2017 Winter Project Week/SlicerShape
From NAMIC Wiki
Revision as of 15:34, 13 January 2017 by Jfishbaugh (talk | contribs)
Home < 2017 Winter Project Week < SlicerShape
Key Investigators
- Beatriz Paniagua (Kitware, Inc.)
- Johan Andruejol (Kitware, Inc.)
- JC Fillon-Robin (Kitware, Inc.)
- James Fishbaugh (NYU)
- Priscille de Dumast (UofM)
Project Description
Objective | Approach and Plan | Progress and Next Steps |
---|---|---|
|
|
|
Breakout session notes
- Slicer.org bridging with all Slicer customizations available
- We will need to change infrastructure of slicer.org to acknowledge all the different customized versions of the program. Bea will give this a go.
- Tutorials, forums, mailing lists should be included there.
- Architectural design decisions
- Improving support for vtkMRMLModelNode.
- Visualization - Color maps, Surface pre-processing algorithms
- Support for 4D support - Sequences
- New vtkMRML*Node classes
- S-reps
- Improving support for vtkMRMLModelNode.
- Improving user experience
- Providing a lot of feedback to the user
- Quality of inputs and outputs
- Speed, computation time
- Obtain some sort of benchmarks of computational performance based on the OS requirements. kwsyslibrary can provide information of the program running. Another idea is to run a small subset of the batch processing and provide some information of how much did it take and how much the rest of the computation is going to take.
- The user's frustration comes from the unexpected time delay when you run a method (expecting 1 min, waiting 1 h) Slicer has progress bars (Andras Lasso) is saying progress bar have phases and we should use it. They have a preview for parameter tuning and then they run the long method (preview on small data)
- Running in separate process is good because you can kill the process anytime
- Providing a lot of feedback to the user
- Thinking about the specific needs of the user base of slicerSALT
- These methods take a lot of time to compute and usually involve a batch process
- Greg Sharp SlicerRT has a stand-alone tool that submits and logs jobs into Slicer
- Ron Kikinis' input is that Slicer is not really designed for not interactive things, but you can have non-interactive mode and run things
- Girder has ways to submit jobs and save outputs into it (Breakout session tomorrow)
- These methods take a lot of time to compute and usually involve a batch process
Shape Regression
A prototype extension for shape regression is working in Slicer (see screenshot at the top of the page). We learned about the sequences extension which can play back sequences of various data types, including surfaces. It already has a lot of the functionality we will need for visualizing 4D shape models, allowing for more rapid prototyping and development in general. Plotting features such as volume will also be straightforward.
Shape regression can take several minutes to several hours, so the user needs some useful feedback. One idea that the community seems to favor is a dashboard application which can launch and monitor the progress of jobs. For processes that can take hours, this seems to be a better idea than a progress bar.