The aim of this project is to, in light of emerging Grid technology, develop a European-wide database of mammograms that will be used to investigate a set of important healthcare applications as well as the potential of this Grid to support effective co-working between healthcare professionals throughout the EU. The project aims to concentrate on the application of Grid technology rather than merely focusing on its further development. Medical conditions such as breast cancer, and mammograms as images, are extremely complex with many dimensions of variability across the population.
Among the benefits of having a European-wide database are to provide:
- a statistically significant numbers of examples and conditions;
- a more diverse epidemiology;
- a wider variation in quality of images and diagnosis;
- an abstract interface for accessing heterogeneous databases;
- a potential knowledge discovery in the diagnosis and understanding of breast cancer.
1. Evaluate current Grids technologies and determine the requirements for Grid-compliance in a pan-European mammography database.
2. Implement the MammoGrid database, using novel Grid-compliant and Federated-DB technologies to provide improved access to distributed data and allow rapid deployment of software packages to operate on locally stored information.
3. Deploy enhanced versions of a standardisation system to enable comparison of mammograms in terms of intrinsic tissue properties independently of scanner settings, and to explore its place in the context of medical image formats (DICOM).
4. Develop software tools to automatically extract image information that can be used to perform quality controls on the acquisition process of participating centres (e.g. average brightness, contrast).
5. Develop software tools to automatically extract tissue information to be used in clinical studies (e.g. breast density, presence, number and location of micro-calcifications), for increasing the performance of breast cancer screening.
6. Use the annotated information and the images in the database to benchmark the performance of the software described in points 3, 4 and 5.
7. Exploit the MammoGrid database and the algorithms to propose initial pan-European quality controls on mammographic acquisition and ultimately to provide a benchmarking system to third party algorithms.
DESCRIPTION OF WORK
The goals of the project are to provide a "proof of concept" for each of the objectives. Longer term goals, beyond the lifetime of the project, will include the addition of new clinical participants, the extension of the standardisation software (s/w) to include the most frequent acquisition protocols across the EU, a facility for the benchmarking of Computer Aided Diagnosis and Detection (CADi, CADe) algorithms and the exploitation of the database (DB) to control mammographic image quality across the centres and develop full epidemiological studies on the images. Each of the objectives is associated with a measurable deliverable.
An information infrastructure ready to support a growing federated DB will be built according to the needs of the users and the applications to be deployed. This infrastructure will be tried and tested over the duration of the project including hardware performance, local DB architectures, package deployment, and DB performance according to the user requirements specified in the first phase of the project.
In order to compare mammograms it is necessary to obtain a representation independent of scanner settings and that holds intrinsic information on the breast tissue. The standardisation s/w achieves this goal from the digitised film and a series of acquisition parameters routinely obtained on a set of commercial scanners in particular screening programmes. The project will enable an extension of this algorithm to cater for other scanners including direct digital machines and the routine quality control (QC) parameters recorded in a variety of programmes. The package will be shown to display the current performance under the new conditions.
As a consequence of this effort and of the increased understanding of different QC protocols, the consortium will develop s/w packages for the verification of image quality. The deployment of these packages across the information infrastructure as a means for auditing the local contents of the federated DB and comparing results between centres will be tested during the project. The QC statistics will be compared against the annotated information. Poor QC or variations in quality are thought to be one of the major downfalls of the breast screening program considered as a whole process involving not only the control of the hardware related to image production (X-ray tube, sensors, etc.), but also image display and interpretation; in particular, images themselves and also interpretation ability can vary. A subset of the consortium has a world-class track record of developing CADi and CADe software. During the project these algorithms will be developed in the context of data-mining applications to enable users to perform image-based queries and the compilation of statistics regarding the distribution of particular anatomical characteristics and pathologies. These algorithms will be benchmarked against the annotated information in the DB providing a measurable indication of performance.
Clinical support will be engaged during the whole duration of the project to define the end-user requirements, provide digitised annotated data, test the functionality of the new algorithms, monitor the technical development from a clinical perspective and, crucially, to evaluate the performance of the MammoGrid in the context of proof-of-concept tests. The clinical teams will evaluate the clinical feasibility of the system in the two applications developed during this project: as a data-mining tool for epidemiology research in breast cancer, and as QC device in breast screening programmes. The overall clinical applicability of the project will be assessed according to the results of these two exercises.
Funding SchemeCSC - Cost-sharing contracts
OX1 2EP Oxford, Oxfordshire
CB2 1TN Cambridge
OX1 2JD Oxford
BS16 1QY Bristol