ISSN:
1741-0444
Keywords:
Feature analysis
;
Image representation
;
Medical imaging
Source:
Springer Online Journal Archives 1860-2000
Topics:
Biology
,
Chemistry and Pharmacology
,
Medicine
Notes:
Abstract Medical imaging is usually performed in such a way that for a particular type of examination the global structure of the resulting images will be more or less the same. A model can be developed to describe such a class of images. By using this model, images can be described by a set of parameters which define possible variations of the common image structure. In this way the essential image information is made easily accessible for further processing, while a large amount of invariant image data is deleted. This representation method has been applied to thallium-201 myocardial scintigrams. After normalising the image descriptions obtained, a mean normal image could be defined. By matching this with a given image set, the diagnostic importance of the described image properties could be estimated. Furthermodre, it is shown that images can be reconstructed from their descriptions with good resemblance to the originals. With an artificially composed set of images an experiment was performed to evaluate the influence of the image properties on human visual analysis. Experimental results corresponded to those obtained by objective image analysis.
Type of Medium:
Electronic Resource
URL:
http://dx.doi.org/10.1007/BF02442848
Permalink