EMSegmenter-Tasks:CT-Hand-Bone

Return to EMSegmenter Task Overview Page

Description

Single channel automatic segmentation of CT hand scans into the finger bones based on the publication ( ). The task can be applied to right and and left hand scans. Subjects should be scanned with a similar protocol as atlas, which is ....

The pipeline consist of the following steps:

• Step 1: Flip/Mirror the atlas (based on right hand scans) if scan of the subject is of the left hand.
• Step 2: Create Binary MAP of the hand for the atlas template and CT scan. For each scan:
  • Generate binary map of the scan by thresholding the scan above 150
  • Fill holwes in

Blur resulting binary map ( via itk::DiscreteGaussianImageFilter (parameter Variance=1.5, MaximumKernelWidth 5)

• • Binarize blurred image image

Use itk::BinaryThresholdImageFilter to remove artifacts. Set everything between 0 and 30 to 0, otherwise 255

• • extract largest component

Use itk::BinaryThresholdImageFilter to extract label 255.

Use itk::ConnectedComponentImageFilter to label the objects in the binary image. Each distinct object is assigned a unique label.

Use itk::RelabelComponentImageFilter to sort the labels based on the size of the object: the largest object will have label #1, the second largest will have label #2, etc.

Use itk::BinaryThresholdImageFilter to extract label 1. This is the largest object in the data set.

• Step 2: Register the binarized atlas template to the binarized CT scan via BRAINSFit (Johnson et al 2007)
  • Step 2a: Use BRAINSFit to perform a affine registration.
  • Register the atlas template linear to the subject scan and save the linear transformation. (BRAINSFit Rigid,Affine)
  • Step 2b: Use BRAINSDemonWarp to perform a non-linear registration.*
• Register the atlas template non-linear to the subject scan. Use the linear transformation as initialization. (BRAINSDemonWarp)
  • Step 2c: Use BRAINSResample together with the --deformationVolume option to resample the atlas files.
• Step 3: Compute the intensity distributions for each structure
  

Compute intensity distribution (mean and variance) for each label by automatically sampling from the MR scan. The sampling for a specific label is constrained to the region that consists of voxels with high probability (top 95%) of being assigned to the label according to the aligned atlas.

• Step 4: Automatically segment the CT scan into the structures of interest using EM Algorithm (Pohl et al 2007)

Anatomical Tree

The anatomical tree represents the structures to be segmented. Node labels displayed below contain a human readable structure name and in parentheses the internally used structure name.

• Hand
  • Air
  • Tissue
  • Index finger / digitus secundus (II)
    • Proximal (II)
    • Medial (II)
    • Distal (II)
  • Middle finger / digitus medius (III)
    • Proximal (III)
    • Medial (III)
    • Distal (III)
  • Ring finger / digitus annularis (IV)
    • Proximal (IV)
    • Medial (IV)
    • Distal (IV)
  • Little finger / digitus minimus (V)
    • Proximal (V)
    • Medial (V)
    • Distal (V)

Result

• 

Collaborators

Vincent Magnotta (University of Iowa)

Acknowledgment

The construction of the pipeline was supported by funding from NIH NCRR 2P41RR013218 Supplement.

Citations

• Pohl K, Bouix S, Nakamura M, Rohlfing T, McCarley R, Kikinis R, Grimson W, Shenton M, Wells W. A Hierarchical Algorithm for MR Brain Image Parcellation. IEEE Transactions on Medical Imaging. 2007 Sept;26(9):1201-1212.
• S. Warfield, J. Rexilius, P. Huppi, T. Inder, E. Miller, W. Wells, G. Zientara, F. Jolesz, and R. Kikinis, “A binary entropy measure to assess nonrigid registration algorithms,” in MICCAI, LNCS, pp. 266–274, Springer, October 2001.
• Johnson H.J., Harris G., Williams K. BRAINSFit: Mutual Information Registrations of Whole-Brain 3D Images, Using the Insight Toolkit, The Insight Journal, July 2007
• T. Vercauteren, X. Pennec, A. Perchant, N. Ayache. Symmetric Log-Domain Diffeomorphic Registration: A Demons-based Approach. MICCAI 2008