WWW 94 Conference Paper: Networked Biomedical Image Hyperbase
Networked Biomedical Image Hyperbase
Authors
E. Duval, K. Hendrikx, H. Olivié
Computer Science Department, Katholieke Universiteit Leuven
Celestijnenlaan 200 A, B-3001 Leuven, Belgium
Tel: +32.16.20.10.15 (ext.3643); Fax: +32.16.20.53.08
Email: {Erik.Duval,Koen.Hendrikx,olivie}@cs.kuleuven.ac.be
P. O'Hanlon, D. Jameson
Audiovisual Centre, University College London
Windeyer Building, Cleveland Street, London W1P 6DB, UK
Tel: +44.71.636.8333 (ext.3056); Fax: +44.71.580.0995
Email: piers@livenet.ac.uk
J. Williams
Educational Technology Service, Bristol University
Royal Fort Annexe, Tyndall Avenue, Bristol BS8 1UJ, UK
Tel: +44.272.303500; Fax: +44.272.255985
Email: j.williams@bristol.ac.uk
Abstract
This paper presents a distributed hypermedia system, accessible through the World-Wide Web.
The main components are an image store and a database.
Our infrastructure is currently used in different educational scenarios.
In order to be able to carry out realistic field experiments, we have included the Bristol Biomedical Videodisc
in the image store. This is a collection of
ca. 20,000 high quality medical, veterinary and dental images.
We explain in some detail how WWW clients interact with the different components of our server.
Some general points are made regarding the WWW approach
and its applicability in a large-scale multimedia environment.
Partners
This paper presents some results of a collaborative research
and development effort, involving three academic partners:
We started this collaboration in the context of the European projects CAPTIVE
('Collaborative Authoring, Production and Transmission of Interactive Video for
Education', 1989-1991) and MTS ('Multimedia TeleSchools', 1992-1994)
[Beckwith et al.93,
Duval et al.94a,
Duval et al.94b,
Jameson et al.93].
Both projects were funded by the European Commission, in the framework of the
DELTA ('Developing European Learning through Technological Advance') program.
Our aim is to develop an infrastructure for access to educational
resources, enabling authors, developers, lecturers and students to locate and
access material relevant to their needs. For this purpose, we have developed:
- an image store, based on an analogue WORM (Write Once, Read Many
times) videodisk, suited for storage of still images and short video sequences;
- a database, called HyperDB, with descriptive data about educational
resources.
The material described in HyperDB can include traditional educational
resources, like books, video and audio tapes, slides, etc. As this material
cannot be made available on-line, the descriptive data stored in HyperDB
indicate how it can be obtained (e.g. by sending an electronic mail request). If
the relevant resources are stored on the image store, then they can be
obtained through our infrastructure, as explained further on.
Typical Scenario
In a typical scenario, an end user first accesses
HyperDB, in order to find out what material suits his needs. The role of HyperDB
is similar to that of an OPAC (Online Public Access Catalog) in a traditional
library. The underlying data model however is quite different:
- An OPAC is typically record-oriented: each record describes one book and consists of a
fixed set of description fields (author, title, etc.).
- Data in HyperDB are structured according to a hypermedia data model, i.e. as a network of linked
nodes. Links, nodes, webs and the participation of nodes in links can all be described by
an arbitrary set of (attribute,value)-pairs. HyperDB relies on a
DataBase Management System (DBMS) for lower level data management.
Using the DBMS query facility, HyperDB can act as a Hyperdocument library
[Engelbart90].
Once the relevant resources have been identified, they can be accessed
through a variety of means, depending on the type of resource and the
infrastructure available at the receiving site. If the resource is a still image
or a video sequence stored on the image store, then several options are
available:
- The image store is integrated in the LIVE-Net optical fiber network
that links different university colleges in London. This network is in turn
connected to a satellite uplink, so that images can be transmitted by Direct
Broadcast Satellite (DBS). This is especially appropriate in distance education
scenarios, when images need to be transmitted to a number of receiving sites.
Courses on electronical engineering are currently delivered to schools in North-Wales
[Brockley94] in this way.
- Although the images are stored in an analogue format, they can be digitized
at run-time, as the image store is connected to a codec (COder/DECoder) device. This makes it
possible to transmit a digitized version of the images over the Internet.
Because of the bandwith limitations and the resulting transmission delays, users
can first request transmission of thumbnail versions of the images in order to assess
their relevancy. This approach is obviously more appropriate for individual
consultation.
In order to be able to carry
out realistic field experiments, we have included the Bristol Biomedical
Videodisc in the image store. This is a collection of ca. 20,000 high quality
medical, veterinary and dental images with an accompanying database. The data
from that database have been incorporated into HyperDB.
Medicine, in general, is a highly visual subject and images form the basis of
much of the teaching material. The Bristol Biomedical Videodisc was developed to
meet the needs of teachers, particularly those creating computer-based teaching
material, for an easily accessible archive of images.
The videodisc was created as a shared resource with the images having been
donated by members of the leading medical, veterinary and dental institutions
worldwide. Donors can therefore access not only their own images but a
continually expanding collection of thousands more. It is distributed on a non-
profit-making basis, with the intention of making it as widely available as
possible to the general medical community.
A note on large-scale experiments:
In the work we describe here, the videodisc essentially functions as a source
of test material. We believe that is essential to have more of these experiments
involving substantial amounts of data, in order to investigate wether and how
the WWW can cope with these. The viability of the WWW approach for access to a
relatively limited number of data per server (rarely more than ca. 50 files) has
been proven by the enormous growth in popularity and the everyday extensive use
of the web. However, servers with tens of thousands of images, as well as data
describing the images, represent a different order of magnitude (see also next section).
Integrated Access through the World-Wide Web
The image store and HyperDB together constitute a distributed server. We rely heavily on the World-
Wide Web for interaction with this server. For this purpose, we have set up WWW
servers at each of the sites where a component of our server is located, in casu
in London (U.K.) for the image store and in Leuven (Belgium) for HyperDB.
The following figure gives an overview of the different components and the communication protocols involved:
Forms stored on each of the WWW servers define the user interface for access
to the corresponding component (Image Store or HyperDB).
The functionality itself is implemented in separate software that is
called when the form query is submitted, as explained in the sections about the
Interaction between client and HyperDB
and about the Interaction between client
and image store.
The role of the WWW in this set-up is two-fold:
- The forms facilities of HTML allow us to use WWW software
as a user interface management system: forms on our WWW
servers define the kind of objects (e.g. a text input field, a menu, etc.)
to be displayed to the end user. The WWW client will display these
objects as appropriate in the context of the end user environment (e.g. with a
motif look and feel under X-Windows).
- The HTTP protocol that defines interaction facilities for WWW clients and
servers makes it possible to provide access to the WWW servers (and through them
to the Image Store and HyperDB) from anywhere on the internet. This enables us
to achieve our aim of providing access to educational resources for a
potentially very large community of authors, developers, students and teachers.
It should be clear that none of the relevant data (apart from the forms) are stored
in HTML files, as is mostly the case with more conventional WWW servers. When a
form query is submitted to one of our WWW servers, a HTML document is generated
at run time by the program that implements the functionality. It has been
mentioned before (both on the Usenet newsgroup devoted to the WWW and in the www
-talk mailing list) that this approach is more suited for more advanced
applications.
We strongly believe that generating the HTML document to be delivered at run time
is more appropriate in large-scale environments:
- Special purpose hardware can be used for multimedia data storage
and processing: the image store e.g. is an
analogue WORM videodisk connected to a codec device for digitalization of the
images. An appropriate digital version of the image can be generated when
appropriate, e.g. a thumbnail picture icon for judging the relevancy of a
relatively large set of images, or a full-size, high quality picture for final
delivery of a more restricted set of images.
- The descriptive data in HyperDB are managed by a DataBase Management System
that provides higher level functionality than a file system:
- it supports the ACID properties of transactions (for Atomicity, Consistency, Isolation and
Durability);
- it includes a powerful query facility that enables users to
identify relevant data by specifying search constraints in a declarative manner,
rather than solely relying on browsing alone, which quickly becomes inadequate
when a large amount of data is involved.
The
Leuven WWW server holds a
form that enables end users to specify search constraints involving
characteristics of biomedical images (breed, sex, species, etc.). The purpose of
the constraints is to identify relevant images that satisfy the requirements of
the individual user.
(You can try it out with userid='guest' and password='notneeded'!).
Some search constraints (on Species, Breed, etc.) can be selected in a scrollable list.
Other comparison values must be typed in by the end user in a text field.
Such a value can include search patterns:
- '%' stands for an arbitrary character string;
- '_' stands for one arbitrary character.
The form enables end user to define search constraints, using a direct manipulation approach reminiscent of
Query By Example [Shneiderman 92]. Comparisons involving two attributes
describing the biomedical images are not supported (e.g. "Species > Description")
as they do not make sense in this case. A more general query facility should
support such comparisons as well.
When the form query is submitted to the Leuven WWW server (using the POST
method), a program is executed through the
Common Gateway Interface. This
program translates the original request into a query for the HyperDB.
As HyperDB runs under a Relational DataBase Management System (RDBMS),
the database query is formulated in SQL and executed by the RDBMS.
HyperDB delivers data describing the images that satisfy the search
constraints. The Leuven server thus effectively acts as a Hyperdocument library
[Engelbart 90], that enables end users to locate relevant resources.
A similar form has been defined for query based access to data about
organisations, persons, and journal articles. In fact, once the form query is
submitted to the WWW server, the program that is called through the Common
Gateway Interface (CGI), in order to process the query, is the same for these
forms and for the form concerning the descriptive data about the biomedical
images.
The access to the LaserDisc is achieved through a
CGI gateway. This has been
written so as to provide digitised images on demand, thus providing quick access
to a large image base. The real-time, on demand, digitising saves the use of
large quantities of disk space. The gateway allows the user to specify the frame
number on the disk (with possible exclusions due to copyright), in addition the
size and type of the image can be specified thus offering the opportunity to
provide both "thumbnail" images and full size images. This work has come from
work in related areas[1,2,3].
A low level interface has been developed to the LaserDisc, which offers
control of the device through it's serial port. The Sun Workstation connected to
the LaserDisc player also contains a VideoPix frame grabbing card. When a
request is made the frame number is verified as legal and a connection to the
LaserDisc is established, which selects the appropriate image from the disc (
access time is under 0.5 second). The VideoPix card is enabled and the image
grabbed, and according to the request the image is resized and converted to the
desired type before transmission.
An interface to the image store has been developed which will allow
individual access to the LaserDisc player, independently of the database program
. This allows users to browse the disc.
Plans for the Future
We have a number of plans for the immediate future:
- Access to HyperDB through the WWW, although operational,
needs further development:
- Not all facilities are currently implemented: it is e.g. envisaged that an
end user will be able to ask for the number of hypermedia resources described in
HyperDB that satisfy his search constraints before actually submitting his query,
so that he can either refine or broaden his search criteria accordingly.
- In parallel with our work involving the WWW, we have also
developed components that can be assembled into a customizable client in Tcl/Tk,
a user interface management system developed at the University of Berkely
[Ousterhout 91]. As there is a Tcl/Tk WWW client (tkWWW), it will be interesting
to investigate wether our Tcl/Tk components (a graphical browser and a input+
query tool) can be integrated with tkWWW.
- Access to the LaserDisc is fully functional though there are a number of
enhancements which we hope to make:
- Introduction of real-time movie generation in such formats as MPEG-I; Smal
video clips could be generated from the LaserDisc and delivered to the client
stations.
- Develop multiple server implementations so that images maybe delivered from
other sources on demand.
References
- [Beckwith et al. 93] R.C. Beckwith, D.G. Jameson, P. O'Hanlon, and E. Duval.
Interactive satellite teaching and conferencing using an image server.
Published in the Proceedings of the ESA Olympus Conference, April, 20-22, 1993,
Seville.
- [Brockley94]T. Brockley.Telecommunication Training and Education
in Rural Area Context (Wales). Eleventh International Conference on Technology
and Education - London, March 1994. Proceedings pp.330-332.
- [Duval et al. 94] E. Duval, H. Olivié, P. O'Hanlon, D.G. Jameson,N. Ismail,
S. Wilbur, and R.C. Beckwith. A multimedia information resource for
education. Eleventh International Conference on Technology
and Education - London, March 1994. Proceedings pp.1008-1010.
- [Duval et al.94b]E. Duval, H. Olivié, P. O'Hanlon, D. G. Jameson,
N. Ismail, S. Wilbur and R. Beckwith. Managing Networked Multimedia Data.
ACM SIGGRAPH Computer Graphics, Volume 28, No. 1, 1994, pp.15-19.
- [Duval & Olivié 94] E. Duval and H. Olivié, HOED:
Hypermedia Online Educational Database. ED-MEDIA 94, World Conference on Educational
Multimedia and Hypermedia, June 25-29, 1994, Vancouver, Canada.
- [Engelbart90] D. C. Engelbart, Knowledge-Domain Interoperability and an
Open Hyperdocument System, in ACM CSCW Proceedings 1990, pp.143-156.
- [Jameson et al. 93] D.G. Jameson, P. O'Hanlon, R.C. Beckwith, E. Duval, C.
Liekens, and H. Olivié. Distance learning by satellite - using an
information resource to provide a flexible learning system. In TELETEACHING
93, International Conference on Learning and working independent of time and
distance, Trondheim, Norway, 20-25 August 1993. International Federation for
Information Processing.
- [Ousterhout 91] J. K. Ousterhout, An X11 Toolkit Based on the Tcl Language.
1991 Winter Usenix Conference, also available via anonymous ftp from sprite.berkeley.edu in tcl/tkUsenix91.ps.
- [Shneiderman 92] B. Shneiderman, Designing the user interface:
strategies for effective human-computer interaction, 2nd. edition, Addison-Wesley.