Changes in the choroid have been suggested to be of critical relevance to some eye diseases, e.g., geographic atrophy (GA) of age-related macular degeneration (AMD). Several groups have worked to develop automated choroidal segmentation algorithms in optical coherence tomography (OCT) images. However, the reported algorithms mainly focus on OCT images with early AMD (or eye diseases with mild/moderate retinal damage). GA, an advanced stage of AMD, frequently presents severe retinal damage and those algorithms may be confounded by the zones of increased choroidal reflectivity due to GA damage. The fundus autofluorescence (FAF) signal from GA regions is more distinguishable and detectable than on OCT images. We propose to use a three-dimensional (3-D) graph-based approach with the complementary information from the companion FAF images to handle the 3-D choroidal segmentation difficulty due to GA. We compare the automated approaches with and without using the FAF-derived GA information against the ground truth. With the FAF-derived information, the segmentation performance regarding both choroidal borders is significantly improved (all p < 0.01). The robustness of our segmentation approach may be of great value for future large-scale quantitative and longitudinal studies of the choroid, particularly in the setting of atrophic AMD.
Atrophic age-related macular degeneration (AMD) or geographic atrophy (GA), and atrophic juvenile macular degeneration (JMD) or Stargardt atrophy, have been proven to be the leading cause of blindness respectively in older adults, and in children and young adults. Automated techniques of timely screening and detection of such atrophic diseases would appear to be of critical importance in prevention and early treatment of vision loss. We first developed a deep learning-based automated screening system using the residual networks (ResNet), which can differentiate the eyes with atrophic AMD and JMD from normal eyes on fundus autofluorescene (FAF) images. We further developed another deep learning-based automated system to segment the atrophic AMD and JMD lesions using a fully convolutional neural network - U-Net. Transfer learning based on a pre-trained model was applied for ResNet to facilitate the algorithm training, and excessive data augmentation techniques for both ResNet and U-Net were applied to enhance the algorithm generalization ability. In total, 320 FAF images from normal subjects, 320 with atrophic AMD, and 100 with atrophic JMD were included. The performance of the algorithms were evaluated by comparing with manual gradings by reading center graders. For the screening system, there was no reported algorithm and our algorithm demonstrated a high screening accuracy with 0.98 for atrophic AMD and 0.95 for atrophic JMD. For the segmentation system, our algorithm presented a high overlapping ratio with 0.89 ± 0.06 for atrophic AMD and 0.78 ± 0.17 for atrophic JMD.
Geographic atrophy (GA) is an end-stage manifestation of the advanced age-related macular degeneration (AMD), the leading cause of blindness and visual impairment in developed nations. Techniques to rapidly and precisely detect and quantify GA would appear to be of critical importance in advancing the understanding of its pathogenesis. In this study, we develop an automated supervised classification system using deep convolutional neural networks (CNNs) for segmenting GA in fundus autofluorescene (FAF) images. More specifically, to enhance the contrast of GA relative to the background, we apply the contrast limited adaptive histogram equalization. Blood vessels may cause GA segmentation errors due to similar intensity level to GA. A tensor-voting technique is performed to identify the blood vessels and a vessel inpainting technique is applied to suppress the GA segmentation errors due to the blood vessels. To handle the large variation of GA lesion sizes, three deep CNNs with three varying sized input image patches are applied. Fifty randomly chosen FAF images are obtained from fifty subjects with GA. The algorithm-defined GA regions are compared with manual delineation by a certified grader. A two-fold cross-validation is applied to evaluate the algorithm performance. The mean segmentation accuracy, true positive rate (i.e. sensitivity), true negative rate (i.e. specificity), positive predictive value, false discovery rate, and overlap ratio, between the algorithm- and manually-defined GA regions are 0.97 ± 0.02, 0.89 ± 0.08, 0.98 ± 0.02, 0.87 ± 0.12, 0.13 ± 0.12, and 0.79 ± 0.12 respectively, demonstrating a high level of agreement.
Historically, regular drusen and geographic atrophy (GA) have been recognized as the hallmarks of nonneovascular age-related macular degeneration (AMD). Recent imaging developments have revealed another distinct nonneovascular AMD phenotype, reticular pseudodrusen (RPD). We develop an approach to semiautomatically quantify retinal surfaces associated with various AMD lesions (i.e., regular drusen, RPD, and GA) in spectral domain (SD) optical coherence tomography (OCT) images. More specifically, a graph-based algorithm was used to segment multiple retinal layers in SD-OCT volumes. Varying surface feasibility constraints based on the presegmentation were applied on the double-surface graph search to refine the surface segmentation. The thicknesses of these layers and their correlation with retinal functional measurements, including microperimetry (MP) sensitivity and visual acuity (VA), were investigated. The photoreceptor outer segment layer demonstrated significant thinning with a reduction in MP sensitivity and VA score when atrophic AMD lesions were present. Regular drusen and RPD were separately segmented on SD-OCT images to allow their characteristics and distribution to be studied separately. The mean thickness of regular drusen was found to significantly correlate with the VA score. RPD appeared to be distributed evenly throughout the macula and regular drusen appeared to be more concentrated centrally.
Recently, much attention has been focused on determining the role of the peripapillary choroid - the layer between the outer retinal pigment epithelium (RPE)/Bruchs membrane (BM) and choroid-sclera (C-S) junction, whether primary or secondary in the pathogenesis of glaucoma. However, the automated choroidal segmentation in spectral-domain optical coherence tomography (SD-OCT) images of optic nerve head (ONH) has not been reported probably due to the fact that the presence of the BM opening (BMO, corresponding to the optic disc) can deflect the choroidal segmentation from its correct position. The purpose of this study is to develop a 3D graph-based approach to identify the 3D choroidal layer in ONH-centered SD-OCT images using the BMO prior information. More specifically, an initial 3D choroidal segmentation was first performed using the 3D graph search algorithm. Note that varying surface interaction constraints based on the choroidal morphological model were applied. To assist the choroidal segmentation, two other surfaces of internal limiting membrane and innerouter segment junction were also segmented. Based on the segmented layer between the RPE/BM and C-S junction, a 2D projection map was created. The BMO in the projection map was detected by a 2D graph search. The pre-defined BMO information was then incorporated into the surface interaction constraints of the 3D graph search to obtain more accurate choroidal segmentation. Twenty SD-OCT images from 20 healthy subjects were used. The mean differences of the choroidal borders between the algorithm and manual segmentation were at a sub-voxel level, indicating a high level segmentation accuracy.
Geographic atrophy (GA) is a manifestation of the advanced or late stage of age-related macular degeneration (AMD). AMD is the leading cause of blindness in people over the age of 65 in the western world. The purpose of this study is to develop a fully automated supervised pixel classification approach for segmenting GA, including uni- and multifocal patches in fundus autofluorescene (FAF) images. The image features include region-wise intensity measures, gray-level co-occurrence matrix measures, and Gaussian filter banks. A k-nearest-neighbor pixel classifier is applied to obtain a GA probability map, representing the likelihood that the image pixel belongs to GA. Sixteen randomly chosen FAF images were obtained from 16 subjects with GA. The algorithm-defined GA regions are compared with manual delineation performed by a certified image reading center grader. Eight-fold cross-validation is applied to evaluate the algorithm performance. The mean overlap ratio (OR), area correlation (Pearson’s r), accuracy (ACC), true positive rate (TPR), specificity (SPC), positive predictive value (PPV), and false discovery rate (FDR) between the algorithm- and manually defined GA regions are 0.72±0.03, 0.98±0.02, 0.94±0.00, 0.87±0.01, 0.96±0.01, 0.80±0.04, and 0.20±0.04, respectively.
Age-related macular degeneration (AMD) is the leading cause of blindness in people over the age of 65. Geographic atrophy (GA) is a manifestation of the advanced or late-stage of the AMD, which may result in severe vision loss and blindness. Techniques to rapidly and precisely detect and quantify GA lesions would appear to be of important value in advancing the understanding of the pathogenesis of GA and the management of GA progression. The purpose of this study is to develop an automated supervised pixel classification approach for segmenting GA including uni-focal and multi-focal patches in fundus autofluorescene (FAF) images. The image features include region wise intensity (mean and variance) measures, gray level co-occurrence matrix measures (angular second moment, entropy, and inverse difference moment), and Gaussian filter banks. A k-nearest-neighbor (k-NN) pixel classifier is applied to obtain a GA probability map, representing the likelihood that the image pixel belongs to GA. A voting binary iterative hole filling filter is then applied to fill in the small holes. Sixteen randomly chosen FAF images were obtained from sixteen subjects with GA. The algorithm-defined GA regions are compared with manual delineation performed by certified graders. Two-fold cross-validation is applied for the evaluation of the classification performance. The mean Dice similarity coefficients (DSC) between the algorithm- and manually-defined GA regions are 0.84 ± 0.06 for one test and 0.83 ± 0.07 for the other test and the area correlations between them are 0.99 (p < 0.05) and 0.94 (p < 0.05) respectively.
Spectral-domain optical coherence tomography (SD-OCT) is a three-dimensional imaging technique that allows direct visualization of retinal morphology and architecture. The various retinal layers may be affected differentially by various diseases. An automated graph search algorithm is developed to sequentially segment 11 retinal surfaces in SD-OCT volumes using a three-stage approach. In stage 1, the four most easily discernible and/or distinct surfaces are identified in four-times-downsampled images and are used as a priori information to limit the graph search for the other surfaces in stage 2. Eleven surfaces were then detected in two-times-downsampled images in stage 2, and refined in the original images in stage 3 using the graph search integrating the estimated morphological shape models. Twenty macular SD-OCT volume scans from 20 normal subjects are used in this initial study. The overall mean and absolute mean differences in border positions between the automated and manual segmentation for the 11 surfaces are −0.20±0.53 voxels (−0.76±2.06 μm ) and 0.82±0.64 voxels (3.19±2.46 μm ), respectively. Intensity/reflectivity and thickness properties in various retinal layers are also investigated. This investigation in normal subjects may provide a comparative reference for subsequent adaptations in eyes with diseases.
Spectral-domain optical coherence tomography (SD-OCT) is a 3-D imaging technique, allowing direct visualization of retinal morphology and architecture. The various layers of the retina may be affected differentially by various diseases. In this study, an automated graph-based multilayer approach was developed to sequentially segment eleven retinal surfaces including the inner retinal bands to the outer retinal bands in normal SD-OCT volume scans at three different stages. For stage 1, the four most detectable and/or distinct surfaces were identified in the four-times-downsampled images and were used as a priori positional information to limit the graph search for other surfaces at stage 2. Eleven surfaces were then detected in the two-times-downsampled images at stage 2, and refined in the original image space at stage 3 using the graph search integrating the estimated morphological shape models. Twenty macular SD-OCT (Heidelberg Spectralis) volume scans from 20 normal subjects (one eye per subject) were used in this study. The overall mean and absolute mean differences in border positions between the automated and manual segmentation for all 11 segmented surfaces were -0.20 ± 0.53 voxels (-0.76 ± 2.06 μm) and 0.82 ± 0.64 voxels (3.19 ± 2.46 μm). Intensity and thickness properties in the resultant retinal layers were investigated. This investigation in normal subjects may provide a comparative reference for subsequent investigations in eyes with disease.
Segmenting vessels in spectral-domain optical coherence tomography (SD-OCT) volumes is particularly challenging
in the region near and inside the neural canal opening (NCO). Furthermore, accurately segmenting them
in color fundus photographs also presents a challenge near the projected NCO. However, both modalities also
provide complementary information to help indicate vessels, such as a better NCO contrast from the NCO-aimed
OCT projection image and a better vessel contrast inside the NCO from fundus photographs. We thus present
a novel multimodal automated classification approach for simultaneously segmenting vessels in SD-OCT volumes
and fundus photographs, with a particular focus on better segmenting vessels near and inside the NCO
by using a combination of their complementary features. In particular, in each SD-OCT volume, the algorithm
pre-segments the NCO using a graph-theoretic approach and then applies oriented Gabor wavelets with oriented
NCO-based templates to generate OCT image features. After fundus-to-OCT registration, the fundus image
features are computed using Gaussian filter banks and combined with OCT image features. A k-NN classifier is
trained on 5 and tested on 10 randomly chosen independent image pairs of SD-OCT volumes and fundus images
from 15 subjects with glaucoma. Using ROC analysis, we demonstrate an improvement over two closest previous
works performed in single modal SD-OCT volumes with an area under the curve (AUC) of 0.87 (0.81 for our
and 0.72 for Niemeijer's single modal approach) in the region around the NCO and 0.90 outside the NCO (0.84
for our and 0.81 for Niemeijer's single modal approach).
Optical coherence tomography (OCT), being a noninvasive imaging modality, has begun to find vast use in
the diagnosis and management of ocular diseases such as glaucoma, where the retinal nerve fiber layer (RNFL)
has been known to thin. Furthermore, the recent availability of the considerably larger volumetric data with
spectral-domain OCT has increased the need for new processing techniques. In this paper, we present an
automated 3-D graph-theoretic approach for the segmentation of 7 surfaces (6 layers) of the retina from 3-D
spectral-domain OCT images centered on the optic nerve head (ONH). The multiple surfaces are detected
simultaneously through the computation of a minimum-cost closed set in a vertex-weighted graph constructed
using edge/regional information, and subject to a priori determined varying surface interaction and smoothness
constraints. The method also addresses the challenges posed by presence of the large blood vessels and the optic
disc. The algorithm was compared to the average manual tracings of two observers on a total of 15 volumetric
scans, and the border positioning error was found to be 7.25 ± 1.08 μm and 8.94 ± 3.76 μm for the normal and
glaucomatous eyes, respectively. The RNFL thickness was also computed for 26 normal and 70 glaucomatous
scans where the glaucomatous eyes showed a significant thinning (p < 0.01, mean thickness 73.7 ± 32.7 μm in
normal eyes versus 60.4 ± 25.2 μm in glaucomatous eyes).
The optic disc margin is of interest due to its use for detecting and managing glaucoma. We developed a
method for segmenting the optic disc margin of the optic nerve head (ONH) in spectral-domain optical coherence
tomography (OCT) images using a graph-theoretic approach. A small number of slices surrounding the Bruch's
membrane opening (BMO) plane was taken and used for creating planar 2-D projection images. An edge-based
cost function - more specifically, a signed edge-based term favoring a dark-to-bright transition in the
vertical direction of polar projection images (corresponding to the radial direction in Cartesian coordinates)
- was obtained. Information from the segmented vessels was used to suppress the vasculature influence by
modifying the polar cost function and remedy the segmentation difficulty due to the presence of large vessels.
The graph search was performed in the modified edge-based cost images. The algorithm was tested on 22
volumetric OCT scans. The segmentation results were compared with expert segmentations on corresponding
stereo fundus disc photographs. We found a signed mean difference of 0.0058 ± 0.0706 mm and an unsigned
mean difference of 0.1083 ± 0.0350 mm between the automatic and expert segmentations.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.