater than 256 will fail in the -autorecon2 stage during mri_fill, because the wm.mgz file is 256³ while the norm.mgz and aseg.mgz files are FOV³ (where FOV > 256). The work-around is to use the --cropsize256 flag with recon-all with -autorecon1.

• The surface files ?h.volume are not produced by recon-all yet, but are mentioned in mris_preproc. Production of the ?h.volume files by recon-all (calculating ?h.area * ?h.thickness) will be available in a future release.
• The -log option of mris_anatomical_stats does not work. A fix will appear in the next release.

• In qdec, a popup window containing the message 'FsgdfPlot_IsWindowShowing' may appear. To fix this, copy this script to your $FREESURFER_HOME/lib/tcl directory: fsgdfPlot.tcl. This fix will appear in the next release.

• make_average_subject can fail with the message 'Cannot allocate memory'. A fix will appear in the next release.
• mri_ca_register and mri_ca_label can segfault on the Mac OS. A fix will appear in the next release.
• In tkmedit, when tal coords are entered, the cursor goes to the wrong location. A fix will appear in the next release.
• mris_preproc, when transforming multiple volumes from the same subject, each file converted overwrites the previous. A fix will appear in the next release.

---

Stable v4.0.1

18 September 2007

Stable release version 4.0.1 is a bug fix release.

What's New

• Bug fixes:
  • Fixed seg-faulting mri_convert in unpacksdcmdir.
  • Fixed tkmedit problem whereby volume display after loading talairach.xfm was incorrect.
  • Fixed seg-faulting mris_annot_to_segmentation.
  • Fixed tksurfer problem whereby curvature and overlay displays could not be turned off.
  • Fixed qdec problem where the display could be hidden on small screens (now has scroll bars).

  • Updated spatialsmooth-sess such that it no longer uses FSL's 'ip' binary.

Known issues

• When making flatmaps with mris_flatten, sometimes the map never unfolds properly (where there can be a section where it folds over itself). The fix for this bug will appear in a future release.

• The qdec app produces a 'Bad Match' error when started on some Linux platforms. The work-around is to setenv XLIB_SKIP_ARGB_VISUALS 1 if using tcsh, or export XLIB_SKIP_ARGB_VISUALS=1 if using bash.

• Input data having a FOV greater than 256 will fail in the -autorecon2 stage during mri_fill, because the wm.mgz file is 256³ while the norm.mgz and aseg.mgz files are FOV³ (where FOV > 256). The work-around is to use the --cropsize256 flag with recon-all with -autorecon1.

• The surface files ?h.volume are not produced by recon-all yet, but are mentioned in mris_preproc. Production of the ?h.volume files by recon-all (calculating ?h.area * ?h.thickness) will be available in a future release.
• mris_make_surfaces will let a surface in callosal regions extend across midline in some cases. This bug will be fixed in the next release. There is no work-around.

• unpacksdcmdir (mri_convert) issue: "ERROR: incorrectly formatted version string syngo MR B15". A fix for this problem will appear in the next release.

• The -tal-check option for checking the Talairach alignment step does not always catch failures. An improved checker will appear in the next release.

• mri_fill might fail with the message "could not find any points where lh and rh wm are nbrs". A fix will appear in the next release.

---

Stable v4.0.0

21 August 2007

Stable release version 4.0.0 is a major release with new features and bug fixes.

What's New

• New features:
  • The corpus callosum is now segmented into five separate labels in the subcortical segmentation volume 'aseg.mgz'
  • White matter is now parcellated for gyral regions
  • mris_wm_volume - calculates the total white matter volume
  • Improved talairach alignment, with automatic failure detection
  • Improved atlas-based skull-strip
  • Longitudinal analysis (recon-all -long, see help section in recon-all --help)

  • Qdec - group analysis GUI tool

  • DTI recon stream: see dt_recon --help, also new: dmri_poistats (see --help) Linux Centos only
  • FSFast:
    • Full NIFTI compliance
    • mkanalysis GUI -- set up analysis AND contrasts
    • selxavg3-sess -- now replaces stxgrinder, implicit intensity normalization, better whitening
    • tksurfer-sess -- replaces paint-sess and surf-sess
    • tkmedit-sess -- accepts multiple contrasts
    • isxconcat-sess + mri_glmfit -- intersubject averaging (RFx, FFx, weighted RFx), replaces func2sph-sess, func2tal-sess, and isxavg-??-sess, GRF
    • No longer needed: stxgrinder-sess, paint-sess, func2sph, fun2tal, isxavg-??, mkcontrast, inorm-sess
  • Average subject subcortical segmentation
  • Updated numerics libraries (NRC and GSL replaced by VXL), paving the way for an open-source release.
• Bug fixes:
  • Corrected midline parcellation (during cortical parcellation, areas of the anterior cingulate cortex (rostral anterior cingulate in the parcellation atlas), were mislabeled as corpus callosum or "unlabeled subcortical")
  • Corrected aseg: removal of grey matter around callosum
  • Corrected average subject surface area
  • Corrected problem with average subject thickness generation (mismatched number of vertices)
  • Corrected problem with mris_anatomical_stats, where incorrect results were reported for the measures 'average integrated rectified mean curvature', 'average integrated rectified Gaussian curvature', 'folding index', and 'intrinsic curvature index'.

Known issues

• When using unpacksdcmdir, mri_convert will fail with a segmentation fault. A fix will appear in the next release.
• In tkmedit, the displayed volume after loading the talairach.xfm file is not correct. The work-around is to either load the talairach.lta file, or use tkregister2 instead. This will be fixed in the next release.
• mris_annot_to_segmentation may segfault. A fixed version will appear in the next release.
• In tksurfer, if you have both curvature and an overlay loaded, and you turn both off, the overlay stays on. This will be fixed in the next release.
• When making flatmaps with mris_flatten, sometimes the map never unfolds properly (where there can be a section where it folds over itself). The fix for this bug will appear in a future release.
• In qdec, the Analyze button can be hidden on small screens. This will be fixed in the next release.
• The qdec app produces a 'Bad Match' error when started on some Linux platforms. There is no work-around at this time.

• Input data having a FOV greater than 256 will fail in the -autorecon2 stage during mri_fill, because the wm.mgz file is 256³ while the norm.mgz and aseg.mgz files are FOV³ (where FOV > 256). The work-around is to use the --cropsize256 flag with recon-all with -autorecon1.

• The surface files ?h.volume are not produced by recon-all yet, but are mentioned in mris_preproc. Production of the ?h.volume files by recon-all (calculating ?h.area * ?h.thickness) will be available in a future release.

---

Stable v3.0.5

7 Feb 2007

Stable release version 3.0.5 is a bug-fix release.

What's New

• Bug fixes:

  • Fixed bug in NIFTI reader in FreeSurfer (see details below in v3.0.4 Known Issues).

  • Fixed bug in 'unpacksdcmdir' and 'mri_convert' commands on the Intel Mac distribution.
  • Fixed bug in recon-all such that now xopts is correctly processed by mri_nu_correct.mni.

Known Issues

• Input data having a FOV greater than 256 will fail in the -autorecon2 stage during mri_fill, because the wm.mgz file is 256³ while the norm.mgz and aseg.mgz files are FOV³ (where FOV > 256). The work-around is to use the --cropsize option of mri_convert (see its --help for details) to convert the orig.mgz file to 256³.

• During cortical parcellation, areas of the anterior cingulate cortex (rostral anterior cingulate in the parcellation atlas), are mislabelled as corpus callosum or "unlabelled subcortical". A fix for this is in the works, possibly for the next release.
• When making flatmaps with mris_flatten, sometimes the map never unfolds properly (where there can be a section where it folds over itself). The fix for this bug will appear in a future release.
• On the Mac, make_average_subject (and make_average_volume and make_average_surface) will fail if more than 252 subjects are specified. The work-around is to type 'limit descriptors unlimited' before running these scripts. A fix will appear in the next release.
• mris_anatomical_stats, as executed by recon-all, reports incorrect results for the measures 'average integrated rectified mean curvature', 'average integrated rectified Gaussian curvature', 'folding index', and 'intrinsic curvature index'. The other reported measures ('total white matter volume', 'total surface area', 'total gray matter volume' and 'average cortical thickness') are correct as reported by recon-all's execution of mris_anatomical_stats. There is no work-around. A fix will appear in the next release.

---

Stable v3.0.4

25 Nov 2006

Stable release version 3.0.4 is a bug-fix release.

What's New

• Bug fixes:
  • Fixed problem with left hemi of fsaverage subject (see v3.0.3 Known Issues, below, for details).
  • Fixed bug whereby the pial surface can cross the midline in the cingulate/callosum area to the other hemisphere. The change [attachment:bert-medial-wall-diffs.tiff affects this area].
  • Fixed 'glutInit' bug occuring on Fedora Core and Debian Linux installations using tksurfer.
  • Fixed bug in mri_normalize which core dumped for low SNR input volumes.
  • Fixed bug in tksurfer with "Don't cross functional threshold" parameter.
  • Fixed problem with public declaration of Tcl/Tk/Tix/BLT environment variables which caused problems with FSL and 3D Slicer.
• New features:

  • Label file named <subjid>/label/?h.cortex.label is a single label for all of cortex.

Known Issues

• Input data having a FOV greater than 256 will fail in the -autorecon2 stage during mri_fill, because the wm.mgz file is 256³ while the norm.mgz and aseg.mgz files are FOV³ (where FOV > 256). The work-around is to use the --cropsize option of mri_convert (see its --help for details) to convert the orig.mgz file to 256³.

• The 'unpacksdcmdir' command, and 'mri_convert' of some DICOM formatted data, will fail on the Intel Mac distribution v3.0.4, due to a byte-endian issue. The v3.0.4b distribution has a fix for these problems.
• recon-all does not process xopts for mri_nu_correct.mni. The fix will appear in the next release.

• There is a small bug in the NIFTI reader in FreeSurfer, which causes the sform/qfrom matrix to be slightly off for oblique volumes that are non-isotropic (the more oblique and the more non-isotropic, the worse). Part of the error involves the matrix being slightly non-orthogonal. The fix for this bug will appear in the next release. There is NO effect of this bug if any of the following apply:

  1. You use mri_convert to convert from DICOM (or any other non-NIFTI format) to NIFTI
  2. Your volumes are not oblique
  3. Your volumes are isotropic
• During cortical parcellation, areas of the anterior cingulate cortex (rostral anterior cingulate in the parcellation atlas), are mislabelled as corpus callosum or "unlabelled subcortical". A fix for this is in the works, possibly for the next release.
• When making flatmaps with mris_flatten, sometimes the map never unfolds properly (where there can be a section where it folds over itself). The fix for this bug will appear in a future release.

---

Stable v3.0.3

24 May 2006

Stable release version 3.0.3 is a bug-fix and feature-addition release.

What's New

• Bug fixes:
  • Fixed a bug in tksurfer in which if an overlay and the curvature was loaded, the curvature would draw under the overlay even if the curvature display was turned off.
  • Modified the way NIFTI1 files are stored (these are the single .nii and .nii.gz files). The change is somewhat esoteric: the vox_offset is now 352 instead of 348. 352 is the NIFTI spec, so we were technically out of compliance before, although AFNI, FSL, and SPM did not seem to have a problem with 348.
  • Fixed bug in make_average_surface and make_average_volume scripts which limited the number of subjects that could be specified on the command line.
• New features:
  • asegstats2table and aparcstats2table - scripts to generate text/ascii tables of freesurfer stats data (eg, aseg.stats). This can then be easily imported into a stats program.

Known Issues

• In tksurfer, if using the -fthresh option, specify the number as a float (-fthresh 2.0) not an integer (don't use -fthresh 2).
• There is a problem with the left hemi of the fsaverage subject, specifically lh.sphere.reg. This will cause some inaccuracies in the visualization of maps on the left hemi of fsaverage and other average subjects (but the statistics themselves should be ok). The corrected lh.sphere.reg for fsaverage will be included in the next release. Or, if you have an average subject that you want to correct, you do not have to rebuild it. Rather, you can simply:

cd youraveragesubject/surf
cp lh.sphere.reg lh.sphere.reg.old # backup
cp rh.sphere.reg lh.sphere.reg # copy rh.sphere to lh.sphere

• There is a bug whereby the pial surface can cross the midline in the cingulate/callosum area to the other hemisphere. There is no workaround. The bug will be fixed in the next release. This change will [attachment:bert-medial-wall-diffs.tiff affect this area].
• tksurfer will fail on Fedora Core and Debian Linux installations which use freeglut, as 'glutInit' is not executed. This bug will be fixed in the next release.
• mri_normalize may core dump for low SNR input volumes. This bug will be fixed in the next release.
• In tksurfer, fills that use the "Don't cross functional threshold" parameter do not correctly fill negative functional areas. This bug will be fixed in the next release (v3.0.4).
• Tcl/Tk/Tix/BLT environment variables declared in freesurfer setup file causes problems with other apps, like FSL and 3D Slicer. This problem will be fixed in the next release. The workaround is to not call the Freesurfer setup script when using the other apps.

---

Stable v3.0.2

18 April 2006

Stable release version 3.0.2 is a bug-fix and feature-addition release.

What's New

• Bug fixes:
  • The --all-info or --version command-line corrupting bug has been fixed.
  • The mris_fix_topology mrisDefectAnalysis bug has been fixed.
  • The bug where control points were being truncated instead of rounded by mri_normalize, which could offset them by 1 voxel, has been fixed.
  • The bug where curvature was being computed incorrectly for RH9 binaries has been fixed (affected spherical morph, and curvature overlay).
  • eTIV calculations are now correct (they were ~25% too large, specifically, multiply pre v3.0.2 eTIV values by .744 to get the correct eTIV).
  • mri_label2vol - fixed bug for segmentation IDs over 999. Runs much faster now.
• New features:
  • bem directory is created in subjects dir (for the MEG people)
  • Better integration with FSL (compatible with FSL-nifti)
  • Uses robust estimates of class statistics in mri_normalize, mri_segment and mris_make_surfaces. Helps in low contrast (such as SPGR-based) images in which the gray matter is oversegmented.

Known Issues

• None at this time.

---

Stable v3.0.1

13 March 2006

Stable release version 3.0.1 is a bug-fix release.

What's New

• FIX_VERTEX_AREA no longer need be set in the environment outside of recon-all.
• In tkmedit, while loading a functional overlay or time course, the registration file name is now read properly.
• The FSFast utility mkanalysis-sess.new now accepts only sequence names matching "TR " (instead of all occurrences of TR).

Known Issues

• When running mri_surfcluster with average7 inside your csd file, make sure to specify --subject average7 --really-use-average7 or else it will underestimate the clustersize and cause the significances to be too low.

• mris_fix_topology may exit recon-all with the following message in the log: mrisDefectAnalysis : defect 123 overlap defect 119, where your defect numbers will be different. If this happens for one of your subjects, there is no work-around. A fix will appear in the next release.

• If the flags --all-info or --version are included at the beginning of a command-line, it will corrupt the subsequent command-line arguments. The work-around is to not include --all-info or --version with other command-line arguments. A fix will appear in the next release.

• On the rh9 distribution, sometimes the ?h.curv files produced are incorrect, displaying as nearly all red (instead of green and red, indicating gyri and sulci). This produces incorrect cortical parcellation data. There is no known work-around at this time. This problem does not appear when using the Centos or Mac distributions.
• tkmedit saves control points for intensity normalization in floating point RAS coordinates, but mri_normalize truncated these when they are read in with the -f command (the default behavior when the tmp/control.dat file exists). This can result in the control points being displaced by a voxel from the location that tkmedit displays them.
• mri_label_volume estimated TIV calculation are about 25% too large. This will be fixed in the next release.

• Be sure to read about the changes to recon-all, detailed below, and also read about the new fsaverage subject.

---

Stable v3.0.0

5 March 2006

Stable release version 3.0.0 is a major release containing many new features over the prior stable release (version 2.2).

Important! The recon-all script has been updated. Please read the ReconAllChanges page which describes these changes! Users of prior 'dev' releases should also take note of the changes.

• COR file format is no longer accepted by recon-all. mgz file's have replaced this old format.

• brainmask.mgz has replaced what was the brain volume. You should visualize and make all edits to brainmask.mgz and not to brain.mgz

• recon-all will now save all your changes by default. Any manual changes/edits to the talairach, setting of control points, edits to volumes etc will be saved and used by default and you must specify when you wish to ignore these changes.

• Data processed with an older release should make use of the -legacy flag in recon-all.

What's New

• Sub-cortical segmentation - Each MRI volume voxel is assigned one of 37 possible neuroanatomical labels, including left and and right caudate, putamen, pallidum, thalamus, lateral ventricles, hippocampus, and amygdala. The labeling allows for volumetric measurement of these structures. The segmentation information also improves the ability to create a filled white-matter volume.

• Automated filling of ventricles and basal ganglia - Improves the ability to create a filled white-matter volume (which is used in the surface creation process). Note: this requires the use of the new average subject fsaverage because the surface registration changes.

• Improved automation - The need to conduct manual edits to account for skull-stripping and cutting-plane problems is reduced. The topology fixer is much better at handling temporal lobe strands. Overall, greater than 90% of hemispheres processed by recon-all do not require manual intervention.

• Improved GUI tools - The GUI tools tkmedit and tksurfer incorporate many usability improvements requested by the user-base.

• New cortical parcellation - A neuroanatomical label is assigned to each location on the cortical surface model based on probabilistic information estimated from a manually labeled training set.

• Surface-based GLM analysis - Utilities for conducting group analysis studies with GLM.

• Multiple comparison correction for surface analysis - False discovery rate (FDR), permutation and Monte Carlo simulations.

• New average subject. $FREESURFER_HOME/subjects/fsaverage. Keeps info about average surface area.

• Average subject customization. Make your own average subject from your cohort.

• Extensive testing - The recon-all processing stream has been tested against several hundred subjects, different pulse-sequence types, and different MRI system platforms.

• Online tutorial of the Freesurfer work-flow - This release is compatible with the on-line tutorial. Tutorial data must be downloaded separately.

• Simplified distribution download and installation - The critical library and toolkit dependencies (MNI, FSL, GSL, Tcl/Tk/Tix/BLT, and Qt) are included to ease installation. These third-party open-sourced packages can be removed if they are already installed on the target system (although full Freesurfer functionality is no longer assured). For Mac OS X, a GUI installation package is provided.

• Compatible with all mouse pads

• Now Y2K compliant!

Known Issues

• Make sure to run "setenv FIX_VERTEX_AREA" if using any commands outside of recon-all, particularly the mri_glmfit and mri_surfcluster programs. To be safe, you can just add the line "setenv FIX_VERTEX_AREA" into FreeSurferEnv.csh or your .cshrc file. This will only affect those downloading the new version. This will automatically be done for those in the Martinos Center.

• In tkmedit, while loading a functional overlay or time course, the registration file name may not be read properly. A workaround is to load the overlay or time course from the command line (see tkmedit for details).

• The surface registration curvature target has changed with the new automatic filling of ventricles and basal ganglia. Do not mix newly reconstructed subjects with those that used the old target. For new data recons, display overlay maps on the new average subject fsaverage. For older data, use average7.

• There are some changes in the way that mri_surfcluster and mri_glmfit computes cluster sizes and how these interact with the average subject (and average7) see the notes fsaverage for more detail.