Exploration of Dynamic Medical
Volume Data
Outline
1.
2.
3.
4.
5.
6.
7.
Introduction
Medical Background
Basic Visualization Techniques
Advanced Visualization Techniques
Case Study: Tumor Perfusion
Case Study: Brain Perfusion
Summary
2
Introduction
Static image data
Only provide a snapshot
Many aspects relevant for diagnostic decisions and
treatment planning cannot be judged by means of a
single snapshot
Introduction
Dynamic image data
Might change over time
Acquired to assess blood flow (perfusion) and
tissue kinetics by tracing the distribution of
contrast agents (CA) or other data changes
A special variant of dynamic data is functional
MRI where activation patterns after stimulation are
recorded
Medical Background
Functional MRI
Where activations of brain areas are imaged
Dynamic SPECT
Where the temporal distribution of a radioactive
tracer is registered
Perfusion data
Have a broad clinical relevance
Dynamic contrast enhanced (DCE) images are
acquired to study these phenomena
5
Medical Background
Parameters
Peak enhancement (PE)
The maximum value (over all points in time)
Time To Peak (TTP)
The point of time where peak enhancement occurs
Integral
The area below the curve is computed
Mean Transit Time (MTT)
MTT specifies the time where the area below the curve
is the same on the left and on the right
6
Basic Visualization Techniques
Techniques to visualize volume data
Cine-movies
Which step through all points in time for a selected slice
Subtraction images
Which depict the intensity difference between two
selected points in time
Color-coded parameter maps for a selected slice
A parameter map is a 2D display of a selected slice, in
which each pixel encodes the value of a selected
parameter
Basic Visualization Techniques
Subtraction
perfusion
images
to
analyze
cerebral
Basic Visualization Techniques
Parametric images for slice 4 of a dynamic
MRI sequence
TTP, MTT, and the integral are depicted as colorcoded images
Advanced Visualization Techniques
Drawbacks of basic visualization techniques
Basic techniques do not permit the integration of
several parameter maps in one image
Dynamic information cannot be integrated with
morphologic information that may be based on
another dataset with higher spatial resolution
10
Advanced Visualization Techniques
Multiparameter visualization
The integrated visualization of several parameters
in a suspicious region is desirable for various
diagnostic tasks
Combining Isolines and Color-coding
Exploration of Multiple Parameter Images with
Lenses
11
Advanced Visualization Techniques
A gray scale MIP of the subtraction volume of
two early points in time is combined with a
color-coded CVP (closest vessel projection)
12
Advanced Visualization Techniques
Combining Isolines and Color-coding
The combination of isolines and colors is
particularly effective and can be easily interpreted
Isolines connect regions where the investigated
dynamic parameter has a certain value
13
Advanced Visualization Techniques
Principle of isoline generation with the
Marching Squares approach
Isolines for isovalue 5 are computed by linear
interpolation along the grid cells
14
Advanced Visualization Techniques
Ten isolines depict a dynamic parameter
derived from MRI mammography
The data and the resulting isolines are smoothed
15
Advanced Visualization Techniques
Exploration of Multiple Parameter Images
with Lenses
Lenses might be employed to show different
information in the lens region
Lenses are useful for showing information relating
to one parameter in the context of a map of another
parameter
16
Advanced Visualization Techniques
Magic Lenses for
multiparameter maps
the
exploration
of
The focus region inside the lens shows the
parameter cerebral blood volume whereas the gray
values show the original MRI data
17
Advanced Visualization Techniques
Exploration of MRI-mammography data with
a lens
18
Advanced Visualization Techniques
Integrating
morphology
dynamic
information
and
It is useful to add spatial reference information in
the regions not containing dynamic information
The integration of dynamic and morphologic
information can be carried out in 2D slice
visualizations or 3D renderings
19
Advanced Visualization Techniques
The visualization of dynamic information is
restricted to the segmented tumor
The surrounding tissue is
conventional volume rendering
displayed
as
20
Case Study: Tumor Perfusion
Tumor perfusion
Perfusion imaging is carried out to evaluate
whether lesions regarded as suspicious in static
images are likely to represent a cancer
Perfusion images support diagnosis of tumor
diseases and therapy monitoring
21
Case Study: Tumor Perfusion
Typical
parameters
mammography
Matrix: 512 512
Slice distance: 2 mm
Number of slices: 6080
Temporal
resolution:
measurements)
for
11.5
DCE
MRI
min
(510
22
Case Study: Tumor Perfusion
Computer support
Software solutions are challenging, due to the
dynamic nature of DCE MRI mammography
Data processing, in particular motion correction, is
more challenging than brain perfusion imaging
23
Case Study: Tumor Perfusion
Visualization techniques
Maximum Intensity Projection
Conventionally used for gray scale volume data in
which the interesting structures have a small volumefilling factor
Closest Vessel Projection
Developed to add depth information to MIP images
Dedicated to the visualization of vascular structures
24
Case Study: Tumor Perfusion
Left: a malignant breast tumor visualized using
a MIP. Right: the same data visualized using a
CVP
25
Case Study: Tumor Perfusion
Left: a lesion is represented as an area of
bright yellow. Right: the graph of the selected
voxel is shown
26
Case Study: Brain Perfusion
Brain perfusion
Ischemic stroke is among the leading causes of
death in all western countries
The identification of tissue at risk (ischemic
penumbra) is crucial before considering any
patient treatment
27
Case Study: Brain Perfusion
Typical parameters for contrast-enhanced MRI
perfusion
Matrix: 128 128
Slice distance: 7 mm
Number of slices: 1015
Temporal resolution: 12
measurements)
seconds
(4080
28
Case Study: Brain Perfusion
CT Imaging
CT perfusion studies only acquire one slice
To reduce image noise, a large slice thickness (10
mm) is employed
Perfusion maps
Brain perfusion maps can be quantified in terms of
absolute blood flow and blood volume
Derived from CT and MRI data
29
Case Study: Brain Perfusion
Visualization techniques
The symmetry of the brain is the basis for
diagnostic evaluation of static and dynamic images
Magic Lenses
Synchronization of ROIs
30
Case Study: Brain Perfusion
Enhancement curves are simultaneously
derived for the symmetric regions in both
hemispheres
31
Case Study: Brain Perfusion
Synchronized lenses in both hemispheres of
the brain support the comparison between the
symmetric regions
32
Summary
Dynamic image data have a great potential for
enhancing diagnosis and therapy monitoring
for important diseases
The acquisition of appropriate data and their
interpretation require long term experience
Focus on the role of visualization to support a
fast and unambiguous interpretation of such
data
33
Clipping, Cutting, and
Virtual Resection
34
Outline
Part 2 start
1. Clipping
2. Virtual Resection
3. Virtual Resection with a Deformable Cutting
Plane
4. Cutting Medical Volume Data
5. Summary
35
Clipping
A fundamental interaction technique for
exploring medical volume data
It is used to restrict the visualization to subvolumes
36
Clipping
The tumor is demonstrated by tilting the clip
plane vertically
37
Clipping
Implementation of clipping
For volume rendering, each voxel affected by the
clipping plane are discarded completely
For surface rendering, each triangle is tested to
determine whether it should be drawn or not
38
Clipping
Volume and surface rendering with a clipping
plane for exploring spatial relations in a CT
head dataset
39
Clipping
Selective clipping
A special variant of clipping
Used to emphasize structures (those not affected
by clipping) while presenting contextual
information (structures which are partially visible
due to clipping)
40
Clipping
Selective clipping
Left: the brain and the ventricles are rendered
completely. Right: the vertical symmetry is used
for selective clipping of a CT head dataset
41
Clipping
Selective clipping with boolean textures
An elegant and efficient way to accomplish
selective clipping is the use of Boolean textures
Boolean textures are constructed by implicit
function, such as quadrics
42
Clipping
Box clipping
Combine six clipping planes to define a subvolume
Useful for exploring a region in detail, for example,
an aneurysm or the region around a tumor
43
Clipping
Box clipping for the analysis of an intracranial
aneurysm
A detailed view of the region of interest is
combined with an overview rendering
44
Clipping
Local volume rendering for the evaluation of
the surrounding of a tumor in CT thorax data
The tumor is visualized as an isosurface whereas
the vascular structures around it are rendered as
direct volume rendering
45
Virtual Resection
Resection refers to the removal of tissue
during a surgical intervention
Virtual resection is a core function of many
intervention planning systems
46
Virtual Resection
Requirements of virtual resection functions
The user must be able to specify a virtual resection
intuitively and precisely
The Modification must be supported to change
virtual resections
Virtual
resections
should
be
visualized
immediately, with high quality
47
Virtual Resection
Specification of virtual resections by erasing
Use scalable 3D shapes as erasers to remove the
touched tissue
Boolean operations on voxel values are used to
decide which subset of voxels should be drawn
the visual quality is limited by the resolution of the
underlying voxel grid
48
Virtual Resection
Left: a resection area specified by erasing
liver tissue with a sphere. Right: the result of
the virtual resection is displayed in a 2D view
49
Virtual Resection
Specification of virtual resections by drawing
on slices
Inspired by the communication between surgeons
and radiologists discussing a resection
The resection is marked by drawing on the slices
with a pen or mouse
This process is time-consuming if the entire
resection volume should be specified
50
Virtual Resection
Virtual liver resection by drawing on the slices
The virtually resected and the remaining portion of
the liver are separated to support the evaluation of
the shape of virtual resections
51
Virtual Resection with a Deformable
Cutting Plane
Based on a surface representation of an organ,
usually achieved with explicit segmentation
The user draws lines on the (3D) surface of an
organ to initialize the cutting plane
The plane is deformed locally to fit the lines
drawn by the user
52
Virtual Resection with a Deformable
Cutting Plane
Defining cutting plane boundaries
The user employs a 2D pointing device and
controls the movement of a cursor on a 3D surface
This control is accomplished by casting a ray from
the viewpoint through the 2D point to the 3D
position on the surface
53
Virtual Resection with a Deformable
Cutting Plane
Definition of the cut path
The Euclidean distance represents the shortest
distance between successive points
The geodesic shortest path connects points on the
3D surface with a path on that surface
54
Virtual Resection with a Deformable
Cutting Plane
Euclidean (left) versus geodesic distance
(right) between surface points
55
Virtual Resection with a Deformable
Cutting Plane
Cut boundary specification specified with a
pen on a digitizer tablet, which is shown
enlarged in the right image
56
Virtual Resection with a Deformable
Cutting Plane
Cut boundary specification with a tactile input
device
57
Virtual Resection with a Deformable
Cutting Plane
Generation of the initial cutting plane
Determine the oriented bounding box of the lines
drawn by the user
Determine the orientation and extent of the cutting
plane
Set the center of the cutting plane
Project the point-set into the cutting plane
Calculate displacements
Smoothing
58
Virtual Resection with a Deformable
Cutting Plane
Definition of the plane E based on the (dashed)
lines P drawn by the user
59
Virtual Resection with a Deformable
Cutting Plane
Modification of virtual resections
The resection can be refined by translating grid
points
The user can define the sphere of influence as well
as the amplitude of the deformation
There is also a facility to translate the whole mesh
60
Virtual Resection with a Deformable
Cutting Plane
Left: fine-tuning of the plane with respect to
blood vessels. Right: the initial cutting plane is
translated with a sphere of influence
61
Virtual Resection with a Deformable
Cutting Plane
Based on the two lines drawn on the object
surface, an initial resection has been specified
that might be refined by the user
62
Virtual Resection with a Deformable
Cutting Plane
The result of a virtual resection by means of a
deformable mesh
63
Virtual Resection with a Deformable
Cutting Plane
Transformation of the resection boundary to a
resection plane
Flat projection
A planar projection of the boundary is deformed to
represent the points specified by a user
Minimal surfaces
They are constructed to exactly match the given
boundary
64
Virtual Resection with a Deformable
Cutting Plane
A flat surface (left) as approximation of the
given boundary compared to a minimal surface
(right) of the same boundary
65
Virtual Resection with a Deformable
Cutting Plane
Application
techniques
areas
of
virtual
resection
Liver surgery
Osteotomy planning
Craniofacial surgery
66
Virtual Resection with a Deformable
Cutting Plane
Conventional osteotomy planning based on a
stereolithographic model (left). Virtual
resection based on a 3D model of the patients
bones (middle, right)
67
Virtual Resection with a Deformable
Cutting Plane
Efficient visualization of virtual resections
Efficiency is an important aspect for virtual
resection and clipping with arbitrary geometries
It is desirable that a high update-rate be achieved
without compromising accuracy
68
Virtual Resection with a Deformable
Cutting Plane
Visualization parameters
The realistic approach (to remove the resection
volume entirely) is only one of several possibilities
The resection volume can be regarded as a new
visualization object that can be flexibly
parameterized
69
Virtual Resection with a Deformable
Cutting Plane
Combination of resection proposals and virtual
resection
A completely different approach to resection
specification is to propose to the surgeon which
part of an organ has to be resected
The resection proposal might be presented as
additional information when the deformable
cutting plane is specified
70
Virtual Resection with a Deformable
Cutting Plane
Combining resection proposals and interactive
resection for liver surgery planning
The resection proposal (dark red) is presented
while the user interactively specifies the resection
region (with the yellow line)
71
Cutting Medical Volume Data
Cutting facilities are important for surgery
simulation
Users move a cutting device through medical
volume data and simulate cutting procedures
Collision detection and tactile feedback are
essential for educational purposes when
prospective surgeons are trained
72
Cutting Medical Volume Data
High-quality representation of cut surfaces
Surgical cutting is a challenging application
because the requirements for accuracy and speed
are high, and arbitrary shapes are involved
If virtual resection is accomplished by a volume
representation, the resolution of the cut surface is
limited to the resolution of the underlying data
73
Cutting Medical Volume Data
Volume cutting
Left: two-dimensional representation of an object
surface. Middle: voxelization of the cutting tool.
Right: the resulting cut surface
74
Cutting Medical Volume Data
Progressive cutting
Left: only the left area has to be considered.
Middle: representation of the new cut surface.
Right: the resulting cut surface
75
Cutting Medical Volume Data
Cutting with a virtual scalpel (left). The result
is shown in high quality (right)
76
Cutting Medical Volume Data
Virtual resection vs. surgery simulation
Virtual resection techniques are intended for
experienced surgeons who are actually planning a
surgical procedure
Virtual resection is not focused on the realistic
simulation of a procedure but on decision support
based on the interaction with the data of a
particular patient
77
Summary
Virtual resection is an essential feature for
surgery planning, particularly for internal
organs, such as the kidney, liver, and pancreas
There are some similarities between virtual
resection and surgery simulation concerning
the representation and visualization of the data
Hardware support for 3D texture-mapping
combined with multi-texturing is essential for
a good performance
78
