TOADS

TOpology-preserving Anatomy Driven Segmentation (TOADS) provides a detailed segmentation of brain structures using statistical and topological atlases. The segmented structures are constrained to have the correct topology.

Input Types

The input image(s) should be skull stripped and have a correct information header. The best result is achieved from a T1-weighted image, but all MR pulse sequences can be used as the input of the algorithm.

Module Parameters

Main inputs

Image modality	Adjusts clustering parameters (eg. initial centroids) based on the modality of the image. T1_SPGR SPGR T1 Sequence parameters T1_MPRAGE MPRAGE T1 Sequence parameters PD PD Sequence parameters T2 T2 Sequence parameters FLAIR FLAIR Sequence parameters
Atlas file	A text file providing the location of the atlases, as well as the predefined clustering parameters for each input pulse sequence.
Output images	Specifies the type of the output images that are desirable. hard segmentation Outputs the topologically constrained hard segmentation of the brain. hard segmentation +memberships In addition to the hard segmentation, it also provides the fuzzy memberships for each structure in a 4D image. cruise inputs In addition to the hard segmentation, it outputs the sulcal CSF, filled WM, and cortical GM memberships, as well as the filled WM mask.  These are necessary for CRUISE algorithm to reconstruct the cortical surfaces. dura removal inputs Similar to cruise inputs but it also includes the cerebellum memberships. These inputs are necessary for Remove Dura algorithm. thalamus segmentation Outputs the hard segmentation and fuzzy memberships for the thalamus.
Output max membership classification	In addition to the topologically constrained segmentation, it also provides the hard segmentation computed directly from fuzzy memberships without any constraint.
Correct inhomogeneity	Models the inhomogeneity by a low degree polynomial in the clustering algorithm.
Output inhomogeneity field	Outputs the estimated inhomogeneity field (if correct inhomogeneity option is selected)

Advanced Options

Atlas prior	Controls the influence of the statistical atlases on the segmentation algorithm.
Smoothing parameter	Controls the amount of the spatial smoothness imposed on the fuzzy memberships.
Maximum difference	Convergence criteria that specifies the minimum amount of relative change in the energy function.
Maximum iterations	Maximum number of iterations before the algorithm stops.
Atlas alignment	The type of transformation used for registering the statistical atlases to the subject space. rigid This is the default option and generates the most robust results. multi_fully_affine A more elaborate registration that can be used if rigid registration does not generate good results.
Connectivity (foreground, background)	The connectivity rule used for defining the simple points in digital topology.

---

Example Usage

Input image.
Hard segmentation
Membership functions	Sulcal CSF Putamen Thalamus
cruise inputs	Cortical GM Filled WM membership Filled WM mask
Inhomogeneity field