Purpose: The purpose of this study was to evaluate the correlation between quantity of intraretinal hyperreflective foci (HRF) in the eye with intermediate AMD and progression to late AMD. Methods: Volume optical coherence tomography (OCT) scans from 114 eyes of 114 patients were retrospectively reviewed. HRF were assessed both qualitatively and quantitatively. Five sequential en face slabs from midretina were thresholded to isolate the HRF. These five slabs were recombined, and HRF area was measured in the whole 6 × 6-mm image (HRFTOT) and within the central 3-mm (HRF3mm) and 5-mm (HRF5mm) regions. These measurements were correlated with the development of late AMD (defined as choroidal neovascularization [CNV] and/or complete RPE and photoreceptor atrophy [cRORA]) after 1 year of follow-up. Results: HRF area in all three regions showed significant correlations with progression to late AMD: R = 0.610 for HRF3mm, R = 0.622 for HRF5mm, and R = 0.614 for HRFTOT (all P < 0.001). Correlations remained significant with progression to cRORA alone, though not for progression to CNV alone. While qualitative assessment of HRF (i.e., presence of HRF: yes or no) also showed a significant correlation with progression to late AMD (R = 0.454, P < 0.001) and atrophy alone (R = 0.445, P < 0.001), they were weaker than by HRF quantification. Conclusions: The area of HRF from en face OCT in eyes with intermediate AMD correlates with the 1-year risk of progression to late AMD, and in particular with the development of atrophy.
To evaluate the correlation between the choriocapillaris (CC) flow alterations around geographic atrophy (GA) and the GA yearly growth rate (yGR) in patients with dry age-related macular degeneration (AMD). We retrospectively reviewed and analyzed spectral domain optical coherence tomography (SD-OCT) and SD-OCT angiography images of consecutive patients with GA acquired using the Cirrus OCT at the Doheny Eye Centers between 2015 and 2017. All eligible patients had one 6 x 6 mm OCTA scan acquired during the first visit (considered as baseline) and two fovea-centered 512 x 128 macular cubes (6 x 6 mm) acquired at baseline and after a minimum of 12 months. The fundus images from the OCT volumes were used to manually delineate the GA area and calculate the yGR after square root transformation. The en-face angiogram at the level of the CC was analyzed for the percentage of flow voids (FV) outside the atrophic lesion (FVOUT) and in the para- and peri-atrophy regions (FV500 and FV1000 respectively; two concentric 500 μm wide rings around the atrophy edge). These values, together with the difference between FV500 and FV1000 (ΔFV), were then correlated with the corresponding yGR. Thirty-three eyes of 23 patients were eligible for the analysis. The mean yGR was 0.23 ± 0.17 mm/years. At baseline, the mean FVOUT was 41.86 ± 2.71%, while FV500 and FV1000 were 46.4 ± 4.17% and 42.51 ± 2.65% respectively. The mean ΔFV was 3.89 ± 2.6%. While in the univariable analysis, the yGR was significantly associated with FV500 and with ΔFV (both p < 0.001), in multivariable model the association remained significant only with ΔFV (p < 0.001). Our study reports a correlation between the CC flow impairment around the atrophic lesions and their yGR in patients with GA. If replicated in future longitudinal studies, the choriocapillaris FV in the para-and peri-atrophy regions may prove to be useful parameters for evaluating the prognosis of these eyes.
Age-related macular degeneration (AMD) affects millions of people and is a leading cause of blindness throughout the world. Ideally, affected individuals would be identified at an early stage before late sequelae such as outer retinal atrophy or exudative neovascular membranes develop, which could produce irreversible visual loss. Early identification could allow patients to be staged and appropriate monitoring intervals to be established. Accurate staging of earlier AMD stages could also facilitate the development of new preventative therapeutics. However, accurate and precise staging of AMD, particularly using newer optical coherence tomography (OCT)-based biomarkers may be time-intensive and requires expert training which may not feasible in many circumstances, particularly in screening settings. In this work we develop deep learning method for automated detection and classification of early AMD OCT biomarker. Deep convolution neural networks (CNN) were explicitly trained for performing automated detection and classification of hyperreflective foci, hyporeflective foci within the drusen, and subretinal drusenoid deposits from OCT B-scans. Numerous experiments were conducted to evaluate the performance of several state-of-the-art CNNs and different transfer learning protocols on an image dataset containing approximately 20000 OCT B-scans from 153 patients. An overall accuracy of 87% for identifying the presence of early AMD biomarkers was achieved.