Abstract Navigating and efficiently retrieving image and
other multimedia object from WWW has become one of the primary efforts
in a wide area of applications. In biomedical field this is justified due
to great number of bio-archives distributed all around the world which,
containing an extensive range of information such as biological text results,
clinical reports, images, audio and video recordings, improve the information
interchange and facilitate medical people in education and training, clinical
analysis and surgery. VHD anatomical images are available through internet
access but due to lack of a database architecture no driven query is allowed.
At VHD-MMS (Milano Mirror Site) we planned to integrate the image repository
with a query framework for content-based image retrieval. Not only the
system will allow retrieving images through their general information but
also by concept with semantic value and visual properties. The system is
built through an Object-Oriented Database Management System. The image
is mapped into an image model which contains attributes (meta, semantic,
visual) and rules to manage them. Graphics User Interface allows the user
to question the database based on keywords (organs, tissues, anatomical
structures, ..) belonging to a well defined knowledge source such as UMLS
and visual navigating the repository based on color, texture properties.
Through a Software Agents (SA) technology, the performance of the system
will be upgraded making the system able to provide a User-View of the database
content according to the specialty of the user. While being the data structure
single, the software agents architecture allows presenting the information
under different points of view. Virtual data models are generated and customized
to user skills. For instance while browsing an image a radiologist can
be interested in skeletal apparatus and neglects soft tissues. The system
can present a walk through into a set of images in which only the bones
are visible.
