2014 Project Week:MultidimensionalDataSupport

Home < 2014 Project Week:MultidimensionalDataSupport

Projects List
Click here to see the video!
Sequence can be created from any MRML node (markup sequence)
4D labelmap statistics module
Multidimensional data browser (tracked ultrasound sequence)
Multidimensional data editor (DCE-MRI sequence)

Key Investigators

Add support for node bundle sequences (e.g., time sequence, multi-parametric images, etc.)
- All node types: not just volumes but also transforms, models, markups, and any custom nodes
- Large data sets: up to 5000 bundles, with any number of nodes in each bundle, any size of nodes
Operations:
- Import from sequence metafile (time sequence of a 2D image and multiple tool tracking data)
- Save into scene file
- Load from scene file
- Browsing sequence items: automatic replay, prev/next item; ability to show multiple items at the same time
- Modification of sequence items: add/delete, change bundle name, parameter

Even after extensive MRML scene optimization, handling of large data sets (more 1000+ nodes) is not feasible due to the inefficiency of MRML widgets (node selector combo boxes, scene hierarchy trees; due to each widget maintaining its own qMRMLSceneModel)
Plan: store one node per sequence and store each item inside that sequence node (similarly to MultiVolume, but for any node type)

Sequence of any kind of nodes can be stored in a multidimensional data node
- markup time sequence: http://screencast.com/t/yUsn0nKoYitB
- DCE-MRI: http://screencast.com/t/Oy8InJnz9FE
Large number of time points can be loaded: tracked ultrasound sequence with 2000 time points is loaded within 15 seconds, can be replayed at 30fps - see video at http://screencast.com/t/2sJbQpWf2l
Example 4D data processing module created: labelmap statistics
Next steps: proper management of node references (display node, storage node, other visible referenced nodes, other hidden referenced nodes)