Minutes of Follow Up Meeting
From IGSTK
IGSTK Year One Review Project Plans
From Meeting Thursday night 10 Nov 2005 and Friday 11 Nov 2005 Compiled by Kevin Cleary on 12 November 2005
Attendees
- Kevin Cleary, ISIS
- Patrick Cheng, ISIS
- Luis Ibanez, Kitware
- Rick Avila, Kitware
- Andinet E., Kitware
- Stephen Alyward, UNC
- Julien Jommier, UNC
- David Gobbi, Atamai
- Kevin Gary, Arizona State University
Overview by Kevin Cleary
- The project ends on August 31, which gives us about eight months to finish
- We mapped out a schedule like
- November: tag current release (release 2) and application
- Tcon Nov 17
- Thanksgiving Nov 24 (no tcon)
- RSNA Dec 1 (Kevin, Rick there, maybe can still have tcon)
- December: Tcons 8, 15, 22
- January: Iteration 7: improved 4ViewTrackingWithCT application
- February: release toolkit?? (to be discussed)
- SPIE meeting and open source workshop
- March: gather user feedback
- April: Iteration 8: Ultrasound guided RF ablation of liver lesions
- May: finish book
- June: Iteration 9: Guidewire tracking application
- July: finish reports
- August: current contract end: final report due
- November: tag current release (release 2) and application
Action plan
- Focus now on applications and use them to drive project
- Refactor components and add new functionality as needed
- There are no additional components to be developed
- Implement consistent patterns and include these checks in code review
- Work on book (Kevin / Patrick / all)
- Build user community
- Flush out remaining state machine issues (Luis / all)
- What about requirements? (may not be a critical issue)
Individual responsibilities and components:
• Kevin Cleary: project management, book, requirements, funding • Rick Avila: project management, visualization (view component) • Luis Ibanez: architect, state machines • David Gobbi: tracking, communications • Patrick Cheng: applications, book • Julien Jomier: spatial objects, website, logos • Stephen Aylward: consultant • Andinet E.: reader
State Machine Discussion
- Implement event passing and work into example application (Luis/Patrick)
- Luis also suggested we start with Hello IGSTK example
- Kevin Gary noted that all state machine should go to the AttemptTo() pattern
- It was agreed we would do this
- David stated if all processing can be done with a set() method we do not need this
- Luis stated that if processing is needed for checking and we can be sure that are not other events coming in we can use this pattern (???)
- Kevin Gary suggested that this be added to the code review checklist
- Need to list the state machines that can be converted: action item for tcon
- Kevin Gary raised these issues which were not settled
- Centralized dispatch
- Hierarchical state machines
Components Discussion
- Logging is pretty much done and has been adopted by ITK. There are some issues with logging as on the ITK mailing list. Andinet will be the IGSTK representative for this.
- Pulse generator. Luis is currently responsible, but could use some help. Real-time clock is one class and pulse generator is another class. Want to decouple it from FLTK. Looking for a volunteer to help here.
- Spatial object (Julien)
- Need to fix bugs in current spatial object
- We will support the notion of a group but it will be generic across spatial spatial objects
- Discussed having anatomy specific spatial objects for the key anatomies in our demo application (liver, lung, vasculature, etc)
- Need to add a few specific spatial objects (tube etc. ???)
- There is an igstkGroupObject and an igstkTubeGroupObject
- View (Rick)
- View position not initialized yet and annotation is not supported yet
- Interaction with user needs to be improved (window/level, pan, zoom) – in which class does this functionality get added?
- Rick could add volume rendering – was considered desirable but not necessary
- Some limitations on transforms would be desirable as a safety feature. Could also add a reset button.
- We need to add a GetPatientOrientation().
- Reader (Andinet)
- Needs to add state machine in the SuperReader class
- Will not have an ImageWriter() but will have screenshots
- Registration (Andinet)
- Need a robust registration verification as in the paper by Jay West et al. – note that Jay now works at Accuray and we can contact him if needed – he is also on our program committee for SPIE medical imaging
- Tracker class (David)
- There are some additional events to add (what are these?)
- Need to “de-Cify�? the code
Iteration 7: lung biopsy with optical tracking
- Rick will come up with a storyboard (description) for this application
- Will basically be an improved 4ViewTrackingWithCT
- Will include path planning (selection of target lesion and skin entry point)
- The nodule could be represented by a sphere or other spatial object
- Will require a representation of the needle (a cylinder?)
- Will require annotation updates in real-time
- Will use respiratory phantom at Georgetown to test
Iteration 8: Ultrasound guided RF ablation of liver lesions (UNC)
- Use flock of birds tracker
- Ultrasound probe spatial object
- Image based registration
- Pre-op CT or MR plus 2D ultrasound
- Oblique image plane
Other notes
- Julien asked if we would support wireframe and point cloud representations. We will not support it in this version.
- Andinet asked what we should do about the invalid method? David proposed a default action per input. Luis suggest we require dense transition tables (note Stephen’s email on this from 11/12). Kevin Gary suggested sending an event back to any input that is not defined.
- As an example for DICOMReader there are 11 input types and 4 state types which makes 44 transitions – 14 are currently defined
- We will discuss this on the tcon this week
End Business
- A motion was made to commend Hee-su Kim for his excellent work on the project and the contribution of the logger. The motion passed unanimously.
