A Grid Service for Medical Image Reconstruction

   Siegfried Benkner, Gerhard Engelbrecht, Rainer Schmidt
   Institute for Software Science
   University of Vienna

The European GEMSS Project is concerned with the creation of medical Grid
service prototypes and their evaluation in a secure service-oriented 
infrastructure for distributed on demand/supercomputing.
One of the applications considered in GEMSS is medical image reconstruction 
for single photon emission computed tomography (SPECT), which is an effective 
and robust method for diagnosis of cancer, heart diseases and functional 
pathologies of the human body. 
Modern fully 3D iterative reconstruction algorithms provide very accurate
image reconstruction by considering principal 3D effects of data acquisition,
as photon scatter and collimator geometry. The high computational requirements 
of these algorithms motivated the development of a Grid service that can make 
use of powerful machines available on demand in a Grid environment. The GEMSS 
image reconstruction service comprises a graphical user interface (for image
acquisition, visualization and for interacting with the Grid environment) and 
a reconstruction service which exposes the parallel reconstruction kernel (MPI) 
as a Web service accessible securely via SOAP.

A typical use case scenario for reconstructing 3D images from 2D projection data
comprises the following actions:

  1. Authentication (once)
  2. Acquisition of 2D projection data (e.g. from a SPECT scanner)
  3. Specification of reconstruction meta-data (region of interest, required
     accuracy, etc.) 
  4. Specification of QoS requirements (e.g. service execution time) (optional).
  5. Service selection (optional or done automatically)
  6. Upload of 2D input data (approx. 4 MBs)
  7. Start of reconstruction
  8. Querying status information (optional)
  9. Download of results
 10. Visualization of reconstructed 3D image

If the user is not satisfied with the results, he may initiate another
reconstruction with different parameters (meta-data) by starting with step 3. 
Since in this case there is no need for uploading the 2D projection data, 
steps 5 and 6 can be omitted.

Within GEMSS, a framework for transforming native applications into Web services
is being developed and utilized. The GEMSS middleware makes use of an XML 
application descriptor in order to expose an application automatically as a 
Web service based on Apache AXIS and Tomcat. Key issues of the GEMSS
middleware are the provision of advanced security mechanisms to ensure
confidentiality of patient data as well as compliance to EU law, and negotiable 
QoS support (e.g. for providing guarantees on service execution time).