Slicer3:Architecture/Features

Presentations

Presentations:

• Slicer 3 Architecture Presentation 2006-04-03

Goals

The main goals of the slicer3 re-work are:

• Provide a clean, well-defined, flexible and functional application structure for medical image computing developers and users
• Learn from user and developer feedback about earlier versions of slicer as well as other packages with similar goals
• Maximize the re-use of code by making it easy use pieces of slicer in other environments and to plug diverse pieces of code into the slicer model

Architecture

At a high level, these goals are to be implemented through:

• A data-centric approach in which the Data Model (implemented as a follow on to the MRML Tree) is implemented independent of the visualization and algorithmic components of the system
  • The Data Model API allows adding, deleting, reading, and modifying medical image data types (Volumes, Models, Transforms, Fiducials, etc).
  • The Data Model provides functionality to serialize and deserialize the contents in an XML structure
  • The Data Model API is structured to allow the either direct access (in the same process space with the algorithms) or through a client/server connection.
  • A set of Adaptor Classes will be provided to allow access to the Data Model API through ITK Image Readers/Writers, VTK Image Readers/Writers, and command line utility commands. These adaptors will support access to the data by programs that have not specifically been written to support the slicer Data Model.

• An Execution Model in which algorithmic functionality is encapsulated such that they can be used in one of several modes:
  • as stand-alone command line executables (for use in testing or to be called from batch scripts; these executables are also important to enable grid/distributed computing)
  • as shared libraries linked into a parent application (i.e. brought into the slicer application address space where the classes can be instantiated and methods invoked with data passed as pointers)
  • a framework to enable algorithm developers to efficiently code these execution modules will be provided using ITK

• A UI application interface that
  • presents a consistent and modern look and feel to end users
  • a 'thin' application layer that is independent of the data and execution implementation
  • a UI description language (XML) so that execution modules can be 'self describing' in terms of their parameters so that, for example, the UI elements can be provided through a --xml-help command line argument
  • a class hierarchy of reusable UI Widgets that enforce consistent look and feel and can be used in multiple modules
  • UI widgets that are 'aware' of the Data Model, UI, and other slicer3 elements to support high level operations (e.g. a coordinate system browser, a volume properties editor, a view controller, etc).
  • A 'trace' capability in the UI to support scripted tests and demos (possibly also user macros).

• A Visualization Environment customized for medical image processing (VTK based)
  • A framework for representing and displaying elements of the Data Model with sub-voxel accurate rendering
  • Ability to represent and display composite scenes with volumes, models, landmarks, and fiducials
  • Ability for modules to add transient display elements (e.g. glyphs and widgets) to support interaction
  • Cross platform high performance rendering infrastructure
  • a framework to support module development

• Other global features and requirements
  • Cross platform to at least: Windows, Linux, Mac. Solaris, SGI, AIX desirable but not critical.
  • As much code coverage as possible in nightly testing
  • Integrated with CPack for distribution and installation.

Features & capabilities "wish-list"

See Feature and Resource Requests to make a new Slicer 3 feature or resource request or to browse existing ones.