Lecture 5b

69
 Volume Visualization

description

MRI medical imaging

Transcript of Lecture 5b

  • Volume Visualization

  • Outline Introduction

    Common Volume Visualization Steps

    Approaches to Volume Visualization

    Marching Cubes Algorithm

    Ray-casting Algorithm

    GPU based Volume Rendering

  • Volume Visualization - Examples

    MRI Face Mask (1994) CVH Visualization

  • Volume Visualization

    Volumetric data are very common.

    3D Volume Data (Voxels) 2D Image Plane (Pixels)

  • Sources of Medical Volume Data Medical Data

    CT (Computed Tomography)

    MRI (Magnetic Resonance Imaging)

    PET (Positron Emission Tomography)

    Angiography Data

    3D US Data

    CT Angiography:Dept. of NeuroradiologyUniversity of Erlangen,Germany

  • Other Special Volume Data

    Visible Human Project

    CT Human Head:Visible Human Project,US National Library of Medicine, Maryland, USA

    Video Show

  • Volume Visualization Terms

    Space/Grid/Lattice

    Grid Traversal

    Voxels

    Cells

    Geometric Primitives

    Extent Planes

  • Grid Types

    Uniform grid (voxels)

    Reconstruction with

    trilinear interpolation

    Unstructured grid

    Decomposed into

    tetrahedra

    Reconstruction with

    linear interpolation

    REAL-TIME VOLUME GRAPHICS

    Daniel Weiskopf

    Institute of Visualization and Interactive Systems, University of Stuttgrat, Germany

  • Different number can change color

  • Common Vol. Vis. Steps Data Acquistion

    Slice Processing

    Volume Reconstruction

    Volume Enhancement

    Data Classification or Thresholding

    Mapping to Primitives

    Shading & Transforming Primitives

    Displaying Primitives

  • Voxelization

    three dimensional scan conversion

    the process of converting a geometric

    representation of a synthetic model into a set

    of voxels that best represents that synthetic

    model within the discrete voxel space.

  • Common Approaches to

    Volume Visualization

    Surface Rendering

    an indirect technique used for visualizating

    volume primitives by first converting them

    into an intermediate surface representation and

    then employing conventional computer

    graphics techniques to render them to the

    screen, e.g. Marching Cube Algorithm.

  • Common approaches ...

    Direct Volume Rendering

    These techniques deal directly with volume

    primitives without any intermediate conversion

    of the volume data to surface representation,

    e.g. ray tracing of volume data, 3D texture

    mapping rendering

  • Surface Rendering Techniques

    Marching Cubes Algorithm

    creates triangles

    floating point representation

    use case table to create triangles

    use general graphics hardware for rendering

    can generate large number of triangles for

    medical data, e.g. 0.5 million triangles for a

    human skull.

    After break

  • Marching CubesAlgorithm Summary

    Create a cube

    Classify each vertex

    Build an index

    Get edge list

    Interpolate triangle vertices

    Calculate and interpolate normals

    References:

    1. http://en.wikipedia.org/wiki/Marching_cubes

    2. William E. Lorensen, Harvey E. Cline: Marching Cubes: A high resolution 3D surface

    construction algorithm. In: Computer Graphics, Vol. 21, Nr. 4, July 1987

  • Results (MC vs ASC)

    MC ASC, N = 1

  • Results (MC vs ASC)

    MC ASC, N = 2

  • Results (MC vs ASC)

    MC ASC, N = 4

  • Results (MC vs ASC)

    MC ASC, N = 8

  • Marching Cubes

  • Marching Cubes

  • Marching Cubes

  • Marching Cubes

  • Marching Cubes

  • Marching Cubes

  • Marching Cubes

    Ratio

  • Marching Cubes

  • Marching Cubes

    Problem - Ambiguous cases can result in holes

  • Marching Cubes

    Inconsistent Choice

  • Marching Cubes

    Results in holes

  • Direct Volume Rendering

    Ray-casting Algorithm

    3D texture mapping for volume rendering

    hardware dependent (in Part II)

  • Data Classification

    Surface rendering

    User picks threshold value

    Direct volume rendering

    User specifies color table map data values to

    meaningful colors

    User specifies opacity table map interesting

    data to opaque, map other data to transparent

  • Ray-Casting Methods

    Cast Rays from image plane through

    volume to find pixel color

    No intermediate surfaces

    renders volume directly

    References:

    1. http://en.wikipedia.org/wiki/Volume_ray_casting

    2. Marc Levoy, Efficient Ray Tracing of Volume Data, ACM

    Transactions on Graphics, 9(3):245-261, July 1990.

  • Less sampling

    More sample

  • Can be less expensive

  • Compositing

    Compositing describes the way that the

    individual contributions from the sample

    points are accumulated

    61

  • Compositing

    Back to front compositing

    the contribution at sample position k is computed

    by the previous contribution weighted by the

    transparency at the current sample, plus the color

    at the current sample

    62

  • Compositing

    Front to back compositing

    Compose samples from the front of the volume

    dataset along the ray to the back end of the dataset

    63

  • Compositing

    Compositing variationsFirst Hit

    sample the casted rays until we find two sample points below and above acertain intensity threshold

    Pseudo X-ray Traverse and accumulate sample values along the rays throughout the

    whole volume

    Threshold sensitive compositing Similar to pseudo X-ray, but considering only sample values above a

    certain intensity threshold

    Maximum intensity projection (MIP) Search for the sample point with the highest intensity value

    Closest vessel projection (CVP) Take the first sample with a local maximum that surpasses a specified

    threshold

    64

  • Compositing

    Left: pseudo X-ray. Right: threshold-sensitive

    compositing

    65

  • Compositing

    Left: maximum intensity projection. Right:

    closest vessel projection

    66

  • Compositing

    Thin slab volume rendering

    Consider only a small number of slices from

    the full volume dataset

    Allow a better representation of the spatial

    coherence between individual slices

    67

  • Compositing

    Slab volume rendering of about 10 cm of CT

    thorax data

    68

  • Compositing

    Left: single slice representation. Right: thin

    slab maximum intensity projection

    69