2013 Project Week:CARMA LA Seg Gopal

From NAMIC Wiki
Revision as of 22:51, 27 February 2013 by Sbengali (talk | contribs)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search
Home < 2013 Project Week:CARMA LA Seg Gopal

Key Investigators

Gopal Veni, Salma Bengali, Greg Gardner, Alan Morris, Josh Cates, Rob MacLeod, Ross Whitaker

Automated LA Segmentation

The goal of this project is to begin Slicer implementation and testing of Gopal and Ross' segmentation method: "Left atrial wall segmentation using intensity profile based feature detector and optimal graph-cuts." Our objective is to implement this method in Slicer. There are several fairly involved steps for the entire segmentation process, including

  1. An LA shape model building phase
  2. User input to identify the center of a region of interest
  3. Computation of the segmentation, given a new LGE-MRI image
  4. Surface reconstruction of the output point set and (optionally) scan-conversion of the surface mesh to a binary segmentation volume.

For the Winter Project Week, our team will focus on steps 2-3 of this pipeline, with the goal of implementing a command-line module to perform the processing. Step 1 of the process will be implemented in a future planned Slicer module, but for now we will use an existing model built from the the CARMA LA shape database (available through the Data section of the CARMA DBP Wiki page).

Objective

Our Project Week objective is to produce a preliminary version of the segmentation computation outlined in steps 2-3 above.

Approach, Plan

We plan to implement this algorithm as a Slicer command-line module. Gopal has existing C++ code, but this code relies upon a 3rd party LGPL library for the Graph Cut algorithm. We will need to identify an alternative to this 3rd party code. Several other NAMIC partners have past projects requiring Graph Cuts, so there may be a solution available. We will investigate.

Progress

Discussed with Jim and Jc on the possibilities of handling/embeding the model-based-data while the module is being uploaded. Module written but needs to be integrated into the Slicer. Currently, dealing with some slicer build based errors. Once sorted, the module should be ready to check-in.