Skip to content

Latest commit

 

History

History
53 lines (36 loc) · 3.21 KB

CONTRIBUTING_IDEAS_SUGGESTIONS.rst

File metadata and controls

53 lines (36 loc) · 3.21 KB
Raw

shell

Ideas and suggestions for contribution

You can contribute in different ways to different modalities, as noted below.

Alignment / Registration

  • Improved support for cross-modal comparison (PET on T1w, or fMRI on T1w etc)
    • edge detection algorithm performance is variable in different modalities due to variations in the properties of noise and contrast levels - hence tweaking it to different modalities e.g. PET , fMRI etc would be great.
  • Additional blending methods
  • Implementation of intensity renormalization methods
  • in the zoomed mode, ability to navigate through different slices with an arrow keys or scrollbar.
  • TO BE UPDATED

Image Quality Metrics (IQM)

  • implementation of relevant IQMs in pure python - some are described well here: QAP.

Tissue segmentation algorithms

Computation of IQMs, esp. more advanced ones, require tissue segmentation in the anatomical T1 mri space. While this can easily be via Freesurfer, FSL or AFNI, the design inclination of VisualQC is:

  • to reduce the number of external dependencies (esp. big installations that are hard to acquire and maintain for a novice user without any system admin experience or help)
  • to being able to compute only what is needed and only when required

So implementation of tissue segmentation algorithms natively in Python would greatly help with a smooth user experience over time for a novice user e.g. reducing the installation difficulties, delivering new features and bug fixes and reducing the amount of computation.

  • implementation of reasonably fast and sufficiently accurate tissue segmentation in pure Python (integrating existing python packages that are well-tested is fine also). This is helpful to QC both fMRI as well T1w MRI scans.
  • TO BE UPDATED

fMRI Preprocessing

  • Implementation of algorithms for minimal preprocessing of fMRI scans such as head-motion correction, slice-timing correction etc. Some work has been done already at pyaffineprep wherein we could contribute to making it more stable and testing it thoroughly.
- Reorder the rows in the carpet plots in interesting groups:
  • such as within each tissue class (from above)
  • using arbitray parcellations, that are either data-driven (e.g. clustering voxels by time-series) or motivated by another user-chosen criteria
  • Incorporating task design to highlight event borders, or to check for task-coupled artefacts (such as motion)
- Ability to mark individual frames (as "corrupted", or for "further review") for subsequent processing (such as censoring) and analyses.
  • This can be achieved with the Notes functionality e.g. including appropriate text bad_frames:{3,17} that can later be parsed programmatically.
  • However, this can be made much easier with clever interface and programming e.g. Ctrl+click on a particular frame can mark it is one idea.
  • TO BE UPDATED

TO BE UPDATED