Dataset Information

Exploring a Structural Basis for Delayed Rod-Mediated Dark Adaptation in Age-Related Macular Degeneration Via Deep Learning.

ABSTRACT:

Purpose

Delayed rod-mediated dark adaptation (RMDA) is a functional biomarker for incipient age-related macular degeneration (AMD). We used anatomically restricted spectral domain optical coherence tomography (SD-OCT) imaging data to localize de novo imaging features associated with and to test hypotheses about delayed RMDA.

Methods

Rod intercept time (RIT) was measured in participants with and without AMD at 5 degrees from the fovea, and macular SD-OCT images were obtained. A deep learning model was trained with anatomically restricted information using a single representative B-scan through the fovea of each eye. Mean-occlusion masking was utilized to isolate the relevant imaging features.

Results

The model identified hyporeflective outer retinal bands on macular SD-OCT associated with delayed RMDA. The validation mean standard error (MSE) registered to the foveal B-scan localized the lowest error to 0.5 mm temporal to the fovea center, within an overall low-error region across the rod-free zone and adjoining parafovea. Mean absolute error (MAE) on the test set was 4.71 minutes (8.8% of the dynamic range).

Conclusions

We report a novel framework for imaging biomarker discovery using deep learning and demonstrate its ability to identify and localize a previously undescribed biomarker in retinal imaging. The hyporeflective outer retinal bands in central macula on SD-OCT demonstrate a structural basis for dysfunctional rod vision that correlates to published histopathologic findings.

Translational relevance

This agnostic approach to anatomic biomarker discovery strengthens the rationale for RMDA as an outcome measure in early AMD clinical trials, and also expands the utility of deep learning beyond automated diagnosis to fundamental discovery.

SUBMITTER: Lee AY

PROVIDER: S-EPMC7745629 | biostudies-literature | 2020 Dec

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Exploring a Structural Basis for Delayed Rod-Mediated Dark Adaptation in Age-Related Macular Degeneration Via Deep Learning.

Lee Aaron Y AY Lee Cecilia S CS Blazes Marian S MS Owen Julia P JP Bagdasarova Yelena Y Wu Yue Y Spaide Theodore T Yanagihara Ryan T RT Kihara Yuka Y Clark Mark E ME Kwon MiYoung M Owsley Cynthia C Curcio Christine A CA

Translational vision science & technology 20201215 2

<h4>Purpose</h4>Delayed rod-mediated dark adaptation (RMDA) is a functional biomarker for incipient age-related macular degeneration (AMD). We used anatomically restricted spectral domain optical coherence tomography (SD-OCT) imaging data to localize de novo imaging features associated with and to test hypotheses about delayed RMDA.<h4>Methods</h4>Rod intercept time (RIT) was measured in participants with and without AMD at 5 degrees from the fovea, and macular SD-OCT images were obtained. A dee ...[more]

PMID: 33344065

Similar Datasets

Project description:PurposeHyperreflective foci (HRF) are OCT biomarkers for the progression of nonneovascular age-related macular degeneration (AMD) attributed to anteriorly migrated retinal pigment epithelial cells. We examined associations between rod- and cone-mediated vision and HRF plus smaller hyperreflective specks (HRS); we identified a histologic candidate for HRS.DesignCross-sectional study and histologic survey.ParticipantsPatients with healthy maculae (n = 34), early AMD (n = 26), and intermediate AMD (n = 41).MethodsAMD severity was determined by color fundus photography. In OCT scans, HRF and HRS were counted manually. Vision tests probed cones (best-corrected visual acuity [VA], contrast sensitivity), mixed cones and rods (low-luminance VA, low-luminance deficit, mesopic light sensitivity), or rods (scotopic light sensitivity, rod-mediated dark adaptation [RMDA]). An online AMD histopathologic resource was reviewed.Main outcome measuresVision in eyes assessed for HRF and HRS; histologic candidate for HRS.ResultsIn 101 eyes of 101 patients, HRF and HRS were identified in 25 and 95 eyes, respectively, with good reliability. Hyperreflective foci were present but sparse in healthy eyes, infrequent in early AMD eyes, and frequent but highly variable among intermediate AMD eyes (mean±standard deviation [SD] number per eye, 0.1 ± 0.2, 0.2 ± 0.5, and 1.9 ± 3.4 for healthy, early AMD, and intermediate AMD eyes, respectively). Hyperreflective specks outnumbered HRF in all groups (mean±SD, 4.5 ± 3.2, 6.3 ± 5.8, and 19.4 ± 22.4, respectively). Delayed RMDA was associated strongly with more HRF and HRS (P < 0.0001). Hyperreflective foci also were associated with worse low-luminance VA (P = 0.0117). Hyperreflective specks were associated with worse contrast sensitivity (P = 0.0278), low-luminance VA (P = 0.0010), low-luminance deficit (P = 0.0031), and mesopic (P = 0.0018) and scotopic (P < 0.0001) sensitivity. By histologic analysis, cone lipofuscin was found in outer retinal layers of 25% of healthy aged eyes.ConclusionsHyperreflective foci and HRS are markers of cellular activity associated with visual dysfunction, especially delayed RMDA, an AMD risk indicator assessing efficiency of retinoid resupply. Hyperreflective specks may represent lipofuscin translocating inwardly within cones. HRF and HRS may serve as structural end points in clinical trials targeting AMD stages earlier than atrophy expansion. These results should be confirmed in a larger sample.

Project description:PurposeTo examine the association between sequence variants in genetic risk factors for age-related macular degeneration (AMD) and delayed rod-mediated dark adaptation (RMDA), the first functional biomarker for incident AMD, in older adults with normal macular health and early AMD.DesignCross-sectional.ParticipantsAdults 60 years of age or older showing normal macular health (defined as both eyes at step 1 on the Age-Related Eye Disease Study 9-step AMD classification system) and those with AMD in one or both eyes (defined as steps 2-9).MethodsSingle nucleotide polymorphisms were genotyped in the complement factor H (CFH) and ARMS2 genes using a Taqman assay. Rod-mediated dark adaptation was assessed in 1 eye after photobleach with targets centered at 5° on the inferior vertical meridian. Rate of dark adaptation was defined by rod intercept time (RIT), duration (in minutes) required for sensitivity to reach a criterion sensitivity level in the latter half of the second component of rod recovery. Associations between CFH and ARMS2 polymorphisms and RMDA were adjusted for age and smoking.Main outcome measureRod intercept time.ResultsThe sample consisted of 543 participants having both genotype and RIT determination; 408 showed normal macular health and 135 demonstrated AMD, most having early AMD (124 of 135). For the combined sample, higher RIT (slower RMDA) was observed for both the A69S variant in ARMS2 and the Y402H variant in CFH (adjusted P = 0.0001 and P = 0.0023, respectively). For healthy participants, the A69S variant in ARMS2 was associated with higher RIT (adjusted P = 0.0011), whereas the Y402H variant in CFH was not (adjusted P = 0.2175). For AMD patients, the A69S variant of ARMS2 and the Y402H variant of CFH were associated with higher RIT (adjusted P = 0.0182 and P = 0.0222, respectively). Those with a larger number of high-risk ARMS2 and CFH alleles showed higher RIT, in both healthy and AMD groups (adjusted P = 0.0002 and P < 0.0001, respectively).ConclusionsWe report a novel association wherein older adults with high-risk ARMS2 and CFH genotypes are more likely to demonstrate delayed RMDA, the first functional biomarker for incident early AMD. Before the AMD clinical phenotype is present, those showing normal macular health with the ARMS2 A69S allele demonstrate delayed RMDA. Understanding ARMS2 function is a research priority.

Project description:PurposeWe evaluate the impact of test target location in assessing rod-mediated dark adaptation (RMDA) along the transition from normal aging to intermediate age-related macular degeneration (AMD). We consider whether RMDA slows because the test locations are near mechanisms leading to or resulting from high-risk extracellular deposits. Soft drusen cluster under the fovea and extend to the inner ring of the ETDRS grid where rods are sparse. Subretinal drusenoid deposits (SDDs) appear first in the outer superior subfield of the ETDRS grid where rod photoreceptors are maximal and spread toward the fovea without covering it.DesignCross-sectional.ParticipantsAdults ≥ 60 years with normal older maculas, early AMD, or intermediate AMD as defined by the Age-Related Eye Disease Study (AREDS) 9-step and Beckman grading systems.MethodsIn 1 eye per participant, RMDA was assessed at 5° and at 12° in the superior retina. Subretinal drusenoid deposit presence was identified with multi-modal imaging.Main outcome measuresRod intercept time (RIT) as a measure of RMDA rate at 5° and 12°.ResultsIn 438 eyes of 438 persons, RIT was significantly longer (i.e., RMDA is slower) at 5° than at 12° for each AMD severity group. Differences among groups were bigger at 5° than at 12°. At 5°, SDD presence was associated with longer RIT as compared to SDD absence at early and intermediate AMD but not in normal eyes. At 12°, SDD presence was associated with longer RIT in intermediate AMD only, and not in normal or early AMD eyes. Findings were similar in eyes stratified by AREDS 9-step and Beckman systems.ConclusionsWe probed RMDA in relation to current models of deposit-driven AMD progression organized around photoreceptor topography. In eyes with SDD, slowed RMDA occurs at 5° where these deposits typically do not appear until later in AMD. Even in eyes lacking detectable SDD, RMDA at 5° is slower than at 12°. The effect at 5° may be attributed to mechanisms associated with the accumulation of soft drusen and precursors under the macula lutea throughout adulthood. These data will facilitate the design of efficient clinical trials for interventions that aim to delay AMD progression.

Project description:PurposeWe investigate whether ocular and person-based characteristics were associated with dark adaptation (DA).DesignCross-sectional, single-center, observational study.ParticipantsOne hundred sixteen participants older than 50 years of age with a range of age-related macular degeneration (AMD) severity.MethodsParticipants underwent best-corrected visual acuity (BCVA) testing, ophthalmoscopic examination, and multimodal imaging. Presence of reticular pseudodrusen (RPD) was assessed by masked grading of fundus images and was confirmed with optical coherence tomography. Eyes also were graded for AMD features (drusen, pigmentary changes, late AMD) to generate person-based AMD severity groups. One eye was designated the study eye for DA testing. Nonparametric statistical testing was performed on all comparisons.Main outcome measuresThe primary outcome of this study was the rod intercept time (RIT), which is defined as the time for a participant's visual sensitivity to recover to a stimulus intensity of 5×10(-3) cd/m(2) (a decrease of 3 log units), or until a maximum test duration of 40 minutes was reached.ResultsA total of 116 study eyes from 116 participants (mean age, 75.4±9.4 years; 58% female) were analyzed. Increased RIT was associated significantly with increasing AMD severity, increasing age (r = 0.34; P = 0.0002), decreasing BCVA (r = -0.54; P < 0.0001), pseudophakia (P = 0.03), and decreasing subfoveal choroidal thickness (r = -0.27; P = 0.003). Study eyes with RPD (15/116 [13%]) had a significantly greater mean RIT compared with eyes without RPD in any AMD severity group (P < 0.02 for all comparisons), with 80% reaching the DA test ceiling.ConclusionsImpairments in DA increased with age, worse visual acuity, presence of RPD, AMD severity, and decreased subfoveal choroidal thickness. Analysis of covariance found the multivariate model that best fit the data included age, AMD group, and presence of RPD (R(2) = 0.56), with the presence of RPD conferring the largest parameter estimate.

Project description:PurposeTo investigate the natural history of dark adaptation (DA) function as measured by the change in rod intercept time (RIT) over 4 years and to correlate RIT change with age-related macular degeneration (AMD) severity.DesignLongitudinal, single-center, observational study.ParticipantsA total of 77 participants aged ≥50 years with a range of AMD severities.MethodsParticipants each contributing a single study eye to the analysis were assigned into person-based AMD severity groups based on fundus characteristics (drusen, pigmentary changes, late AMD, and subretinal drusenoid deposits [SDDs]). The DA function was assessed in study eyes at baseline and 3, 6, 12, 18, 24, 36, and 48 months. Mean change in DA function over time was calculated using the slope of linear regression fits of longitudinal RIT data. Patient-reported responses on a Low Luminance Questionnaire (LLQ) were obtained at baseline and yearly. Nonparametric statistical testing was performed on all comparisons.Main outcome measureThe RIT, defined as the time taken after a photobleach for visual sensitivity to recover detection of a 5×10-3 cd/m2 stimulus (a decrease of 3 log units), was monitored in study eyes over 4 years, and the mean rate of change was computed.ResultsLongitudinal analysis of 65 study eyes followed on the standard testing protocol (mean age, 71±9.3 years; 49% were female) revealed that higher rates of RIT prolongation were correlated with AMD severity group assignment at baseline (P = 0.026) and with severity group assignments at year 4 (P = 0.0011). Study eyes that developed SDD during follow-up demonstrated higher rates of RIT prolongation relative to those that did not (P < 0.0001). Overall, higher rates of RIT prolongation were significantly correlated with greater 4-year decreases in LLQ scores (total mean score, P = 0.0032).ConclusionsLongitudinal decline in DA function, which correlated with patient-reported functional deficits, was accelerated in eyes with greater AMD severity and especially in eyes with SDD both at baseline and at 4 years. The RIT prolongation as a measure of changing DA function may be a functional outcome measure in AMD clinical studies.

Dataset Information

Exploring a Structural Basis for Delayed Rod-Mediated Dark Adaptation in Age-Related Macular Degeneration Via Deep Learning.

Purpose

Methods

Results

Conclusions

Translational relevance

Publications

Exploring a Structural Basis for Delayed Rod-Mediated Dark Adaptation in Age-Related Macular Degeneration Via Deep Learning.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets