Summer2009:Extension of the Command Line XML Syntax/Interface

• 

  JIST Pipeline

• 

  Multi-Core and Grid Processing

• 

  MIPAV Plugins

• 

  Example: Fully automatic surface, DTI, and sub-cortical analysis

• 

  Project Week Main Page

Key Investigators

• Johns Hopkins/Vanderbilt: Bennett Landman
• GE Research: Jim Miller
• Johns Hopkins: Heba Mustufa (not attending in person)

Objective

We will extend the command line interface (CLI) to support more robust/generic communication between SLICER and third party modules. The primary object is to enable cross-compatibility between the MIPAV/JIST module framework and the SLICER CLI framework. Successful completion of this task would enable the Johns Hopkins CRUISE (cortical surface) and CATNAP (diffusion tensor imaging) modules to be run within SLICER. The secondary (and long term) objective is to establish a flexible and usable syntax for module/plugin communication with graphical front-ends which could be reused in other medical image analysis settings.

Approach, Plan

The XML syntax and parser implementation for the SLICER command line interface (CLI) will be studied. Recommendations will be made for extended (backward compatible) syntax. A prototype parser for the new schema will be implemented within SLICER.

Progress

• Surveyed existing CLI interface and identified areas for extension
• Implemented prototype JIST/MIPAV->CLI bridge and discovery procedure.
• Circulated drafts of revised "meta-protocol" layer for adaptation between existing CLI
• Communicated with XIP team about generating a common/compatible interface layer

See detailed results below.

Results

Areas for Investigation with CLI and Existing Limitations

• Multiple “programs” per executable.
• De-couple dependence of executable query (program which is called with “—xml") from the executable which implements desired functionality.
• Provide a mechanism to report non-resource (file-base) results, such as integer, vectors, etc.
• Report to both stdout (via xml markup) [optional]
• Generalize the calling syntax --- especially the ability to control argument structure.
• Allow authors to provide more meta-information about the algorithm (which may be ignored).
• Enable the concept of an “ignorable” parameter which need not be understood to use the program.
• Increase structure for information that authors can provide about their algorithms and moved this to a separate category.
• Increase documentation for “type-of” arguments. Attempted to provide more clarity on a “data-centric” view of the data rather than how these would be represented in ITK/VTK.
• Provide preferred and supported format syntax for resource (file) based communication.
• Ability to model dependencies between flags
• Workflow level perspective – “sample a bunch of points” = harness level things
• Ability to have variable number of output arguments and determine the number / names at RUNTIME

JIST/MIPAV->CLI Bridge

• Automated discovery of JIST/MIPAV algorithms from the command line: edu.jhu.ece.iacl.jist.cli.discover

• Adding Noise                            	edu.jhu.bme.smile.demo.AddingNoise
  Chebyshev Fitting                       	edu.jhu.bme.smile.demo.ChebyShevFitting
  Demo3: Image Algebra                    	edu.jhu.bme.smile.demo.ImageAlgebra
  Image Arithmetic                        	edu.jhu.bme.smile.demo.ImageArithmetic
  Demo2: Generate Random Volumes          	edu.jhu.bme.smile.demo.RandomVol
  Demo4: Scale input image                	edu.jhu.bme.smile.demo.ScaleImageDemo
  staple                                  	edu.jhu.bme.smile.demo.SmileAlgorithmDemo
  ...
  

• Automated self-documenting executables

• usage: edu.jhu.ece.iacl.plugins.registration.MedicAlgorithmFLIRTCollection
         [-h] [-inApply <arg>] [-inCoarse <arg>] [-inCost <arg>] [-inDegrees
         <arg>] [-inFine <arg>] [-inInput <arg>] [-inMaximum <arg>]
         [-inMinimum <arg>] [-inMultiple <arg>] [-inNumber <arg>]
         [-inNumber2 <arg>] [-inOutput <arg>] [-inReference <arg>]
         [-inRegistration <arg>] [-inSkip <arg>] [-inSource <arg>]
         [-inSubsample <arg>] [-inTarget <arg>] [-inUse <arg>]
         [-outExecution] [-outRegistered] [-outTransformation]
         [-outTransformation2] [-outTransformations] [-xDir <arg>] [-xFile
         <arg>]
  
  Optimized Automatic Linear Registration 1.1.1.1 1.1.1.1 unk
  Linear Registration algorithm based on FLIRT.
   -h,--help               Print this message.
   -inApply <arg>          Apply rotation [option:All|X|Y|Z] (required,
                           default=All)
   -inCoarse <arg>         Coarse angle increment [double] (required,
                           default=15.0)
   -inCost <arg>           Cost function [option:Correlation ratio|Least
                           squares|Normalized cross correlation|Normalized
                           mutual information] (required,
                           default=Correlation ratio)
   -inDegrees <arg>        Degrees of freedom [option:Rigid - 6|Global
                           rescale - 7|Specific rescale - 9|Affine - 12]
                           (required, default=Affine - 12)
   -inFine <arg>           Fine angle increment [double] (required,
                           default=6.0)
   -inInput <arg>          Input Weighted volume [file] (optional)
   -inMaximum <arg>        Maximum angle [double] (required, default=30.0)
   -inMinimum <arg>        Minimum angle [double] (required, default=-30.0)
   -inMultiple <arg>       Multiple of tolerance to bracket the minimum
                           [integer] (required, default=10)
   -inNumber <arg>         Number of iterations [integer] (required,
                           default=2)
   -inNumber2 <arg>        Number of minima from Level 8 to test at Level 4
                           [integer] (required, default=3)
   -inOutput <arg>         Output interpolation [option:Trilinear|Bspline
                           3rd order|Bspline 4th order|Cubic
                           Lagrangian|Quintic Lagrangian|Heptic
                           Lagrangian|Windowed sinc|Nearest Neighbor]
                           (required, default=Trilinear)
   -inReference <arg>      Reference Weighted volume [file] (optional)
   -inRegistration <arg>   Registration interpolation
                           [option:Trilinear|Bspline 3rd order|Bspline 4th
                           order|Cubic Lagrangian|Quintic Lagrangian|Heptic
                           Lagrangian|Windowed sinc] (required,
                           default=Trilinear)
   -inSkip <arg>           Skip multilevel search (Assume images are close
                           to alignment) [boolean] (required, default=false)
   -inSource <arg>         Source Volumes [file collection: semi-colon
                           delimited list] (required, default=)
   -inSubsample <arg>      Subsample image for speed [boolean] (required,
                           default=true)
   -inTarget <arg>         Target volume [file] (required)
   -inUse <arg>            Use the max of the min resolutions of the two
                           datasets when resampling [boolean] (required,
                           default=true)
   -outExecution           Execution Time [string] (required)
   -outRegistered          Registered Volumes [file collection: semi-colon
                           delimited list] (required, default=)
   -outTransformation      Transformation Matrix [matrix semi-colon
                           delimited rows in text] (optional)
   -outTransformation2     Transformation Matrices [file] (optional)
   -outTransformations     Transformations [file collection: semi-colon
                           delimited list] (required, default=)
   -xDir <arg>             Request Output: Processing Directory (default
                           current) [directory] (optional)
   -xFile <arg>            Request Output: Results File (default output.txt)
                           [file] (optional)
  
  Provided by: JIST (Java Image Science Toolkit) Command Line Interface v0.1
  http://www.nitrc.org/projects/jist/
  

• These features will be released in the next version of JIST.

Framework for Interoperability of Neuroimaging Software

• Existing Slicer CLI can be bridge to almost any command line system through use of shell scripts.
• These scripts are platform specific and are complex to extend for larger projects --- do we write script factories? Factories for platform specific factories?
• If we are going to address the problem of inter-communication, perhaps it would be best to develop a SIMPLE architecture agnostic XML syntax for functionality query, programmatic direction, and result response.
• Started a NITRC project (FOPA - Framework for Open Programmatic Access) to develop multi-platform reference implementations. http://www.nitrc.org/projects/fopa/

References

• MIPAV: http://mipav.cit.nih.gov/
• JIST: http://www.nitrc.org/projects/jist/
• SLICER CLI: http://www.slicer.org/slicerWiki/index.php/Slicer3:Execution_Model_Documentation