Engineering
 The Center’s success depends on how well the techniques developed in Core 1 can be used to solve
the biological problems posed by the four DBPs of Core 3. Core 2 is the link between the innovative
techniques of Core 1 and the biological questions of the Core 3 end-user practitioners. To build this
link, Core 2 will establish software architectures and software processes that will empower the Core 1
algorithm developers to create robust, well-designed software and interfaces. In addition to software
from Core 1, Core 2 will build interfaces to existing software and data sources. The Core 3 end-users
will define the requirements for the applications developed by Core 2.
These Core 2 activities require
innovative solutions from our experienced team of biomedical software engineers and researchers.
Defining Software Architectures and Frameworks
Software architecture defines the overall structure of a system including its partition into subsystems
and their interactions. Frameworks define internal software structures that provide uniform
approaches to complex control sequences and algorithms. The Core 2 architecture will adapt to
existing toolkits, applications, and data access software. New algorithms will be added to existing
toolkits. The architecture will support distributed execution and will scale according to hardware
resources. The new software will include algorithms developed by Core 1 and applications defined by
Core 3. The software architecture will accommodate and interface to the existing software inventory
of NA-MIC members. In addition, Core 2 will define and provide open interfaces that will accept
software and data from resources outside of this alliance.
Creating a Software Process
A well defined process brings discipline and control to software development. Core 2 will define and
provide tools to support a lightweight software engineering process. The process will combine objectoriented
design techniques with the Extreme Programming development methodology. A pragmatic
rather than dogmatic approach will encourage Core 1 algorithm developers to adhere to the process
principles without restricting innovation and productivity. Core 2 will enhance the Core 1 software by
generalizing designs and implementations. Throughout the process, Core 2 will refactor designs to
improve the existing code by altering the internal structure of the software.
Building Software Tools
Software engineering processes can place a large, seemingly bureaucratic burden on software
developers. Although guidelines and common architectures can produce long-term benefits, manual
application of these requirements can cause developers to ignore or bypass the procedures. To
minimize the burden of the software process, Core 2 will develop software tools to automate software
quality assurance, support cross-platform software development, and manage cross-platform
software distribution. These tools will build on existing software that Core 2 has developed under
other funding.
Software and Data Integration
Most of the neuroimaging and neuroinformatics software developments in the past decade have
utilized the functionalities of object-oriented programming languages (C, C++ and Java), scripting
languages (perl and tcl), and a multitude of visualization libraries (VTK and Java 3D). To date, there
is no direct way of integrating, merging or linking neuroimaging tools that are developed under twodifferent languages or using different libraries. Many software and library interfaces have been
developed for transferring, interactive manipulation, automatic processing, tracking, visualization, and
archiving of neuroimaging data. Only a limited suite of tools, however, exists at or within any one
geographic location, research group, or computer architecture. We need to augment the modeling
and software development efforts with new integration initiatives that provide foundations for bridging
the gaps between investigators in relation to hardware, software, and neuroscientific expertise.
The full potential of neuroimaging and neurological clinical data to describe brain structure and
function can be realized only if powerful neuroimaging, neuroengineering, neuroinformatics,
database, and visualization tools are integrated in a functional, extensible and portable graphic
environment. Our goal is to create an infrastructure that will ensure efficient data management,
reliable processing, and interactive data visualization for large-scale brain-mapping and neuroimaging
studies, substantially increasing our ability to efficiently perform large state-of-the-art imaging
investigations.
The Grid
Grid computing is rapidly gaining acceptance as a routine way to transparently access distributed
computational power and data for biological and biomedical informatics applications. These
techniques are being employed in academia as well as industry. Enabling biomedical code to run in
such an environment, however, requires “Grid-enabling” of such code along with the associated data
repositories. This process of bringing software programs and data resources together generally
involves development of a layer of software. These grid-enabling codes and databases will leverage
emerging Grid practices and experience gained within the BIRN Coordinating Center.
Applications
A robust, well-designed application environment, with a logical user interface and a set of
complementary image analysis and data manipulation tools, provides a critical component for the
success of the overall effort. In addition, the end-user application environment must be flexible
enough to support a variety of interaction styles including interactive versus batch processing modes
and support for different operating systems and machine architectures. The application software will
start with existing, proven applications. Core 2 will integrate Core 1 algorithms into the applications. |
