PRINCIPLES AND APPROACHES 3D Medical Imaging
PRINCIPLES AND APPROACHES
3D Medical Imaging
Introduction (I) – Purpose and Sources of Medical Imaging
Purpose Given a set of multidimensional images, output qualitative /
quantitative information about the object/object system under study in these images.
Sources of Images 2D: digital radiography
Computerized tomography (CT) Magnetic resonance imaging (MRI) Positron emission tomography (PET) Single Photon emission computed tomography (SPECT) Ultrasound (US) Functional MRI (fMRI)
3D: A time sequence of 2D images or a volume of tomography 4D: A sequence of 3D images of a dynamic object 5D and up Among tomographic modalities, CT, MRI, and US provide
structural/anatomical information; PET, SPECT, and fMRI as well as doppler US provide functional information
2
Introduction (II) – Objects and Classification of Study
Objects of Study Rigid (e.g., bones) vs. deformable (e.g., soft-tissue structures) Static (e.g., skull) vs. dynamic (e.g., heart, joints) Mixed characteristics, such as MRI 3D study of the head:
white matter, gray matter, and cerebrospinal fluid Qualitative (e.g., visually) vs. quantitative information (e.g.,
statistically)Classification
Operations: preprocessing, visualization, manipulation, analysis
Viewing medium: computer monitor, holography, head-mounted display
Systems physician display console (by imaging device vendors) Image processing/visualization workstations supplied by
workstation vendors 3D imaging software (commercial products) University-based 3D imaging software (often freely available)
3
Introduction (III) – Schematic Representation of 3D Imaging Systems
4
Introduction (IV) – Basics and Terminology
5
Introduction (V) – Basics and Terminology
Object, Object system (a collection of objects)Body regionImaging devicePixel, voxelScene, scene domain, intensity, binary sceneK-th slice, pixel size, slice thickness, slice
location, slice spacingStructure, structure systemRendition of a scene/structure/structure systemCoordinate systems: imaging device, scene,
structure, display (viewing)
6
Introduction (VI) – Object Characteristics
Graded composition Voxels constituting the femur have a
gradation of density values; however, they “hang together” to form the femur
Hanging-Togetherness (Gestalt) A configuration, pattern, or organized
field having specific properties that cannot be derived from the summation of its component parts; a unified whole
7
Preprocessing – ROI/VOI
Region of Interest (ROI)/Volume of Interest (VOI)A sub-scene with reduced sized of the scene domain
and/or the intensityROI/VOI operations may
Specify a rectangle/rectangular volume, or Drawing and painting, or Specify ROI/VOI loosely, indicate a region containing ROI but
exclude unwanted regions with similar property[Figure; from left to right, top to bottom]
A region of interest specified by a rectangular box in the scene (a); the output is shown in (b); region of arbitrary shape by drawing (c) and painting (d)
8
Preprocessing – Filtering (Enhancing)
Filtering operations convert a given scene into another scene to enhance wanted (object) information and to suppress unwanted (noise, background) information Edge Enhancing
Gradient
9
8-connectivity in 2D
6-connectivity in 3D
432
443322
876
887766
654
665544
812
88112201
)()()()()()(
)()()()()()()(
www
vfwvfwvfw
www
vfwvfwvfw
www
vfwvfwvfw
www
vfwvfwvfwvf
321
56324231102
)()()()()()()(
www
vfvfwvfvfwvfvfwvf
voxel.a ofintensity thegives (.)
ly.respective weights,are
f
wi
))()(),()(),()(()( 5624310 vfvfvfvfvfvfvf
Preprocessing – Edge Enhancing
A slice of a 3D MR scene ofa patient’s head (a) and itsedge-enhancing filtered output with a 2D (b) and a 3D neighborhood (c).
10
Preprocessing – Filtering (Suppressing)
Edge suppressing, smoothing, or averaging – low-pass filtering
Illustration of a smoothing 2D GaussianFilter (b), a 3D Gaussian filter (c), anda median filter (d) for the scene in (a)
Interpolation Scene-Based Interpolation Methods Object-Based Interpolation Methods
11
n
jj
n
jjj
w
vfw
vf
0
00
)(
)(
Preprocessing – Diffusion
Intensity gradients in a given scene are considered to cause a “flow” within the scene whose functional dependence on gradient is controlled through a parameter K.
12
Preprocessing
Registration Scene-Based Registration Methods
Rigid Deformable
Object-Based Registration Methods Rigid Deformable
13
Preprocessing
Segmentation Hard, Boundary-Based, Automatic Methods
Iso-Surfacing Methods Gradient-Based Methods
Fuzzy, Boundary-Based, Automatic Methods Hard, Boundary-Based, Assisted Methods
Active Contours Live Wire/Lane
Hard, Region-Based, Automatic Methods Thresholding Clustering
Fuzzy, Region-Based, Automatic Methods Hard, Region-Based, Assisted Methods Fuzzy, Region-Based, Assisted Methods
14
Visualization
Scene-Based Visualization Methods Slice Mode Volume Mode
Maximum Intensity Projection (MIP) Surface Rendering Volume Rendering
Object-Based Visualization Methods Maximum Intensity Projection Surface Rendering Volume Rendering
Misconceptions and Challenges in Visualization
15
Further Topics
Manipulation Rigid model Deformable model
Analysis Scene-Based Object-Based
Sources of difficulty in 3D imaging Qualitative validation Quantitative validation
16