Search

Purpose: A deep learning model was developed to detect nonexudative macular neovascularization (neMNV) using OCT B-scans. Design: Retrospective review of a prospective, observational study. Participants: Normal control eyes and patients with age-related macular degeneration (AMD) with and without neMNV. Methods: Swept-source OCT angiography (SS-OCTA) imaging (PLEX Elite 9000, Carl Zeiss Meditec, Inc) was performed using the 6 × 6-mm scan pattern. Individual B-scans were annotated to distinguish between drusen and the double-layer sign (DLS) associated with the neMNV. The machine learning model was tested on a dataset graded by humans, and model performance was compared with the human graders. Main Outcome Measures: Intersection over Union (IoU) score was measured to evaluate segmentation network performance. Area under the receiver operating characteristic curve values, sensitivity, specificity, and positive predictive value (PPV) and negative predictive value (NPV) were measured to assess the performance of the final classification performance. Chance-corrected agreement between the algorithm and the human grader determinations was measured with Cohen’s kappa. Results: A total of 251 eyes from 210 patients, including 182 eyes with DLS and 115 eyes with drusen, were used for model training. Of 125 500 B-scans, 6879 B-scans were manually annotated. A vision transformer segmentation model was built to extract DLS and drusen from B-scans. The extracted prediction masks from all B-scans in a volume were projected to an en face image, and an eye-level projection map was obtained for each eye. A binary classification algorithm was established to identify eyes with neMNV from the projection map. The algorithm achieved 82%, 90%, 79%, and 91% sensitivity, specificity, PPV, and NPV, respectively, on a separate test set of 100 eyes that were evaluated by human graders in a previous study. The area under the curve value was calculated as 0.91 (95% confidence interval, 0.85–0.98). The results of the algorithm showed excellent agreement with the senior human grader (kappa = 0.83, P < 0.001) and moderate agreement with the junior grader consensus (kappa = 0.54, P < 0.001). Conclusions: Our network (code is available at https://github.com/uw-biomedical-ml/double_layer_vit) was able to detect the presence of neMNV from structural B-scans alone by applying a purely transformer-based model.

Purpose: En face OCT imaging was investigated as a method for the detection and monitoring of calcified drusen in eyes with nonexudative age-related macular degeneration (AMD). Design: Retrospective case series of a prospective study. Participants: Patients with nonexudative AMD. Methods: A retrospective review was performed of same-day color fundus (CF), fundus autofluorescence (FAF), near-infrared (NIR), and en face swept-source (SS) OCT images to identify eyes with nonexudative AMD and calcified drusen. The appearance and progression of these lesions were compared using the different imaging methods. Main Outcome Measures: Comparison between the presence of calcified drusen observed on CF images with the detection of these lesions on FAF, NIR, and en face SS OCT images. Results: Two hundred twenty eyes from 139 patients with nonexudative AMD were studied, with 42.7% of eyes containing calcified drusen either at baseline or during follow-up visits. On the en face SS OCT images, calcified drusen appeared as dark focal lesions referred to as choroidal hypotransmission defects (hypoTDs) that were detected in the choroid using a sub–retinal pigment epithelium (RPE) slab. The corresponding B-scans showed drusen with heterogenous internal reflectivity, hyporeflective cores, and hyperreflective caps. In most calcified drusen, choroidal hypertransmission defects (hyperTDs) were observed to develop over time around the periphery of the hypoTDs, giving them the appearance of a donut lesion on the en face SS OCT images. These donut lesions were associated with significant attenuation of the overlying retina, and the corresponding FAF images showed hypoautofluorescence at the location of these lesions. The donut lesions fulfilled the requirement for a persistent hyperTD, which is synonymous with complete RPE and outer retinal atrophy (cRORA). Six eyes displayed regression of the calcified drusen without cRORA developing. B-scans at the location of these regressed calcified drusen showed deposits along the RPE, with outer retinal thinning in the regions where the calcified lesions previously existed. Conclusions: En face OCT imaging is a useful method for the detection and monitoring of calcified drusen and can be used to document the evolution of these drusen as they form donut lesions or foci of cRORA.

Purpose: Multimodal imaging was used to identify and characterize the cause of hyperpigmentation seen on color fundus images (CFIs) of eyes with intermediate age-related macular degeneration (iAMD). Design: Retrospective review of a prospective study. Participants: Patients with iAMD. Methods: Color fundus images with macular hyperpigmentation were compared with same-day images obtained using fundus autofluorescence (FAF), near infrared reflectance (NIR), and swept-source (SS) OCT imaging. Two SS OCT en face slabs were generated: a retinal slab to identify hyperreflective foci within the retina and a slab from beneath the retinal pigment epithelium (RPE; the sub-RPE slab) that was used to detect regions that cause decreased light transmission into the choroid, also known as hypotransmission defects. All images were registered to allow for qualitative comparisons by 2 independent graders. Main Outcome Measures: Comparison between foci of macular hyperpigmentation seen on CFIs with the detection of these regions on FAF, NIR, and SS OCT en face images. Results: Compared with CFIs, FAF imaging seemed to be the least sensitive method for the detection of hyperpigmentation, whereas NIR and SS OCT imaging reliably detected these hyperpigmented areas. Although NIR imaging detected most of the hyperpigmentation seen in CFIs, SS OCT imaging detected all the areas of hyperpigmentation and anatomically localized these areas by using both en face and B-scan images. En face OCT slabs of the retina and sub-RPE region were registered to the CFIs, and areas of hyperpigmentation were shown to correspond to hyperreflective foci in the retina and regions of thickened RPE seen on OCT B-scans. Although both hyperpigmentation and early atrophic lesions appeared bright on NIR imaging, en face SS OCT imaging was able to distinguish these lesions because hyperpigmentary changes appeared dark and early atrophic lesions appeared bright on the sub-RPE slab. Conclusions: En face OCT imaging in conjunction with OCT B-scans were able to identify and localize the hyperpigmentation seen in CFIs reliably. This hyperpigmentation was not only associated with intraretinal hyperreflective foci, but also corresponded to areas with a thickened RPE.