Hi, at long last we have an official release of 7.3. See the release notes can be found here: https://surfer.nmr.mgh.harvard.edu/fswiki/ReleaseNotes and also copied below. Thanks to all who tested the beta version.
best
doug

Bug fix when converting diffusion tensor data with a negative determinant from DICOM. There was a bug in the old code that made the bvecs incorrect in this case. The bvecs will now be correct regardless of whether you use the old code or the new -dcm2niix option (see below)
Bug fix: ability to fill in underlabeled ventricles. This is not done by default yet. To implement, add "-vent-fix -1 7000 1" to the mri_ca_label command line. This can be incorporated into recon-all by an expert options file with the line "mri_ca_label -vent-fix -1 7000 1".
Bug fix: ability to force white matter (WM) between putamen and cortex. This is not done by default yet. To implement, add "-insert-wm-bet-putctx 1" to the mri_ca_label command line. This can be incorporated into recon-all by an expert options file with the line "mri_ca_label -insert-wm-bet-putctx 1". If you want to use both the ventricle fix and the WM insertion, then create a line "mri_ca_label -vent-fix -1 7000 1 -insert-wm-bet-putctx 1" to the expert options files
Bug fix: improve surface placement in inferior frontal area near putamen. There are some cases where the white surface is not placed well in these areas because of the close proximity of putamen to cortex; it is probably the case that the surface is not well-defined in this region anyway, but the placement can be improved by creating an expert options file with the line "WhitePreAparc --rip-bg-no-annot"
Bug fix/New: ability to automatically edit the filled.mgz to remove voxels from amygdala and parts of hippocampus. These voxels can often cause topological defects that are not fixed properly. This is not done by default yet. To implement, create an expert options file with the line "mri_edit_wm_with_aseg -fix-scm-ha 1".
Bug fix: cross hemisphere analysis (xhemi). In previous versions, the surface registration (surfreg) would use the annotation during the registration procedure which caused some misalignment. Now surfreg will require that you specify either --annot (old behavior) or --no-annot (new recommended). The errors are small, so, if you have an ongoing analysis, you can stick with the old behavior.
Global Expert options file. If you have a file called global-expert-options.txt in your $SUBJECTS_DIR, recon-all will automatically read in this file as an expert options file
Defect labels will now only encompass actual cortical areas. Before, they would also include vertices on the medial wall, which was not very useful
New: Conversion from DICOM can now be done with dcm2niix as a backend using -dcm2niix with either dcmunpack or mri_convert. This should give the same result as when using the old code except it will be much faster and can now handle all cases that dcm2niix can handle, including enhanced dicom. At some point, this will become the default method. Thanks to Chris Rorden and Yujing Huang for getting this working.
New: Gradient unwarping. There is a new program called mri_gradunwarp; you still have to supply the gradient unwarping spherical harmonics tables.
New: Minimally Invasive Defacing tool (MiDeFace).
New: segment_subfields beta version new software to run subfield segmentation (hippocampus, amygdala, thalamic). The old software required matlab. The new tool is written in python but will not give exactly the same result but will be extremely close. Does not work with T2 yet.
New: SynthStrip deep learning tool to strip skulls
New: ScLimbic subcortical limbic deep learning segmentation tool
New: HypothalamicSubunits deep learning segmentation of hypothalamic subunits
New: SynthSeg deep learning contrast-independent whole brain segmentation (including cortical parcellation)
New: Segmentation-based (SB) cost function for multimodal registration in mri_robust_register
New: better version control. You can create a file called fs-allowed-versions.txt in $SUBJECTS_DIR with a list of allowable versions for your analysis. If this file exists, then both the current version of FS and the version listed in $subject/scripts/build-stamp.txt must be in the list for recon-all to proceed. This helps to prevent using different/incompatible versions of FS within the same analysis while allowing the mixture of compatible versions.
New: interface for X11 virtual frame buffer (fsxvfb). When you run freeview from fsxvfb, no X server is needed. This is good for taking a bunch of pictures when you don't want freeview popping up all the time. Eg, fsxvfb freeview orig.mgz -ss mypic.png --quit