MeVisLab

1
MIP Prototyping

2
MeVisLab
 http://www.mevislab.de/
 In more than 20 years of development, MeVisLab has become one of the
most powerful development platforms for medical image computing
research.
 image processing, visualization and interaction modules can be combined to
complex image processing networks using a graphical programming
approach
 can easily be integrated using a modular, platform independent C++ class
library.
 JavaScript or Python components can be added to implement dynamic
functionality on both the network and the user interface level.
 based on the Qt application framework and the OpenInventor 3D
visualization toolkit
 ITK and VTK AddOns

3
 Rapid Application Prototyping Environment
 Cross-platform (Windows, Mac OS X, Linux)
 Free for non-commercial usage
 Supported file formats
 DICOM, TIFF, DICOM/TIFF, RAW, LUMISYS, PNM, Analyze,
PNG, JPEG
 Currently 920+ Standard modules in the MeVisLab SDK
core, 3000+ modules delivered in total
 with 360+ ITK modules, 1470+ VTK modules, and 300+
modules in the Fraunhofer MEVIS release

4
MeVisLab development
 Three levels
 Visual level
 Programming with “plug and play”
 Individual image processing, visualization and interaction modules can be
combined to complex image processing networks using a graphical
programming approach.
 Scripting level
 Creating macro modules and applications based on macro modules
 Python scripting components can be added to implement dynamic
functionality on both the network and the user interface level.
 C++ level
 Programming modules
 New algorithms can easily be integrated using the modular, platform-
independent C++ class library.

5
Image Processing
 Filters
 Diffusion filters, morphology filters, kernel filters, Hessian, and
vesselness filters
 Segmentation
 Region growing, live wire, fuzzy connectedness, threshold, manual
contours
 Transformations
 Affine transformations, distance transformations, projection and
Radon transforms, manual registration
 Statistics
 Histograms, global image statistics, box counting dimension
 Other
 Unary/binary arithmetic, resampling/reformatting, dynamic data
analysis, noise/test pattern generators

6
Modules for Visualization
 MeVisLab provides modules for visualizing image data and
other data objects in 2D and 3D.
 A set of lookup table (LUT) modules allows applying basic
window/level adjustment or flexible color encoding
schemes.
 The visualization functionality in MeVisLab is based on the
well-established visualization and interaction
library Open Inventor.

7
High-quality Volume Renderer:
MeVisLab Giga Voxel Renderer
 MeVisLab features a high-quality volume renderer that is
based on OpenGL and its extensions.
 It supports the rendering of large volume datasets, even if
they do not fit into the main memory.
 An optimized, multi-resolution technique based on an
octree representation and 3D textures adaptively selects
the best resolution depending on camera position, volume
of interest, and available resources.

8
MeVisLab Software Development Kit
(SDK)
 Using the MeVisLab Software Development Kit
(SDK), a developer is able to implement and test own
algorithms, visualization or interaction methods, or even
complete processing workflows.
 The MeVisLab SDK offers a variety of features that
support module programming, scripting, and network
development.

9
Open Inventor
 An object-oriented 3D toolkit developed by Silicon
Graphics (SGI)
 offering a comprehensive solution to interactive graphics
programming problems
 Most of the visualization modules of MeVisLab make use
of Open Inventor.

10
Open Inventor (OIV)
 Direct Open Inventor node support
 Open Inventor:
 Scene graph paradigm
 Object, rendering, transformation, property, … nodes
 Based on OpenGL
 Extensions to support 2D image viewing/manipulation
 Mixed ML/Open Inventor Modules
 http://www.mevislab.de/mevislab/features/open-inventor/

11
Open Inventor Scene Graph
 Scene objects are represented by nodes
 Size and position is defined by transformation nodes
 A rendering node represents the root of the scene graph

12
Integration of Visualization,
Segmentation and Registration Toolkits
 The Insight Segmentation and Registration Toolkit (ITK) is
an extensive collection of leading-edge algorithms for
registration, segmentation, and analysis of
multidimensional data.
 It is an open-source, cross-platform software package written
in C++ and supported by the US National Library of Medicine.
 The Visualization ToolKit (VTK) is an open source, freely
available software library for 3D computer graphics,
image processing, and visualization.
 It has become one of the most popular open source toolkits
for visualization purposes and is used by thousands of
researchers and developers around the world.

13
MeVisLab User Interface

14
MeVisLab Modules

15
Image Processing Pipeline

16
 Connectors

 Connections

17
Network Layout

18
Network Quick Search

19
Using Groups

20
Using Notes

21
Scripting (MDL)
 User interfaces are created with the “Module Definition
Language” (MDL)
 Abstract hierarchical GUI language
 Interpreted at run-time, allows rapid prototyping
 www.mevislab.de/fileadmin/docs/html/mdl/

22
Getting Started: Chapter 11. GUI Design
in MeVisLab

23
Application Prototyping
 Hide network complexity
 Design user interfaces
 Scripting for dynamic components

24
View2D Module

25
View3D Module

26
Implementing the Contour Filter

27
28
Creating a New Group

29
Parameter Connection for
Synchronization

30
Getting Started: Chapter 5. Defining
a Region of Interest
 a network that allows defining a 2D region of interest
(ROI), that is by selecting a region of the image in the first
viewer, the selected region is displayed as a subimage in a
second viewer

31
Getting Started: Chapter 7. Creating
an Open Inventor Scene
 a dynamically definable applicator (needle for minimally
invasive surgeries) shall be placed at a position and an
angle relative to the rendering of an anatomical image

32
2D Viewers
 Modular 2D Viewer Library (SoView2D)
 Hardware accelerated using textures and shaders
 Supports interactive LUT even on large images
 Extension mechanism supports:
 Overlays
 Markers
 ROIs
 Contours
 User extensions can add drawing
and event handling

33
Winged Edge Mesh Library (WEM)
 Data structure proposed by Baumgart, 1975
 Mesh consists of Nodes, Edges and Faces
 Dense pointer structure of incident primitives
 Fast access to neighboring structures

Pointer links in a neighborhood

34
WEM Modules Overview
 Generation:
 WEMIsoSurface
 Processing:
 WEMCollapseEdges
 WEMSmooth
 WEMPurge
 WEMClip
 …
 Rendering:
 SoWEMRenderer
 Different Render Modes
 Optional Coloring by LUT Values

35
 WEM Sceneshots

Network with iso surface A liver surface colored by a LUT

36
Winged Edge Mesh IsoSurface
 Four subnetworks, each showing different features of the
WEMIsoSurface

37
Contour Segmentation Objects (CSO)
 CSO library
 provides data structures and modules for interactive or
automatic generation of contours in voxel images
 Contours can be analyzed, maintained, grouped and
converted back into a voxel image

 CSO consists of a number of seed points and a number

38
CSO Modules Overview
 Generation (without interaction):
 CSOIsoGenerator
 Processing (with interaction):
 CSOFreehandProcessor, CSOLiveWireProcessor,
CSOIsoProcessor, CSOBulgeProcessor, …
 Rendering
 SoView2DCSOEditor, SoCSO3DVis
 Misc
 CSOConvertToImage, CSOConvertTo3DMask, CSOFilter,
CSOManager, CSOLoad / CSOSave, …

39
SoView2DCSOEditor Example Network

40
SoView2DCSOExtensibleEditor
Example Network

41
SoCSO3DVis Example Network

42
3D

43
44
DICOM Support
 Import of 2D/3D/4D DICOM datasets
 MeVisLab DICOM Service runs as Windows Service or
UNIX Daemon and receives data from PACS even when
user is logged out
 Export of DICOM slices to disk
 DICOM-Store allows to send data to PACS

45
Fuzzy
 FuzzyCluster
 an implementation of the fuzzy c-means algorithm that
classifies an image into different clusters depending on the gray
values
 FuzzyConnectDistance
 a segmentation algorithm based on Fuzzy Connectedness extended by
the possibility to use a property based on the distance of image
elements to the center of the object to be segmented while
calculating membership values
 FuzzyObjectLabeling

46
FCM

47
48
ITK Wrapper
 ITK – Insight Toolkit (www.itk.org)
 Open Source Library for Medical Image Processing and Registration
 about 200 Modules for Standard Image Processing such as
 Image Arithmetics
 Kernel-based and Diffusion Filtering
 Levelset and Segmentation Filtering
 Warping, Resampling Filters
 about 90 Modules Registration-Related Algorithms
 Interpolators
 Metrics
 Optimizers
 Transformations

49
ITK Book Examples

50
ITK Watershed

51
Example Network of itkWatershedImageFilter

52
53
VTK Wrapper
 VTK – Visualization Toolkit (www.vtk.org)
 Visualization, Image Processing and Filtering Library for images,
meshes, grids, data sets etc.
 about 1000 Modules for
 2D/3D Image Processing
 Grid, Mesh, Surface, and Data Filtering
 Pickers
 Properties and Actors
 Mappers
 Renderers, Widgets,Viewers
 Sources, Readers and Writers
 Transformations

54
VTK Example 1: Contour Filter

55
VTK Example 2: VTK/OIV mix
 SoVTK module allows VTK rendering as part of an Open Inventor
scene graph

56
vtkBoxWidget2 Example Network

57
ITK Image Registration

58
Skeletonization
 Skeletonization of a binary image by successive erosion of
border voxels
 vessel centerline extraction
 in 2d and 3d

59
60