2009 Summer Project Week Project Segmentation of Muscoskeletal Images

• 

  Projects List

• 

  Knee MRI Image

• 

  Filled Femur Label Map

• 

  EM Segmented Output

• 
• 

  64 Affine Registered on 58

• 

  64 Pipeline Affine Registered on 58

• 

  64 Bspline Registered on 58

• 

  Masked 42 MRI

• 

  Masked 64 MRI

• 

  Registered Image Difference of 42 and 64

• 

  BSpline Registered Output of 42 and 64

Key Investigators

• Stanford: Harish Doddi, Saikat Pal, Scott Delp
• Harvard: Ron Kikinis
• Steve Pieper, Isomics, Inc.
• Luis Ibanez, Kitware, Inc.

Objective

The aim of this project is to develop an automatic/semi-automatic methodology to convert whole body imaging datasets into three-dimensional models for neuromuscular biomechanics and finite element simulations.

Specific Aims

• Explore the segmentation techniques for knee MRI datasets with special focus on patella, femur and tibia bones.

Approach, Plan

We are working on understanding the capabilities of RegisterImage module in Slicer to apply image-to-image registration on knee datasets. We are implementing masking algorithms to isolate specific knee bones, and perform parameter exploration to evaluate the sensitivity of registered images to input parameters. We are exploring the feasibility of applying python ICP-based registration algorithms to directly morph a surface model to a target image geometry.

Our goals for the project week are:

• Implement an algorithm to acquire masked regions of interest from MR datasets.
• Perform and evaluate results from a parameter space exploration study of RegisterImages Batchmake module on knee dataset.
• Resolve issues in building Python modules from slicer source code.
• Explore possibility of model-to-image registration using existing atlas (.vtk, .stl) to a target image using Python ICP Registration module.

Progress

• Investigated the masking algorithm and completed the module to mask and register images. Need to explore the parameters technique for this code.
• Resolved issues in building Python
• Cluster set up completed
• Converted Model to masked label map volume to further a masked label volume. Now we have the problem of registering label map volume to an MRI image.