@MASTERSTHESIS\{IMM2012-06271,
    author       = "D. H. Hald and R. N. Wiener",
    title        = "Segmentation of Craniofacial Anatomy using Deformable Models",
    year         = "2012",
    school       = "Technical University of Denmark, {DTU} Informatics, {E-}mail: reception@imm.dtu.dk",
    address      = "Asmussens Alle, Building 305, {DK-}2800 Kgs. Lyngby, Denmark",
    type         = "",
    note         = "Supervised by Professor Rasmus Larsen, rl@imm.dtu.dk, {DTU} Informatics",
    url          = "http://www.imm.dtu.dk/English.aspx",
    abstract     = "This thesis presents methods for assessing and estimating the intracranial volume in children with unicoronal synostosis, on the basis of {3D} {CT} images. Obtaining such a measurements provides the opportunity to compare the intracranial volume in these children with i.a. to normal data, in order to comment on possible deviations between such two groups. Furthermore, it provides a tool for a possible surgery evaluation, given as pre- and post measurements. To solve this problem, two deformable models have been chosen; an image registration, and a graph cut based algorithm. The image registration model transforms a template image into a given reference, especially by means of a {B-}spline transformation. By extracting the transformation parameters and applying them to a manual segmented mask, derived from the template image, an estimation of the volume of interest is obtained. The graph cut model takes a gradient based approach to the problem, and based on construction of a weighted graph, consisting of nodes and edges, it finds the optimum cut i.e. the segmentation. The true challenge for this model, lies within the construction of this framework, i.e. establishing the proper weights and neighborhood connections. The performance of the two models is validated on the basis of a voxel deviation found between the model based segmentation and a semi-automatic segmentation, performed with manual editing. Furthermore, visual interpretation of the segmentation surfaces has been performed. Both models showed great results, and presented a good foundation for further applicability."
}