View source for 2014 Project Week:MultidimensionalDataSupport

__NOTOC__
<gallery>
Image:PW-SLC2014.png|[[2014_Winter_Project_Week#Projects|Projects List]]
Image:MultidimEditor.png|Multidimensional data editor
Image:MultidimBrowser.png|Multidimensional data browser
</gallery>

==Key Investigators==
* Queen's: Andras Lasso
* BWH: Andriy Fedorov
* Isomics: Steve Pieper
* Kitware Jean-Christophe Fillion-Robin
* UHN: Kevin Wang
* Duke: James Cook

==Project Description==
<div style="margin: 20px;">
<div style="width: 27%; float: left; padding-right: 3%;">
<h3>Objective</h3>
* 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
</div>
<div style="width: 27%; float: left; padding-right: 3%;">
<h3>Approach, Plan</h3>
* 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)
</div>
<div style="width: 27%; float: left; padding-right: 3%;">
<h3>Progress</h3>
* 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
* Implementation is available: https://subversion.assembla.com/svn/slicerrt/trunk/MultidimData/src
</div>
</div>

==Use cases==
* Tracked ultrasound image sequence review
* Adaptive RT image sequence review
* 4D Cardiac CT volume rendering
* DCE-MRI review