Vessel Density Features of Optical Coherence Tomography Angiography for Classification of Optic Neuropathies Using Machine Learning.

BACKGROUND: To evaluate the classification performance of machine learning based on the 4 vessel density features of peripapillary optical coherence tomography angiography (OCT-A) for classifying healthy, nonarteritic anterior ischemic optic neuropathy (NAION), and optic neuritis (ON) eyes. METHODS: Forty-five eyes of 45 NAION patients, 32 eyes of 32 ON patients, and 76 eyes of 76 healthy individuals with optic nerve head OCT-A were included. Four vessel density features of OCT-A images were developed using a threshold-based segmentation method and were integrated in 3 models of machine learning classifiers. Classification performances of support vector machine (SVM), random forest, and Gaussian Naive Bayes (GNB) models were evaluated with the area under the receiver-operating-characteristic curve (AUC) and accuracy. RESULTS: We divided 121 images into a 70% training set and 30% test set. For ON-NAION classification, best results were achieved with 50% threshold, in which 3 classifiers (SVM, RF, and GNB) discriminated ON from NAION with an AUC of 1 and accuracy of 1. For ON-Normal classification, with 100% threshold, SVM and RF classifiers were able to discriminate normal from ON with AUCs of 1 and accuracies of 1. For NAION-normal classification, with 50% threshold, the SVM and RF classified the NAION from normal with AUC and accuracy of 1. CONCLUSIONS: ML based on the combined peripapillary vessel density features of total vessels and capillaries in the whole image and ring image could provide excellent performance for NAION and ON distinction.

Vessel Density Features of Optical Coherence Tomography Angiography for Classification of Glaucoma Using Machine Learning.

PRCIS: Machine learning (ML) based on the optical coherence tomography angiography vessel density features with different thresholds using a support vector machine (SVM) model provides excellent performance for glaucoma detection. BACKGROUND: To assess the classification performance of ML based on the 4 vessel density features of peripapillary optical coherence tomography angiography for glaucoma detection. METHODS: Images from 119 eyes of 119 glaucoma patients and 76 eyes of 76 healthy individuals were included. Four vessel density features of optical coherence tomography angiography images were developed using a threshold-based segmentation method and were integrated into 3 models of machine learning classifiers. Images were divided into 70% training set and 30% test set. Classification performances of SVM, random forest, and Gaussian Naive Bayes models were evaluated with the area under the receiver operating characteristic curve (AUC) and accuracy. RESULTS: Glaucoma eyes had lower vessel densities at different thresholds. For differentiating glaucoma eyes, the best results were achieved with 70% and 100% thresholds, in which SVM classifier discriminated glaucoma from healthy eyes with an AUC of 1 and accuracy of 1. After SVM, the random forest classifier with 100% thresholds showed an AUC of 0.993 and an accuracy of 0.994. Furthermore, the AUC of our ML performance (SVM) was 0.96 in a subgroup analysis of mild and moderate glaucoma eyes. CONCLUSIONS: ML based on the combined peripapillary vessel density features of total vessels and capillaries in the whole image and ring image could provide excellent performance for glaucoma detection.

Disagreement of Radial Peripapillary Capillary Density Among Four Optical Coherence Tomography Angiography Devices.

Purpose: This prospective study evaluated the agreement among four optical coherence tomography angiography (OCTA) devices in the assessment of radial peripapillary capillary (RPC) density. Methods: The study included 48 eyes of 48 subjects (14 healthy, 19 glaucomatous, and 15 non-glaucomatous optic neuropathy). Each participant was scanned using four OCTA devices in a random sequence: RTVue XR Avanti (RTVue), DRI OCT Triton (Triton), Revo NX 130 (Revo), and PLEX Elite 9000 (PlexE). All 6 × 6-mm grayscale OCTA images from each device were analyzed for RPC density using a customized algorithm. Agreement between each pair of devices was assessed using intraclass correlation coefficients (ICCs) and Bland-Altman plots. Results: There was a poor correlation between devices in all comparisons (RTVue-Triton, ICC = 0.34; RTVue-Revo, ICC = 0.31; RTVue-PlexE, ICC = 0.28; Triton-Revo, ICC = 0.31; Triton-PlexE, ICC = 0.17; Revo-PlexE, ICC = 0.34). Significant proportional biases (P < 0.05) and wide limits of agreement with apparent constant biases were identified in all comparisons. The mean difference was greatest for the RTVue-Revo pair (-49.3%; 95% confidence interval [CI], -52.9 to -45.8) and smallest for the Triton-PlexE pair (-7.7%; 95% CI, -10.1 to -5.3). Conclusions: The RPC densities obtained from each device had poor inter-device agreement and significant biases and cannot be used interchangeably. Translational Relevance: RPC density obtained from different OCTA devices is not interchangeable; thus, the progression of optic neuropathy should be monitored using the same OCTA device.

Choroidal Thickness in Different Types of Inherited Retinal Dystrophies.

PURPOSE: To compare the choroidal thickness among eyes with retinitis pigmentosa (RP), Stargardt disease, Usher syndrome, cone-rod dystrophy, and healthy eyes of sex- and age-matched individuals. METHODS: In this comparative study, 503 eyes with RP (n = 264), cone-rod dystrophy (n = 109), Stargardt disease (n = 76), and Usher syndrome (n = 54) were included. To validate the data, 109 healthy eyes of 56 sex- and age-matched individuals were studied as controls. Choroidal imaging was performed using enhanced depth imaging-optical coherence tomography. Choroidal thickness was measured manually using MATLAB software at 13 points in nasal and temporal directions from the foveal center with the interval of 500 µm and the choroidal area encompassing the measured points was calculated automatically. RESULTS: The mean age was 36.33 ± 13.07 years (range, 5 to 72 years). The mean choroidal thickness at 13 points of the control eyes was statistically significantly higher than that in eyes with RP (P < 0.001) and Usher syndrome (P < 0.05), but not significantly different from that in eyes with Stargardt disease and cone-rod dystrophy. Among different inherited retinal dystrophies (IRDs), the choroidal thickness was the lowest in eyes with RP (P < 0.001). Choroidal thickness in the subfoveal area correlated negatively with best-corrected visual acuity (r = - 0.264, P < 0.001) and the duration of ocular symptoms (r = - 0.341, P < 0.001) in all studied IRDs. No significant correlation was observed between the subfoveal choroidal thickness and central macular thickness (r = - 0.24, P = 0.576). CONCLUSION: Choroidal thinning in four different types of IRDs does not follow a similar pattern and depends on the type of IRD and the duration of ocular symptoms. A larger cohort is required to verify these findings.

Erratum: Retinal image mosaicking using scale-invariant feature transformation feature descriptors and Voronoi diagram (Erratum).

[This corrects the article DOI: 10.1117/1.JMI.7.4.044001.].

Retinal image mosaicking using scale-invariant feature transformation feature descriptors and Voronoi diagram.

Purpose: Peripheral retinal lesions substantially increase the risk of diabetic retinopathy and retinopathy of prematurity. The peripheral changes can be visualized in wide field imaging, which is obtained by combining multiple images with an overlapping field of view using mosaicking methods. However, a robust and accurate registration of mosaicking techniques for normal angle fundus cameras is still a challenge due to the random selection of matching points and execution time. We propose a method of retinal image mosaicking based on scale-invariant feature transformation (SIFT) feature descriptor and Voronoi diagram. Approach: In our method, the SIFT algorithm is used to describe local features in the input images. Then the input images are subdivided into regions based on the Voronoi method. Each pair of Voronoi regions is matched by the method zero mean normalized cross correlation. After matching, the retinal images are mapped into the same coordinate system to form a mosaic image. The success rate and the mean registration error (RE) of our method were compared with those of other state-of-the-art methods for the P category of the fundus image registration database. Results: Experimental results show that the proposed method accurately registered 42% of retinal image pairs with a mean RE of 3.040 pixels, while a lower success rate was observed in the other four state-of-the-art retinal image registration methods GDB-ICP (33%), Harris-PIIFD (0%), HM-2016 (0%), and HM-2017 (2%). Conclusions: The proposed method outperforms state-of-the-art methods in terms of quality and running time and reduces the computational complexity.

Forming Optimal Projection Images from Intra-Retinal Layers Using Curvelet-Based Image Fusion Method.

BACKGROUND: Image fusion is the process of combining the information of several input images into one image. Projection images obtained from three-dimensional (3D) optical coherence tomography (OCT) can show inlier retinal pathology and abnormalities that are not visible in conventional fundus images. In recent years, the projection image is often made by an average on all retina that causes to lose many intraretinal details. METHODS: In this study, we focus on the formation of optimum projection images from retinal layers using Curvelet-based image fusion. The latter consists of three main steps. In the earlier studies, macular spectral 3D data using diffusion map-based OCT were segmented into 12 different boundaries identifying 11 retinal layers in three dimensions. In the second step, projection images are attained using conducting some statistical methods on the space between each pair of boundaries. In the next step, retinal layers are merged using Curvelet transform to make the final projection images. RESULTS: These images contain integrated retinal depth information as well as an ideal opportunity to better extract retinal features such as vessels and the macula region. Finally, qualitative and quantitative evaluations show the superiority of this method to the average-based and wavelet-based fusion methods. Overall, our method obtains the best results for image fusion in all terms such as entropy (6.7744) and AG (9.5491). CONCLUSION: Creating an image with more and detailed information made by the Curvelet-based image fusion has significantly higher contrast. There are also many thin veins in Curvelet-based fused image, which are absent in average-based and wavelet-based fused images.

Optical Coherence Tomography Angiography in Papilledema Compared With Pseudopapilledema.

Purpose: The purpose of this study is to evaluate differences in optical coherence tomography angiography (OCT-A) findings between patients with papilledema and pseudopapilledema. Methods: In this prospective, comparative study, 41 eyes of 21 subjects with papilledema, 27 eyes of 15 subjects with pseudopapilledema, and 44 eyes of 44 healthy normal subjects were included and were imaged using OCT-A. In addition to peripapillary total vasculature maps obtained with commercial vessel density mapping, major vessel removal using customized image analysis software was also used to measure whole image capillary density and peripapillary capillary density (PCD). Peripapiilary retinal nerve fiber layer (RNFL) and macular ganglion cell complex (GCC) were recorded. Results: Average RNFL thicknesses were greater in papilledema eyes than in pseudopapilledema and control subjects. GCC thickness was not different among three groups. Peripapillary vasculature values were significantly lower in papilledema (58.5 ± 6.1%) and pseudopapilledema (58.9 ± 4.7%) eyes compared with healthy eyes (63.2 ± 3.1%) using commercial machine software, without a difference between papilledema and pseudopapilledema eyes. However, using our customized software, peripapillary "capillary" density of papilledema eyes was 29.8 ± 9.4%, which was not significantly different from healthy subjects (31.8 ± 7.4%; P = 0.94). Pseudopapilledema eyes with peripapillary density of 25.5 ± 8.3% had significantly lower capillary values compared with control eyes (P = 0.01). There was a significantly lower whole image and nasal sector peripapillary capillary density of inner retina in pseudopapilledema eyes than papilledema eyes (P = 0.03 and P = 0.02, respectively). Conclusions: Whole image and nasal peripapillary sector capillary densities using OCT-A had diagnostic accuracy for differentiating true and pseudo-disc swelling.

Optical Coherence Tomography Angiography in Optic Disc Swelling.

PURPOSE: To compare optical coherence tomography angiography (OCT-A) of peripapillary total vasculature and capillaries in patients with optic disc swelling. DESIGN: Cross-sectional study. METHODS: Twenty nine eyes with acute nonarteritic anterior ischemic optic neuropathy (NAION), 44 eyes with papilledema, 8 eyes with acute optic neuritis, and 48 eyes of normal subjects were imaged using OCT-A. Peripapillary total vasculature information was recorded using a commercial vessel density map. Customized image analysis with major vessel removal was also used to measure whole-image capillary density and peripapillary capillary density (PCD). RESULTS: Mixed models showed that the peripapillary total vasculature density values were significantly lower in NAION eyes, followed by papilledema eyes and control eyes, using commercial software (P < .0001 for all comparisons). The customized software also showed significantly lower PCD of NAION eyes compared with papilledema eyes (all P < .001), but did not show significant differences between papilledema and control subjects. Our software showed significantly lower whole image and PCD in eyes with optic neuritis than papilledema. There was no significant difference between NAION and optic neuritis using our software. The area under the receiver operating curves for discriminating NAION from papilledema eyes and optic neuritis from papilledema eyes was highest for whole-image capillary density (0.94 and 0.80, respectively) with our software, followed by peripapillary total vasculature (0.9 and 0.74, respectively) with commercial software. CONCLUSIONS: OCT-A is helpful to distinguish NAION and papillitis from papilledema. Whole-image capillary density had the greatest diagnostic accuracy for differentiating disc swelling.