Release Name: rev80

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

=========================
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.