The purpose of this retreat is to review the progress made with the slicer 3 project and plan the work for the upcoming months. In particular, we are planning to have a release by the next AHM in January and we need to make sure we're covering the needed bases.
- 1 Logistics
- 2 Where are we now
- 3 Lessons learned
- 4 What are the priorities
Start at 10am going to mid afternoon.
- Steve Pieper
- Nicole Aucoin
- Alex Yarmarkovich
- Wendy Plesniak
- Jim Miller
- Bill Lorensen
- Sebastien Barre (Kitware)
- Will Schroeder
- Mathieu Malaterre
- Andy Cedilnik
Where are we now
Current state of the implementation
- Review of the slides used at the Core 1 retreat
- Node Types
- Undo/Redo functionality
- Interactions with Logic and Widgets
- KWWidget Classes
- SlicerGUI Classes
- VTK 3D Widgets (aka WWidgets)
- Inserting menu items before Exit in the File menu
- Avoid Slice widget (option menu) resize but allow people to read full name of currently selected node. Maybe with a tooltip of the full name of the widget
=>Exactly. Each time you select something in that menu, intercept the event and reset the balloon help string of the option menu button with the full name of the data (vtkKWWidget::SetBalloonHelpString)
- Getting access to the RenderWindowInteractor from a KWRenderWidget (protected?)
- Can the 3D Widgets be made 'hot' so that they highlight on mouse-over events
=> Check with Will
- Good strategy for assessing the resolution of a machine's display, and setting Slicer3 layout parameters in a compatible way
- Reformatting the main window to display various view options (lightbox, 3D-only, slice-only, 4-up, tabbed scene view, etc.)
- Typing Alt-F-X too fast can cause the UI to hang (Windows)
=> Type it slower ;) Slicer is great, why do you want to quit so fast ? Seriously, could not reproduce. My Slicer crashes and burn (leaks + crash) anyway, so I won't exit cleanly in any cases right now
sp> This happens before any crashes :) But I can show you the behavior when we are together...
- KWWidgets method to "eventually render" like an 'after idle render' in tcl/tk
Slicer widgets and KWWidgets we need to design and build
- Window / Level / Threshold Editor (close to being done)
- Better Slice Controller widget; one that manages FG/BG layers but also allows other layers to be defined, with visibility and compositing attributes
- MRML Scene browser (tree widget)
- MRML properties browser/editor
- 3D View navigation widget and Slice Window zoom widget (toggle on mouseover Slice Windows)
- A filebrowser without "motif" look -- which uses the registry to figure out favorite directories, maybe which contains previews of files
- Workflow wizard?
- Ibrowser (but maybe this is a scriptable widget)
- Color palette browser/editor widget
Other GUI Issues
- Progress on tracing
- Cross platform look and feel (font differences)
- (native mac port worth the effort?) See Cross-Plaform Issues: Mac OS X with default Tcl/Tk Aqua
- Tile extension likely to be available?
- Example applications for Core 1 to use as template
- Build/Pack Environment
- Autodetection by Slicer
- Building the GUIs in KWWidgets
Report on feedback from Cores 1 and 3
- Efficient coding for fleshing out the GUI (keyboard shortcuts, widget geometry management, clean cross-platform look&feel).
- How to make all this code clean and maintainable
- How to be sure there are enough shoulders on the oars to get the work done
- Building and Launchers
- Path and LD_LIBRARY_PATH management for resources and dependent shared libraries (for Base and for Modules)
- Template CMake framework for Module writers
- More efficient builds?
- Example ITK code using the Execution Module framework
- Bundling Modules for distribution
- "Check for Updates" feature
- Use hits on the update website to track usage patterns
- Standing agenda item to discuss Slicer3 projects during the engineering tcons
- User-feedback sessions on GUI and Functionality
- Randy Gollub
- BWH PNL Users
- Dartmouth Users
What are the priorities
- Core 1 priorities
- They like the execution model
- Core 3 priorities
- How can we focus time at the Project Week to move things ahead?
- Can we give Core 1 examples of using the Execution Model? What help do they need?
Best of Breed
This is Ron's list of features where we are "best of breed" or close in Slicer 2.x.
It is important that we don't loose ground when moving to Slicer3.
- DTI: We have a clear advantage compared to almost everybody.
- Interactive Editor: must-have basic component
- IGT: Best integrated package: tracker, registration, segmentation and visualization
- fMRI: one-stop shopping for combination of analysis and visualization/multi-modality
- Morphometric Analysis: EM is close to the quality achieved by Freesurfer on data for which it was calibrated, but is easier to calibrate and can handle pathology (e.g. WML)
One of the biggest UI challenges today: many of the interface modules were designed by engineers and not sufficiently organized for the needs of end-users.
Another big challenge is that researchers prefer not to learn a "big environment" when they feel they can get their papers published with a small set of tools. Researchers also typically do not write their code for maintainability (or even when they do, they don't actually maintain it).
- Execution Model discussion
- Decision to flesh out examples of the command line modules in preparation for project week
- Decision to work on automatic generation of GUI interface during project week
- Move to ITK 2.8 Monday
- Alex and Steve to rework the GAD module for Bill and Jim to use as template for CLModule
- The next budget year GE and Kitware will allocate resources for Slicer3 development to meet development goals for January NA-MIC AHM.