12 Segmentation of Symptomatic Exudate-Associated …
347
solve SEAD segmentation in 3-D OCT [24]. By introducing both a boundary term
and a regional term into the energy function, the method computed a minimum cost
s/t cut on an appropriately constructed graph [17, 18]. For multiple object-region
segmentation, an interaction term can be introduced to the energy function as a hard
geometric constraint [11]. The overall problem can also be solved by computing an s/t
cut with a maximum-flow algorithm. The conventional graph-cut framework can be
applied to objects with different topological shapes, but it cannot avoid segmentation
leaks in lower resolution images.
12.2.2 Optimal Surface Approach—Graph-Search Approach
Optimal surface approach (GS methods) [12, 25–27] is another graph based method
which is important for the analysis of multiple intra-retinal layers in 3-D OCT images
[28, 29]. Take SEAD cases as example, most of the subretinal fluid, lesions intraretinal fluid and the pigment epithelial detachments are all associated with surrounding retinal layers. The GS methods modeled the boundaries between layers as terrainlike surfaces and suggested representing the terrain-like surface as a related closed
set. GS methods segment the terrain-like surface by finding an optimal closed set. For
the multiple-surface case, the optimal surface approach constructed a corresponding
subgraph for each terrain-like surface [25], and added weighted inter-graph arcs,
which enforced geometry constraints between subgraphs. The multiple optimal surfaces segmentation could be solved simultaneously as a single s/t cut problem by
using a maximum-flow algorithm. The method worked well in finding stable results
of globally optimal terrain-like surfaces. However, it was limited by the prior shape
requirement. For the multiple-SEAD in a single OCT image, it can be modeled as a
problem with multiple regions interacting with multiple surfaces. A surface-region
graph-based method was proposed to segment multiple regions and multiple surfaces
simultaneously [30].
12.3 Probability-Constrained Graph Search-Graph Cut
The graph search-graph cut method consists of two main steps: initialization and
segmentation (Fig. 12.2). In the initialization step, preprocessing steps are applied
first to the input OCT image. The preprocessing steps include: segmenting the layers,
fitting a surface to the bottom [retinal pigment epithelium (RPE)] layer, determining
SEAD footprints [31], ignoring points within the SEAD footprints, and flattening
the scan images; a texture classification based method is employed producing the
initialization results. Following initialization step, probability normalization refines
the initialization results. In the segmentation step, the GS-GC method synergistically
integrates the results from the initialization.
347
solve SEAD segmentation in 3-D OCT [24]. By introducing both a boundary term
and a regional term into the energy function, the method computed a minimum cost
s/t cut on an appropriately constructed graph [17, 18]. For multiple object-region
segmentation, an interaction term can be introduced to the energy function as a hard
geometric constraint [11]. The overall problem can also be solved by computing an s/t
cut with a maximum-flow algorithm. The conventional graph-cut framework can be
applied to objects with different topological shapes, but it cannot avoid segmentation
leaks in lower resolution images.
12.2.2 Optimal Surface Approach—Graph-Search Approach
Optimal surface approach (GS methods) [12, 25–27] is another graph based method
which is important for the analysis of multiple intra-retinal layers in 3-D OCT images
[28, 29]. Take SEAD cases as example, most of the subretinal fluid, lesions intraretinal fluid and the pigment epithelial detachments are all associated with surrounding retinal layers. The GS methods modeled the boundaries between layers as terrainlike surfaces and suggested representing the terrain-like surface as a related closed
set. GS methods segment the terrain-like surface by finding an optimal closed set. For
the multiple-surface case, the optimal surface approach constructed a corresponding
subgraph for each terrain-like surface [25], and added weighted inter-graph arcs,
which enforced geometry constraints between subgraphs. The multiple optimal surfaces segmentation could be solved simultaneously as a single s/t cut problem by
using a maximum-flow algorithm. The method worked well in finding stable results
of globally optimal terrain-like surfaces. However, it was limited by the prior shape
requirement. For the multiple-SEAD in a single OCT image, it can be modeled as a
problem with multiple regions interacting with multiple surfaces. A surface-region
graph-based method was proposed to segment multiple regions and multiple surfaces
simultaneously [30].
12.3 Probability-Constrained Graph Search-Graph Cut
The graph search-graph cut method consists of two main steps: initialization and
segmentation (Fig. 12.2). In the initialization step, preprocessing steps are applied
first to the input OCT image. The preprocessing steps include: segmenting the layers,
fitting a surface to the bottom [retinal pigment epithelium (RPE)] layer, determining
SEAD footprints [31], ignoring points within the SEAD footprints, and flattening
the scan images; a texture classification based method is employed producing the
initialization results. Following initialization step, probability normalization refines
the initialization results. In the segmentation step, the GS-GC method synergistically
integrates the results from the initialization.