Data Structures and Workflows for ICME
25
flexible image processing techniques applicable to n-D images, including robust
approaches for multimodal image registration. ITK utilizes a pipeline construct
to build workflows for image processing problems and is particularly suited to
processing 3D medical imaging modalities, such as computed tomography and
magnetic resonance imaging. ITK on its own is a pure library; the open-source
3D Slicer application provides a graphical front end to many ITK functionalities,
including registration, with capabilities for 3D visualization and volume rendering
[40, 41]. 3D Slicer leverages the Visualization Toolkit (VTK), which provides
a platform-agnostic rendering engine along with a wide variety of geometric
processing tools, such as connectivity, smoothing, and mesh fairing [42, 43].
Another open-source software tool for analyzing biomedical information is
SCIRun, supported by the Center for Integrative Biomedical Computing [44, 45].
SCIRun provides a graphical programming interface for building simulation and
analysis workflows tailored to biomedical data, with a focus on bioelectric fields.
This visual programming approach is similar to the interface paradigm adopted
by DREAM.3D. For application-agnostic data mining and machine learning tasks,
the open-source Orange application provides a visual programming front end built
on top of Python’s rich set of available analytics libraries [46, 47]. A tool with
similar capabilities to Orange is the open-source Java-based Waikato Environment
for Knowledge Analysis (Weka) [48, 49]. Weka provides several machine learning
functionalities, and it also provides plugin support for data-driven image segmentation in Fiji.
The above examples motivate a more generic need for extensible data processing
and handling in scientific analysis. As the materials community begins to broadly
adopt the ICME paradigm, it is prudent to take advantage of the strides made in other
fields in implementing workflow tools, particularly in medical and biomedical image
analysis. Leveraging the lessons learned from these previous tools can accelerate the
development of materials applications, allowing for the allocation of development
resources toward addressing fundamental materials data problems that are not
shared in other fields.
4 Building an Extensible ICME Data Schema and Workflow
Tool
We now consider the critical aspects that define a successful ICME workflow
tool: a scalable, efficient data structure; modularity and plug-and-play capability
for building workflows; and standardized data access and metadata labeling. The
primary interest is for processing data that is accessible via a spatiotemporal index.
We refer to this sort of data in general as field data. These kind of data are
naturally generated by many types of materials characterization and simulation
approaches. Note that we do not directly consider scalar material properties,
such as thermophysical constants. While these constants are integrally important
25
flexible image processing techniques applicable to n-D images, including robust
approaches for multimodal image registration. ITK utilizes a pipeline construct
to build workflows for image processing problems and is particularly suited to
processing 3D medical imaging modalities, such as computed tomography and
magnetic resonance imaging. ITK on its own is a pure library; the open-source
3D Slicer application provides a graphical front end to many ITK functionalities,
including registration, with capabilities for 3D visualization and volume rendering
[40, 41]. 3D Slicer leverages the Visualization Toolkit (VTK), which provides
a platform-agnostic rendering engine along with a wide variety of geometric
processing tools, such as connectivity, smoothing, and mesh fairing [42, 43].
Another open-source software tool for analyzing biomedical information is
SCIRun, supported by the Center for Integrative Biomedical Computing [44, 45].
SCIRun provides a graphical programming interface for building simulation and
analysis workflows tailored to biomedical data, with a focus on bioelectric fields.
This visual programming approach is similar to the interface paradigm adopted
by DREAM.3D. For application-agnostic data mining and machine learning tasks,
the open-source Orange application provides a visual programming front end built
on top of Python’s rich set of available analytics libraries [46, 47]. A tool with
similar capabilities to Orange is the open-source Java-based Waikato Environment
for Knowledge Analysis (Weka) [48, 49]. Weka provides several machine learning
functionalities, and it also provides plugin support for data-driven image segmentation in Fiji.
The above examples motivate a more generic need for extensible data processing
and handling in scientific analysis. As the materials community begins to broadly
adopt the ICME paradigm, it is prudent to take advantage of the strides made in other
fields in implementing workflow tools, particularly in medical and biomedical image
analysis. Leveraging the lessons learned from these previous tools can accelerate the
development of materials applications, allowing for the allocation of development
resources toward addressing fundamental materials data problems that are not
shared in other fields.
4 Building an Extensible ICME Data Schema and Workflow
Tool
We now consider the critical aspects that define a successful ICME workflow
tool: a scalable, efficient data structure; modularity and plug-and-play capability
for building workflows; and standardized data access and metadata labeling. The
primary interest is for processing data that is accessible via a spatiotemporal index.
We refer to this sort of data in general as field data. These kind of data are
naturally generated by many types of materials characterization and simulation
approaches. Note that we do not directly consider scalar material properties,
such as thermophysical constants. While these constants are integrally important