ReleaseNotes.txt

######################################################
ASHS (Automatic Segmentation of Hippocampal Subfields)
Release Notes
######################################################

=========================
Revision 110, Nov 2019
=========================

New in this Revision:
---------------------

* Started to multimodal support to ASHS. In this revision, the option to perform 
  multimodal registration was added. In future revisions, multimodal label fusion
  and multimodal corrective learning.

* In some application (e.g. MTL segmentation), additional affine alignment is needed
  to accurately align the two modalities for each side separately in addition to 
  global whole image rigid registration.

=========================
Revision 100, July 2018
=========================

New in this Revision:
---------------------

* Added support for non-local-mean denoising and upsampling as preprocessing steps

* User can specify additional arguments for GNU parallel


=========================
Revision 90, May 2012
=========================

New in this Revision:
---------------------

* Fixed various bugs resulting from changes in revision 80

* Changed the way cross-valudation is performed in ashs_train. Now, for each
  cross-valudation experiment, we set up a "mock atlas" directory and run the
  main ASHS script to perform cross-validation. 


=========================
Revision 81, April 2012
=========================

New in this Revision:
---------------------

* Resolved some issues with clashes between PATH and other environment vars
  and the variables exported by ASHS. All variables exported by ASHS now
  start with the prefix ASHS_. The PATH is set in each ashs_*_qsub.sh script
  so that the user's global PATH set in the .bash-profile does not override
  the ASHS path.

=========================
Revision 80, April 2012
=========================

New in this Revision:
---------------------

* You no longer need SGE to run ashs_main. There are three ways to run ASHS. You
  can just run it in a single process, which is really slow. You can run a lot
  of ashs_main scripts, each in a separate process, by using SGE. Or you can let
  ASHS launch sub-jobs using SGE. This was the behaviour in the older revisions.
  For the latter option, you will need to use the -q or -Q options.

* The training component (ashs_train) still requires SGE and uses it to launch 
  sub-jobs. 


=========================
Revision 76, April 2012
=========================

New in this Revision:
---------------------

* ASHS includes a training component. Given a set of images with corresponding 
  segmentations, you can use it to create your own atlas set. See ashs_train.sh

* ASHS no longer requires manual 'slice markings'. Instead, if your protocol is
  limited to specific slices, ASHS can be informed of this with the help of a 
  heuristic rules file. The rule file can be used to specify that a certain
  label excludes another label, or that a certain label spans a specific range
  of slices relative to the other labels. The heuristics are specified when 
  building an atlas.

* ASHS is no longer tied to a specific segmentation protocol. Each atlas set can
  use its own segmentation protocol.

* Most ASHS parameters can now be set by user by specifying a config file.

* ASHS includes a bootstrap mode, where segmentations obtained after the 
  standard procedure are used to rerun the registrations between atlases
  and the target image. Currently, only one iteration of bootstrap is run.
  Bootstrap is not yet incorporated into the leave-one-out validation in
  ashs_train. 

* ASHS will now try both FLIRT and ANTS for T1-template rigid registration. It 
  will use the transformation that gives the best metric.