ProjectWeek200706:LesionClassificationInLupus

Return to Project Week Main Page

Key Investigators

• MIND/UNM: Jeremy Bockholt, Mark Scully
• MGH:Bruce Fischl
• UIowa: Vincent Magnotta

Objective

Our goal is to automatically, or with little or no manual human rater input, accurately tissue classify our example lupus data-set into gray, white, csf, and lesion classes.

Approach, Plan

Our approach is to utilize existing automatic lesion classification techniques, both within NA-MIC kit and external to it.

Our plan for the project week is to learn and use EMSegment (S. Wells), BRAINS2 (V. Magnotta), and Constructing Image Graphs for Lesion Segmention (M. Prastawa). We will learn and begin to compare and contrast these approaches with a bronze standard (manual rater tracings) of lesions.

Progress

• Provided Kilian Pohl and example anonymous demo case with raw t1, t2, and flair, and coregister t1, t2, flair and manual lesions.
• Mark Scully has installed current Slicer3 and run the EMSegment tutorial with tutorial data
• We are currently struggling with trying to get the EMSegment module to work on the lupus sample data set

Additional Information

• Homepage for the MIND/UNM DBP (II)

Using an exemplar case that has already been processed using non-NAMIC kit tools:

1. Use existing NA-MIC kit to coregister T1, T2, Flair
2. Use the EM Segment in slicer 2-3.X to classify grey, white, csf, and white matter lesion
3. Summarize the volume and location of lesions