Segmentation in 3D Slicer
|An extensive set of tools is available within 3D Slicer to support your segmentation task. The right module will depend on your input data and the underlying question asked. Below is an overview of the main and auxilary modules related to image segmentation. The spectrum ranges from fully automated to manual segmentation. Most modules are generic and can handle any image content, but a few are designed specifically for brain images. This page is organized by methods.|
Voxel Based Methods
This module is designed for experienced user who want to perform atlas based automatic segmentation. The module does not constrain the number of anatomical classes to be segmentation. Each anatomy is characterized via an atlas that captures the variation of that structure within the population. The user defines associates atlases with anatomical structures by completing a predefined workflow. This parameterization of the algorithm then results in a template for the task that than can be applied to a new set of images that were acuiqred with the same imaging protocol. EMSegment Simple is for less experienced users who would like to apply these templates to their own images. EM Segment Command provides users with the same functionality as EM Segment Simple. However, the module can be executed via batch mode.
This filter creates a binary thresholded image that separates an image into foreground and background components. The filter calculates the optimum threshold for those two classes automatically. The module requires a minimal user input.
BRAINSROIAuto is a program whose purpose is to automatically generate a Binary Image (or Mask) to encompass the region in an brain image volume occupied by the head (separate head from the background).
Region Growing Algorithm
This module is designed to segment single structures that have homogeneous intensity pattern. The module is simply initialized by placing fiducials inside the object. Users can interactively modify the resulting segmentation by scrolling through the evolution of the label contour.
These are statistical region growing algorithm. Similar to fast marching, these algorithms are initialized via setting seeds inside the object of interest. At each iteration, these approaches first generates an intensity model of the structure of interest from the the current segmentation and then update thee segmentation by labeling all voxels that are connected to that region and whose intensity values comply with the intensity model
Grow Cut Segmentation is a competitive region growing algorithm using cellular automata. The algorithm works by using a set of user input scribbles for foreground and background. For N-class segmentation, the algorithm requires a set of scribbles corresponding the N classes and a scribble for a don't care class.
Surface Based Approaches
This module is a tool to generate closed contours on a surface in 3D. The contour is initialized with a set of points, and subsequently 'evolves' according to some geometric criterion of the underlying surface (e.g. Surface Normal, mean curvature, second fundamental form, etc) and the embedded curve (e.g. geodesic & normal curvatures, etc ).
ABC is a full segmentation pipeline developed and used at University of North Carolina and University of Utah for healthy brain MRIs. The processing pipeline includes image registration, filtering, and inhomogeneity correction.
As part of the Vascular Modeling Toolkit extension, these modules target easy-to-use semi-automatic segmentation of tubular and blob-like structures. The methods provided by this module are Fast Marching Upwind Gradient Initialization and Geodesic Active Contours and CURVES Evolution.