Scratch

TODO

• consider changing to use kwargs http://pythontips.com/2013/08/04/args-and-kwargs-in-python-explained/
• improved documentation for the function parameters
• include revision control for each feature

2015 Kick-off meeting notes

• need to keep provenance of the preprocessing steps

From earlier wiki

• separate library from Slicer
• consider packaging python module standalone, see http://www.scotttorborg.com/python-packaging/
• revisit naming
• revisit functionality and separation of concerns
• fix formatting
• performance
• at the moment, computation is parallelized at the case level (no parallel calculation of different features, no parallelization within feature calculation)
• add testing
• consider pure python testing - nose? http://pythontesting.net/framework/nose/nose-introduction/
• replace settings dictionary with the API on the classes
• add command line options for initializing parameters, allow JSON file with all options

Parameters:

• modality, patient ID, date, series - go away
• image/label file paths: change to SITK images
• baseimagespacing - goes away
• dimensions - go away
• key parameters: levels, binwidth, resampledpixelspacing
• add: enable/disable feature groups

Consider reusing SITK implementations

• see project http://wiki.na-mic.org/Wiki/index.php/2014_Summer_Project_Week:Pluggable_Label_Statistics

Priorities:

1. Set up testing

2. Naming conventions/styling: Slicer recommends https://www.python.org/dev/peps/pep-0008/, use docstring, see https://www.python.org/dev/peps/pep-0257/.

3. Python library packaging, separate from the Slicer-dependent code

4. Refactor communication of parameters from settings file to API

5. Review for redundancy with SITK

6. Refactor with parallelization in mind

7. Add description to individual features