Grid-Based Breast Cancer Screening and Epidemiology: the MammoGrid Project

Breast cancer as a medical condition and mammograms as images exhibit many
dimensions of variability across any population. Similarly, the way diagnostic
systems are used by clinicians and maintained by technicians varies between
imaging centres and breast screening programmes, as does the appearance of the
mammograms generated.  In general, the sensitivity and specificity of breast
screening programmes is unacceptably poor.  Studies of second screening opinion
suggest that technological support offers the best chance forward.

A geographically distributed database of histologically proven cases, both
positive and negative, that reflects the spread of pathologies across the
population would quickly become an essential evidence base for radiologists in
screening programmes and also those in training.  It would also be an
invaluable tool for the epidemiologist who wishes to study the effects of, say,
lifestyle and diet.

Exploiting existing and emerging technologies, the aim of the MammoGrid project
is to develop a Europe-wide database of mammograms that will be used to
investigate important healthcare applications and to explore the potential of
the grid to support effective co-working between healthcare professionals. In
particular, the project aims to prove that grids infrastructures can be
practically used for collaborative medical image analysis, especially to enable
radiologists from geographically dispersed hospitals to share standardized
mammograms, to compare diagnoses (with and without computer aided detection)
and to perform sophisticated epidemiological studies across national
boundaries.

This leads to several technical issues, including the standardization of
mammograms, design of an appropriate clinical workstation and distribution of
data, images and clinician queries across a grid-based database while
respecting patient confidentiality and security protocols. This talk will
outline the approach taken in MammoGrid to connect radiologists across a Grid
using an "information infrastructure", a DICOM-compliant object model residing
in multiple, distributed data stores in Italy and the UK, and a service-based
approach to managing a federation of grid-connected mammography databases. 


Tony Solomonides was formerly Head of Department and Associate Dean of Faculty
at the University of the West of England, Bristol.  Trained originally in
Mathematics, he worked briefly in topology and its applications to non-linear
mechanics before turning to Computer Science.  He is now a Reader in Computer
Science in the Centre for Complex Cooperative Systems at UWE and a CERN
Associate.  He has researched scientific databases and applications since the
mid-1990's, having also worked in medical informatics, especially the
management of diabetes and pathology of inflammatory bowel disease, before
then.  He has participated in the development and application of the CRISTAL
database to track the construction of a Large Hadron Collider experiment at
CERN.  He has also written on the application of 'description-driven' methods
and systems to the medical field, notably on large scale Electronic Patient
Record, and on the tracking of antibiotic resistant bacteria in the community.
He has also published extensively in the field of data collection through
social survey instruments.