Seg EM

seg_EM is a general purpose intensity based image segmentation tool. In it's simplest form, it takes in one 2D or 3D image and segments it in n classes. This function has implemented several other improvements (Please click the hyperlinks to go to specific sections):

• Input anatomical priors, either as a series of images or as a 4D image.
  • These priors should be pre-registered using for example NiftyReg.
• Markov Random Field for spatial consistency
• Bias field inhomogeneity correction to compensate for B0 non-uniformities as per Van Leemput et al. TMI (1999) PDF.
  • The Bias corrected image can also be outputted
• Outlier detection as in Van Leemput et al. TMI (2001) PDF
• Prior relaxation as in Cardoso et al. MICCAI (2011) Link
• Semi-conjugate prior over the model parameters as explained in Cardoso et al. MICCAI (2011) Link

The help for this function can be obtained by running

seg_EM -h