|
The monograph deals with an important application of computer vision and pattern
recognition in the area of medical science, more specifically reconstructive craniofacial
surgery. Craniofacial fractures are encountered very frequently in today’s fastpaced
society; the major causes being gunshot wounds, motor vehicle accidents, and
sports-related injuries. Surgical reconstruction is challenging because the surgeons
in the operating room have to register the broken bone fragments accurately and
under severe time constraints. Within the broad class of craniofacial fractures, the
emphasis in this monograph is on mandibular fractures since the mandible is often
unprotected and exposed making it especially vulnerable to accidents and injuries.
A typical input to a computer vision-based system for virtual craniofacial surgery is
a sequence of Computed Tomography (CT) images of a fractured human mandible.
The detection of fractures in CT images, the other integral component of reconstructive
craniofacial surgery, is often difficult because of the complexity of the fracture
patterns, missing data, image intensity inhomogeneities, and presence of noise and
undesired artifacts.
A formal treatment of computer vision-guided craniofacial surgery entails the
solving of two broad classes of problems, i.e., computer-aided fracture detection and
virtual reconstruction, both of which raise several important theoretical and practical
issues. From a theoretical standpoint, the monograph discusses several traditional
topics in computer vision and pattern recognition such as image registration, image
reconstruction, combinatorial pattern matching, and detection of salient points and
regions in an image. Several useful algorithms and concepts from two traditionally
diverse disciplines, namely, graph theory and statistics are seen to be applicable
in this context. The relevant topics from graph theory include maximum-weight
graph matching, maximum-cardinality minimum-weight matching for a bipartite
graph, maximum-flow minimum-cut determination in a flow graph, and construction
of automorphs of a cycle graph. The monograph demonstrates how the above
graph-theoretic algorithms can be applied to solve some important problems in computer
vision and pattern recognition that pertain to virtual reconstructive craniofacial
surgery. The various statistical techniques brought to bear include Markov random
fields, hierarchical Bayesian restoration, Gibbs sampling, and Bayesian inference.
This unique text/reference discusses in depth the two integral components of reconstructive surgery; fracture detection, and reconstruction from broken bone fragments. In addition to supporting its application-oriented viewpoint with detailed coverage of theoretical issues, the work incorporates useful algorithms and relevant concepts from both graph theory and statistics. Topics and features: presents practical solutions for virtual craniofacial reconstruction and computer-aided fracture detection; discusses issues of image registration, object reconstruction, combinatorial pattern matching, and detection of salient points and regions in an image; investigates the concepts of maximum-weight graph matching, maximum-cardinality minimum-weight matching for a bipartite graph, determination of minimum cut in a flow network, and construction of automorphs of a cycle graph; examines the techniques of Markov random fields, hierarchical Bayesian restoration, Gibbs sampling, and Bayesian inference. |
|