Information Technology Reference
In-Depth Information
Fig. 4
Diagram of a syntactic image recognition (classification) system
This publication presents the use of graph grammars, as they are a more robust
tool for describing images than sequential or tree grammars. The basic assumption
behind these methods is that it is possible to define a mechanism generating graph
representations of the images considered. This mechanism is the appropriate graph
grammar, whereas the set of all graph representations of images that it generates is
treated as a certain language. We therefore have to build an automaton recognising
elements of this language. This automaton, or more exactly, its software imple-
mentation - the syntactic analyser (parser) -- is responsible for the recognition
procedure and allows the image description written using the proposed language
to be converted into a description reaching down to the semantic sphere, enabling
all material medical facts associated with the image examined to be understood.
Creating a graph model of the 3D structure of the analysed vessels and its linguis-
tic description makes it possible for the computer to analyse the structure obtained
in order to automatically detect the location of the stenosis, its extent and type
(concentric or eccentric). This representation yields a brief, unambiguous descrip-
tion of all elements of the vascular structure, thus supporting the further reasoning
about its correct function or functional irregularities. In the future, it will be possi-
ble to combine this type of a description with the haemodynamic modelling of the
flow in vessels affected by the disease, and this will help to link the lesions found
in the morphology of coronary vessels with pathological blood supply to and the
hypoxia of individual fragments of the heart muscle. In addition, using such se-
mantic descriptions in integrated modules of intelligent medical diagnostics sys-
tems can help in the early detection of pathological stenoses leading to hearth
hypoxia or anginas.
The road to finding the appropriate languages necessary to describe the seman-
tic part of 3D coronary vascularisation reconstructions is long and hard. The de-
scription of semantically important aspects of an image or its part cannot depend
on details which are of secondary significance from the point of view of under-
standing the image contents and which thus produce additional information
surplus which does not contribute anything to the final assessment of the given
image. This is why, apart from developing the linguistic description methods for
3D images and from coming up with intelligent methods of using experts'
Search WWH ::
Custom Search