Model Based Segmentation Flashcards
1
Q
Model based segmentation
A
- incorporate prior knowledge about structure of interest into segmentation process
- active contours and active surface approaches
2
Q
Active contours
A
- snakes: deformable splines influenced by constraints and image forces
- internal Energy -> min (enforces smoothness via sum of elastic and bending energy
- external Energy: represents agreement of contour with the image data
- MINIMIZATION OF ENERGY
- Problems: result depending on number and spacing of control points; boundaries often oversmoothed; good init crutial; self-intersections
- topology adaptation
3
Q
Level Sets
A
- LS function describes an interface between points x with f(x,t)>k and f(x,t)
- interface evolved based on intrinsic properties (boundary smoothness), extrinsic properties (image based i.e. gradient), domain knowledge
- f(x,t) = 0 is boundary
4
Q
types of level sets
A
- stationary level set: wave propagation from C(t), movement in one direction, result is set of points reached at time t
- dynamic level set: interface can change direction and cross points several times
5
Q
Fast marching
A
- increase of efficiency through restriction of updates to voxel in neighborhood of voxels with known arrival times
- shortest arrivel time - shortest path problem
- algorithm: move trial node with shortest arrival time to set of know nodes; move neighbors of trial node from far set to trial set; calculate arrival times for new nodes in trial set
6
Q
When use active contours and surfaces
A
- if parameterizable model of segmentation can be generated
- if data alone does not allow proper outlining of object of interest
7
Q
Shape Models
A
- advanced representation of characteristic shape and location of anatomical structures
- learn normal vs. desease-related shape structures
- assesment of variation from normal via correlation measure
8
Q
Point Distribution Model
A
- shape instances represented via feature points
-
