======================================
InsightSNAP Release Notes
Version 1.9.x
======================================
-----------------
1. New Features
-----------------
1.1. New in Version 1.9
----------------------------------------------
This is an experimental/interim release, preparing us for the major 2.0
release. It is intended for users who want to preview new functionality; users
interested in stability may want to wait until the 2.0 release.
1.1.1. New Features and UI Improvements
- Support for multiple image layers. Users can now load gray and RGB images
as overlays on top of the main image layer. For example, one can display a
statistical map as an overlay over an anatomical image. As of version 1.9.0,
overlays must have the same dimensions as the main image. Support for overlay
management is a work in progress and new UI elements will be added in the
subsequent 1.9.x releases.
- An expanded menu bar. We have split the menu into File, Segmentation and Overlay
menus to provide easier and faster access to the ITK-SNAP features.
- Native file chooser. On Windows and MacOS, ITK-SNAP will use a native file
chooser instead of the FLTK built-in file chooser. On Mac OSX, the native
file chooser can be further enhanced by installing the DTI-TK Quick Look
plugin that supports NIfTI/Analyze image preview (www.nitrc.org/projects/dtitk)
- Automatic check for software update. Users can enable automatic update checking.
1.2. New in Version 1.8
----------------------------------------------
1.2.1. New Features and UI Improvements
- Support for reading floating point images of arbitrary range. SNAP still
represents gray images internally as signed short, but now it can load a
floating point image and remap its intensity range to signed shorts. When
displaying intensity values, it will map back to float.
- A new 'Adaptive' paintbrush. Under the paintbrush tool, it can be selected
using the 'Shape' drop down. This tool can speed up manual segmentation
quite a bit for some users. This brush has the shape of a rectagle. As you
click on a pixel in one of the slice views, the brush will fill a region
that includes the pixel you clicked and has more or less uniform intensity.
For example, in brain MRI, if you click in the ventricles near the caudate,
the brush will fill the ventricle but not the caudate. This is not as
powerful as running the level set segmentation, but it's very local and
great for quickly segmenting structures - or dealing with inhomogeneities.
The underlying algorithm is ITK's watershed segmentation. You can control
the tolerance of the adaptive brush ('granularity' input, lower values
produce smaller, more cohesive regions). The brush can be used in 2D or 3D.
This feature was inspired by a similar tool in ITKGrey, a tool from the
Vista Lab at Stanford that itself is a branch of an older version of
ITK-SNAP. Let us know if this feature works for you. Potentially, we may add
other algorithms in the future, including running the level set inside of
the brush.
- Support for image orientation. This is a major step towards NIFTI
compatibility (part of our R03 effort) and something many users should find
helpful. Formats such as NIFTI, DICOM, and a couple others encode the
orientation of the image axes in patient space, and even allow image axes to
not be parallel to the anatomical axes. SNAP now reads this information from
the image header and uses it to assign anatomical labels and compute
anatomical coordinates. One of the consequences of this change is that the
image IO wizard no longer requires specifying an orientation code (e.g.,
'RAI') when loading an image, since this information is read in the header.
- A new 'reorient image' dialog has been added, so that if the orientation
information in the header is wrong, you can change the orientation and save
the image. For now, the user can only specify reorientations that are
parallel to the anatomical axes.
- World cursor coordinates (under image information) are now displayed in
NIFTI / MNI coordinates as well as ITK coordinates. The difference is that
the NIFTI coordinates incorporate orientation and are in the (L->R, P->A,
I->S) coordinate frame. ITK coordinates are (x * spacing + origin), and
ignore orientation.
- 3D Meshes generated and rendered by SNAP are now represented in NIFTI world
coordinates. Previously, the coordinates were computed using the formula
x_mesh = x_voxel * spacing + origin
In version 1.8 and beyond, the mesh coordinates are computed as
x_mesh = nifti_sform_matrix * [x_voxel; 1]
This means that the meshes output by earlier versions of SNAP may be
translated and rotated relative to the meshes output by version 1.8. This will
not affect users who simply view meshes in SNAP; however users who export
meshes to other programs will be affected.
- Multisession cursor (similar to yoking in MRIcro) now uses these NIFTI
coordinates rather than ITK coordinates. This is a key feature because it
enables users to work with MRI scans acquired during the same session with
different orientations. For example, a coronal T1 scan and an oblique T2
scan can be loaded in two SNAP instances, and the cursor will be correctly
linked across the two.
CAVEAT: SNAP's cursor always falls on voxel centers. This means that the
multisession cursor correspondence is not exact, but rounded to the nearest
voxel. If in session A you move your cursor, the cursor in session B will
move to the voxel center closest to the physical position referenced by the
cursor in session A.
- A new multi-session zoom feature. Similar to the multi-session cursor, this
allows the zoom level to be maintained across multiple SNAP sessions. Useful
if you do a lot of zooming in and out when working with a pair of scans.
This is disabled by default and must be enabled in each SNAP session using
the checkbox under the 'Zoom/Pan Tool'.
- Changes to how zoom works, related to above. Now 'zoom views together' is on
by default, meaning that the zoom factor is the same in axial, coronal and
sagittal windows. Zoom level is specified in px/mm, where px is the number
of screen pixels (in other words, a metric equivalent of dots per inch).
Before, zoom was specified in percent, relative to an optimal zoom that
would best fit all three windows. With the new way you have more control
over the zoom. For example, if your image has 1mm voxels, you can have one
to one correspondence between screen pixels and voxels by setting the zoom
to 1 px/mm.
- Multisession 3D views. When the multisession cursor is selected, the 3D
views are also synchronized across sessions. This works even if the images
opened in the two SNAP sessions have different dimensions, orientation and
spacing. SNAP 3D window now uses NIFTI world coordinates, so as long as the two
images overlap in world space, so will the 3D views of the two images. This
feature is useful when comparing two segmentations of the same image.
- A new ruler display in slice windows. Can be disabled or modified on the
display options dialog.
- Much better tracking of changes to the segmentation image and better
promting to save changes before quitting or loading a new image. The title
bar display is also improved and uses an asterisk to indicate unsaved
changes.
- The command-line options have been updated. You can now load a grey image
without using any flags (e.g., itksnap image.nii) and there is a new '-rgb'
flag for loading an RGB image from command line. The upshot is that you can
now associate SNAP with image file extensions in the operating system and
double-click an image file to open it in SNAP.
- A new 'Tools' dialog on the label editor. This dialog is intended to provide
several tools for merging or modifying labels. The first tool is to combine
a pair of labels into one. Previously, this was possible using the 3D
scalpel tool, but that was not really an intuitive way to relabel images.
- As part of above, a new topological merge tool, developed by Nick Tustison,
Brian Avants and Marcelo Siqueira (I hope I did not forget anyone). Given
adjacent labels A and B, it will replace most voxels in B with the label A,
while preserving the topology of A. This tool is used to preserve topology
during manual segmentation. If A has correct topology and you want to add
some region to A, label this region with label B, and then grow A into B
with topology preservation. This is a work in progress, and feedback would
be welcome on this feature.
- Documented existing keyboard shortcuts and added some new ones. Available
shortcuts can be listed by selecting Help->Shortcuts.
1.2.2. Programmatic/Distribution Changes
- SNAP is now built against ITK 3.8, offering several improvements, especially
in how image orientation is handled.
- IPC communications (technology that allows multisession cursor and zoom) now
has some versioning built into it, so if you are running two versions of
SNAP, they will not clash.
- On LINUX, we now distribute a .tgz archive instead of a script installer.
Some people complained about the latter. We can also make .rpm and .deb
packages although these won't be posted for public download yet.
1.2.3. Bug Fixes
- Level set fix for ITK 3.8 fixes automatic segmentation's weird behavior
1.3. New in Version 1.6.0.1
---------------------------
1.3.1. Bug Fixes
- Major bug in release 1.6.0 involving disabled cursor movement in snake
segmentation mode has been resolved.
1.4. New in Version 1.6.0
-------------------------
1.4.1. New Features and UI Improvements
- You can now save a sequence of all axial, coronal or sagittal slices with
overlays as PNG files (File->Save->Screenshot Series).
- Automatic window and level computation based on the image histogram. The
window and level are set to the 1st and 99th percentiles of the intensity
histogram, respectively. This is much more robust to hypo and
hyper-intensity in medical imaging data. The feature is accessed in the
"Options->Image Contrast" menu (or hit Alt-I in the main window).
- Cursor synchronization across multiple SNAP sessions (similar to the Yoke
feature in MRIcro). The mechanism uses POSIX shared memory. Can be turned
off using the 'Synchronize Cursor' checkbox. Currently, only enabled in
manual segmentation mode; probably will enable in snake mode in the near
future.
--- NOTE FOR MacOS Users ---
MacOS doesn't allow you double-click the application icon to open a
new instance. To open multiple instances of ITK-SNAP, you need to launch
it from the command line.
----------------------------
- SNAP will prompt you before closing if there are unsaved changes.
- A new 'New->Segmentation Image' menu item will clear the current
segmentation.
- Support for RGB (color) images in SNAP. This is great for segmenting in DTI
data (manually, for the time being). RGB images can be loaded as the base
image or as an overlay over the gray. To create these RGB images, use the new
DTI-TK developed by Hui (Gary) Zhang, available from
http://www.picsl.upenn.edu/resources_dtitk.aspx
- Segmentations can be exported as VTK meshes (for example, for loading in
ParaView).
- Multilevel undo/redo functionality for all segmentation operations (polygon,
paintbrush, freehand, 3D segmentation, 3D cutplane). Undo memory is
preserved when loading new segmentation images.
- Freehand drawing support in polygon mode (hold and drag the mouse button).
This feature is especially useful for using SNAP on a tablet.
- Added keyboard shortcuts 'a','s','d' for the opacity slider
- Shortened/simplified some of the menu items
1.4.2. Bug Fixes
- Various bugs have been fixed :)
1.4.3. Distribution Changes
- SNAP website fully migrated to sourceforge.net
- Mac Universal binaries supporting Intel and PCC, Tiger and Leopard are now
available starting with 1.6.0
- Linux binaries will be available starting with 1.6.0
1.5. New in Version 1.4.1
-------------------------
1.5.1. New Features and UI Improvements
- Added paintbrush tool to the main toolbar. Paintbrush can be used to quickly
touch up segmentations. Left mouse button paints with selected label, right
button acts as an erasor
- Went through and added/edited tooltips in the program to be more accurate. It
should be easier to make sense of the program now
- Added a menu option for saving the level set image during active contour
evolution. This is an important feature because it allows users to save
segmentations before sub-voxel accuracy is lost. In particular, this can be
used in conjunction with ParaView to generate meshes from segmentations.
- You can now save and restore the camera settings in the 3D view within a
single SNAP session. This can be useful for generating screen shots of
different segmentation from the same viewpoint. Press 's' in the 3D window
to save the camera state and 'r' to restore it.
1.5.2. Bug Fixes
- MAJOR: fixed bug that was causing crashes on Win32 during polygon drawing
(thanks to Jeff Tsao for this bug fix!)
- Fixed problems with the getsnap.sh linux script
- Some menu items were enabled when they should not have been, now are
disabled.
- Rare bug where speed function very close to 1 was not being rounded
correctly and may have caused crashes on some systems
- Fixed problem where the screen was blank after loading preprocessed image
- Fixed crash when changing bubble radius and then going back to
preprocessing mode
1.5.3. Distribution Changes
- Interim SNAP releases are now hosted on SourceForge. ITK repository will only
be used to host major releases (like 1.6). This allows us to check stuff in
independently of the ITK code freezes. It also makes it easier to add new
developers.
- SNAP CMake files should automatically detect when SNAP is being built
outside of ITK's InsightApplications. This means you can build SNAP on it's
own and the download size is reduced
1.6. New in Version 1.4
-----------------------
1.6.1. New Features and User Interface Improvements
- New and improved label editor. You can easily switch between labels while in
the editor and the interface for adding new labels is more intuitive. You
can now delete labels.
- New and improved interface for intensity reparameterization. The histogram
display is more visible and you have more control over the number of bins in
the histogram and the scaling of the bars (linear or log).
- SNAP remembers all settings associated with loading an image. This means that
any image loaded previously can be reloaded without going throught the
wizard.
- We've added File->Load Previous menu to let you load images quickly
- SNAP can now read DICOM file series (experimental support) and it can read
and write VoxBo CUB image files.
- SNAP remembers more image-associated settings from session to session. For
example, it will remember the intensity reparameterization that you last
used. SNAP will also remember the orientation ("RAI" code) that was last
used to read each image.
- New Image Information window is available under the File menu. It displays
the size of the image and the current cursor position.
- A color map feature has been added in the automatic segmentation mode. The
color map lets you select different color schemes for displaying the
probability map / speed image.
- Small improvements to the active contour 2D example dialog have been made
- A progress monitor has been added for the 3D renderer in main SNAP window.
- New buttons allow taking of snapshots in each of the SNAP image windows
- The tutorial has been updated to reflect the new features.
1.6.2. Bug Fixes.
- SNAP should crash a lot less than before
- The Left-Right orientation should be correctly handled by SNAP. You still
have to supply the correct orientation ("RAI Code") when loading the image.
- The bug with the segmentation being shifted when using "Resample Region"
option has been fixed
- 3D window handles images with non-zero origin better
- Initialization bubbles have been fixed to be floating point
- Lots of other small bugs have been fixed!
1.6.3. Programmatic Enhancements
- SNAP and IRIS now share the sameset of OpenGL windows. This should prevent
crashes on some platforms.
1.6.4. Other
- SNAP available as a universal (Intel/PPC) binary for MacOS at itksnap.org
1.7. New in Version 1.2
-----------------------
1.7.1. User Interface Improvements
- The ability to switch between 4-view mode and single view mode. Each of the
slice views and the 3D view can be expanded to occupy the entire SNAP window.
- A zoom thumbnail is now displayed when a slice view is zoomed in. The thumbnail
view can be used to pan the slice.
- User can specify whether he/she prefers to start in linked zoom mode or in
unlinked zoom mode.
- User can change the appearance of various display elements, including the
crosshairs, the region of interest selection box, the window background and
more.
- SNAP automatically determines the image orientation (RAI) when that information
is available in the image file
- SNAP remembers the last ROI used for each image.
1.7.2. Programmatic Improvements
- The level set segmentation pipeline has been rewritten, taking advantage of
the stop and go functionality of ITK finite difference filters. This means
fewer unexplained crashes and simpler code.
- A state-machine has been added to the user interface logic code. This
machine automatically activates and deactivates UI widgets based on a set of
flags. Rules such as Flag A => Flag B can be added to the state machine.
1.7.3 Bug Fixes
- Slice views update correctly when the SNAP window is resized
- Accepting a polygon now works for high resolution images.
- Fixed a crash on some systems when running edge-based snake segmentation with an
advection term.
2. Known Issues
-----------------
-
3. Wish List
-----------------
- The ability to perform 2D level set segmentation in each slice view.
