244
F. Shi et al.
fluid in diabetic macular edema (DME) and wet AMD. In these cases, the large
variety of layer morphology and degraded image quality caused by abnormalities
greatly affect the segmentation performance. In the first part of this chapter, we focus
on segmentation for retinas with serous pigment epithelium detachments (PEDs),
which are associated with sub-RPE fluid and RPE deformation. The fully automated,
unsupervised 3-D method [15] integrates layer and region segmentation, and can
segment the retina into 10 layers.
With the help of retinal layer segmentation, quantitative analysis of the layer morphology can be conducted. This is especially useful for retinal disease detection or
progress analysis. Certain retinal diseases affect certain retinal layers. The change of
retina layer may occur at an early stage of the disease, and become more evident with
the progress of disease. Most existing layer analysis focus on simple indices such
as layer thickness [16–20] or intensities [21, 22], but texture analysis can also provide important information. In the second part of this chapter, we describe a method
to quantify the disruption of external limiting membrane (ELM) caused by DME
from SD-OCT [23]. With the method, each A-scan is classified as disrupted or nondisrupted based on 6 texture features. In the third part of this chapter, we describe an
extended method for automatic three-dimensional detection of photoreceptor ellipsoid zone (EZ) disruption caused by trauma [24]. 57 features are extracted, and
feature selection and classification is performed to classify each voxel of the EZ as
disrupted or non-disrupted. Both methods can effectively differentiate patients with
normal ones.
10.2 Segmentation of Retinal Layers with Serous Pigment
Epithelial Detachments
10.2.1 Background
Retinal pigment epithelium detachment (PED) is a symptom associated with many
chorioretinal diseases, such as AMD, polypoidal choroidal vasculopathy, central
serous chorioretinopathy, and uveitis [25, 26]. There are three type of PED, namely
serous, fibrovascular, or drusenoid PED. Study shows that the existence of serous
PED in AMD patients is related to the development of choroidal neovascularization
(CNV), which is a main cause of visual acuity loss [26, 27]. In SD-OCT B-scans,
the profile of serous PED is visualized as a dome-shaped elevation of the bright RPE
layer (Fig. 10.1). Note that the layers around the PED may appear discontinuous,
and thus need to be treated specifically during detection.
Following the definition of Iowa Reference Algorithm [14], we define 11 retinal
surfaces (Fig. 10.2), numbered consecutively from top to bottom. The 10 retinal
layers thus defined are nerve fiber layer (NFL), ganglion cell layer (GCL), inner
plexiform layer (IPL), inner nuclear layer (INL), outer plexiform layer (OPL), outer
nuclear layer and inner segment layer (ONL + ISL), connecting cilia (CL), outer
F. Shi et al.
fluid in diabetic macular edema (DME) and wet AMD. In these cases, the large
variety of layer morphology and degraded image quality caused by abnormalities
greatly affect the segmentation performance. In the first part of this chapter, we focus
on segmentation for retinas with serous pigment epithelium detachments (PEDs),
which are associated with sub-RPE fluid and RPE deformation. The fully automated,
unsupervised 3-D method [15] integrates layer and region segmentation, and can
segment the retina into 10 layers.
With the help of retinal layer segmentation, quantitative analysis of the layer morphology can be conducted. This is especially useful for retinal disease detection or
progress analysis. Certain retinal diseases affect certain retinal layers. The change of
retina layer may occur at an early stage of the disease, and become more evident with
the progress of disease. Most existing layer analysis focus on simple indices such
as layer thickness [16–20] or intensities [21, 22], but texture analysis can also provide important information. In the second part of this chapter, we describe a method
to quantify the disruption of external limiting membrane (ELM) caused by DME
from SD-OCT [23]. With the method, each A-scan is classified as disrupted or nondisrupted based on 6 texture features. In the third part of this chapter, we describe an
extended method for automatic three-dimensional detection of photoreceptor ellipsoid zone (EZ) disruption caused by trauma [24]. 57 features are extracted, and
feature selection and classification is performed to classify each voxel of the EZ as
disrupted or non-disrupted. Both methods can effectively differentiate patients with
normal ones.
10.2 Segmentation of Retinal Layers with Serous Pigment
Epithelial Detachments
10.2.1 Background
Retinal pigment epithelium detachment (PED) is a symptom associated with many
chorioretinal diseases, such as AMD, polypoidal choroidal vasculopathy, central
serous chorioretinopathy, and uveitis [25, 26]. There are three type of PED, namely
serous, fibrovascular, or drusenoid PED. Study shows that the existence of serous
PED in AMD patients is related to the development of choroidal neovascularization
(CNV), which is a main cause of visual acuity loss [26, 27]. In SD-OCT B-scans,
the profile of serous PED is visualized as a dome-shaped elevation of the bright RPE
layer (Fig. 10.1). Note that the layers around the PED may appear discontinuous,
and thus need to be treated specifically during detection.
Following the definition of Iowa Reference Algorithm [14], we define 11 retinal
surfaces (Fig. 10.2), numbered consecutively from top to bottom. The 10 retinal
layers thus defined are nerve fiber layer (NFL), ganglion cell layer (GCL), inner
plexiform layer (IPL), inner nuclear layer (INL), outer plexiform layer (OPL), outer
nuclear layer and inner segment layer (ONL + ISL), connecting cilia (CL), outer