Integrating Machine Learning and Traditional Survival Analysis to Identify Key Predictors of Foveal Involvement in Geographic Atrophy.

Purpose: The purpose of this study was to investigate the incidence of foveal involvement in geographic atrophy (GA) secondary to age-related macular degeneration (AMD), using machine learning to assess the importance of risk factors. Methods: Retrospective, longitudinal cohort study. Patients diagnosed with foveal-sparing GA, having GA size ≥ 0.049 mm² and follow-up ≥ 6 months, were included. Baseline GA area, distance from the fovea, and perilesional patterns were measured using fundus autofluorescence. Optical coherence tomography assessed foveal involvement, structural biomarkers, and outer retinal layers thickness. Onset of foveal involvement was recorded. Foveal survival rates were estimated using Kaplan-Meier curves. Hazard ratios (HRs) were assessed with mixed model Cox regression. Variable Importance (VIMP) was ranked with Random Survival Forests (RSF), with higher scores indicating greater predictive significance. Results: One hundred sixty-seven eyes (115 patients, average age = 75.8 ± 9.47 years) with mean follow-up of 50 ± 29 months, were included in this study. Median foveal survival time was 45 months (95% confidence interval [CI] = 38-55). Incidences of foveal involvement were 26% at 24 months and 67% at 60 months. Risk factors were GA proximity to the fovea (HR = 0.97 per 10-µm increase, 95% CI = 0.96-0.98), worse baseline visual acuity (HR = 1.37 per 0.1 LogMAR increase, 95% CI = 1.21-1.53), and thinner outer nuclear layer (HR = 0.59 per 10-µm increase, 95% CI = 0.46-0.74). RSF analysis confirmed these as main predictors (VIMP = 16.7, P = 0.002; VIMP = 6.2, P = 0.003; and VIMP = 3.4, P = 0.01). Lesser baseline GA area (HR = 1.09 per 1-mm2 increase, 95% CI = 1.01-1.16) and presence of a double layer sign (HR = 0.42, 95% CI = 0.20-0.88) were protective but less influential. Conclusions: This study identifies anatomic and functional factors impacting the risk of foveal involvement in GA. These findings may help identify at-risk patients, enabling tailored preventive strategies.

Early Choriocapillaris Loss in a Porcine Model of RPE Cell Debridement Precedes Pathology That Simulates Advanced Macular Degeneration.

Purpose: No large-mammal surgical models exist for geographic atrophy (GA), choroidal neovascularization (CNV), and pachychoroidal vascular remodeling. Our goal was to develop a porcine RPE debridement model of advanced macular degeneration to study photoreceptor cell loss and choroidal remodeling. Methods: Seven 2-month-old female domestic pigs were used for this study. After 25G vitrectomy, the area centralis was detached via subretinal bleb. A nitinol wire (Finesse Flex Loop) was used to debride RPE cells across a 3- to 5-mm diameter region. Fluid-air exchange was performed, and 20% SF6 gas injected. Animals underwent fundus photography, fluorescein angiography, optical coherence tomography (OCT), and OCT-angiography (OCTA) at 2 weeks, 1 month, 2 months, 3 months, and 6 months postoperatively. Retinal histology was obtained at euthanasia, 2 months (n = 3), 3 months (n = 2), or 6 months (n = 2) after surgery. Results: RPE debridement resulted in GA with rapid loss of choriocapillaris, progressive loss of photoreceptors, and pachychoroidal changes in Sattler's and Haller's layers in all seven eyes undergoing debridement within 2 months. OCT and histological findings included subretinal disciform scar with overlying outer retinal atrophy; outer retinal tubulations and subretinal hyper-reflective material. OCTA revealed type 2 CNV (n = 4) at the edges of the debridement zone by 2 months, but there was no significant exudation noted at any time point. Conclusions: Surgical debridement of the RPE results in GA, CNV, and pachychoroid and reproduced all forms of advanced macular degeneration. This surgical model may be useful in examining the role of RPE and other cell replacement in treating advanced macular disease.

The Association Between Metformin Use and New-Onset ICD Coding of Geographic Atrophy.

Purpose: Metformin has been suggested to protect against the development of age-related macular degeneration (AMD) in multiple observational studies. However, the association between metformin and geographic atrophy (GA), a debilitating subtype of AMD, has not been analyzed. Methods: We conducted a case-control study of patients ages 60 years and older with new-onset International Classification of Diseases (ICD) coding of GA in the Merative MarketScan Commercial and Medicare Databases between 2017 and 2021. Cases were matched with propensity scores estimated by age, region, hypertension, and Charlson Comorbidity Index to a control without GA of the same year. Exposure to metformin was assessed for cases and controls in the year prior to their index visit. Conditional multivariable logistic regression, adjusting for AMD risk factors, was used to calculate odd ratios and 95% confidence intervals (CIs). This study design and analysis were repeated in a sample of patients without diabetes. Results: In the full sample, we identified 10,505 cases with GA and 10,502 matched controls without GA. In total, 1149 (10.9%) cases and 1277 (12.2%) controls were exposed to metformin, and in multivariable regression, metformin decreased the odds of new-onset ICD coding of GA by 12% (95% CI, 0.79-0.99). In the sample of patients without diabetes, we identified 7611 cases with GA and 7608 matched controls without GA. Twenty-nine (0.4%) cases and 63 (0.8%) controls were exposed to metformin, and in multivariable regression, metformin decreased the odds of new-onset ICD coding of GA by 47% (95% CI, 0.33-0.83). Conclusions: Metformin may hold promise as a noninvasive, alternative agent to prevent the development of GA. This finding is notable due to shortcomings in recently approved therapeutics for GA and metformin's overall ease of use and few adverse effects. Additional studies are required to explore our findings further and motivate a clinical trial.

Characterizing the Journey of Geographic Atrophy Patients in Routine Ophthalmic Practice.

BACKGROUND AND OBJECTIVE: Geographic atrophy (GA) is a form of late-stage age-related macular degeneration (AMD). This study aims to characterize the journey of patients with GA in real-world ophthalmology practice. PATIENTS AND METHODS: This is a retrospective cohort study of 100 patients with GA and 100 with intermediate AMD (iAMD). RESULTS: Approximately one-third of GA patients' eyes had GA at the time of their initial AMD diagnosis, and nearly half of the iAMD patients' eyes had iAMD at that time. When holding confounders constant, GA patients experienced significantly worse visual acuity outcomes, and a significantly higher proportion required referrals for low vision evaluation, needed assistance for activities of daily living, failed to meet driving standards, and met criteria for legal blindness when compared to iAMD controls. CONCLUSIONS: Many patients have already progressed to GA by the time they receive an AMD diagnosis, emphasizing the importance of providing early detection and intervention, especially considering novel treatment options. [Ophthalmic Surg Lasers Imaging Retina 2024;55:204-210.].

Optical coherence tomography biomarkers in early and intermediate age-related macular degeneration: A clinical guide.

Advanced forms of age-related macular degeneration (AMD), characterised by atrophic and neovascular changes, are a leading cause of vision loss in the elderly population worldwide. Prior to the development of advanced AMD, a myriad of risk factors from the early and intermediate stages of AMD have been published in the scientific literature over the last years. The ability to precisely recognise structural and anatomical changes in the ageing macula, altogether with the understanding of the individual risk implications of each one of them is key for an accurate and personalised diagnostic assessment. The present review aims to summarise updated evidence of the relative risk conferred by diverse macular signs, commonly seen on optical coherence tomography, in terms of progression to geographic atrophy or macular neovascularization. This information may also serve as a basis for tailored follow-up monitoring visits.

Assessment of Choriocapillaris Flow Prior to Nascent Geographic Atrophy Development Using Optical Coherence Tomography Angiography.

Purpose: To assess the relationship between choriocapillaris (CC) loss and the development of nascent geographic atrophy (nGA) using optical coherence tomography angiography (OCTA) imaging. Methods: In total, 105 from 62 participants with bilateral large drusen, without late age-related macular degeneration (AMD) or nGA at baseline, were included in this prospective, longitudinal, observational study. Participants underwent swept-source OCTA imaging at 6-month intervals. CC flow deficit percentage (FD%) and drusen volume measurements were determined for the visit prior to nGA development or the second-to-last visit if nGA did not develop. Global and local analyses, the latter based on analyses within superpixels (120 × 120-µm regions), were performed to examine the association between CC FD% and future nGA development. Results: A total of 15 (14%) eyes from 12 (19%) participants developed nGA. There was no significant difference in global CC FD% at the visit prior to nGA development between eyes that developed nGA and those that did not (P = 0.399). In contrast, CC FD% was significantly higher in superpixels that subsequently developed nGA compared to those that did not (P < 0.001), and a model utilizing CC FD% was significantly better at predicting foci of future nGA development at the superpixel level than a model using drusen volume alone (P ≤ 0.040). Conclusions: This study showed that significant impairments in CC blood flow could be detected locally prior to the development of nGA. These findings add to our understanding of the pathophysiologic changes that occur with atrophy development in AMD.

Visual Sensitivity Loss in Geographic Atrophy: Structure-Function Evaluation Using Defect-Mapping Microperimetry.

Purpose: To examine the structure-function relationship in eyes with geographic atrophy (GA) using defect-mapping microperimetry, a testing strategy optimized to quantify the spatial extent of deep visual sensitivity losses. Methods: Fifty participants with GA underwent defect-mapping microperimetry testing of the central 8°-radius region (208 locations tested once with a 10-decibel stimuli) and fundus autofluorescence imaging in one eye. The GA extent in the corresponding central 8°-radius was derived by manual annotations and image co-registration to examine the global structure-function relationship. The distance of each test location from the GA margin was also derived, and regions defined, to examine the local structure-function relationship. Results: GA extent in the central 8° explained a substantial proportion of variance in the percentage of locations missed (nonresponse) on microperimetry at the global level (R2 = 0.90). At a local level, the probability of missing stimuli at the outer junctional zone (0-500 µm outside the GA margin) and GA margin (probability = 7% and 34%, respectively) was higher than at the outer nonlesional zone (>500 µm outside the GA margin; probability = 2%; P < 0.001 for both). The probability of missing stimuli at the inner junctional zone (0-250 µm inside the GA margin) was also lower than at the inner lesional zone (>250 µm inside the GA margin; probability = 64% and 88%; P < 0.001). Conclusions: This study confirms the expected functional relevance of the region with GA on fundus autofluorescence imaging and underscores the potential effectiveness of defect-mapping microperimetry testing for capturing visual function changes when evaluating new GA treatments.

Deep-learning automated quantification of longitudinal OCT scans demonstrates reduced RPE loss rate, preservation of intact macular area and predictive value of isolated photoreceptor degeneration in geographic atrophy patients receiving C3 inhibition treatment.

OBJECTIVE: To evaluate the role of automated optical coherence tomography (OCT) segmentation, using a validated deep-learning model, for assessing the effect of C3 inhibition on the area of geographic atrophy (GA); the constituent features of GA on OCT (photoreceptor degeneration (PRD), retinal pigment epithelium (RPE) loss and hypertransmission); and the area of unaffected healthy macula.To identify OCT predictive biomarkers for GA growth. METHODS: Post hoc analysis of the FILLY trial using a deep-learning model for spectral domain OCT (SD-OCT) autosegmentation. 246 patients were randomised 1:1:1 into pegcetacoplan monthly (PM), pegcetacoplan every other month (PEOM) and sham treatment (pooled) for 12 months of treatment and 6 months of therapy-free monitoring. Only participants with Heidelberg SD-OCT were included (n=197, single eye per participant).The primary efficacy endpoint was the square root transformed change in area of GA as complete RPE and outer retinal atrophy (cRORA) in each treatment arm at 12 months, with secondary endpoints including RPE loss, hypertransmission, PRD and intact macular area. RESULTS: Eyes treated PM showed significantly slower mean change of cRORA progression at 12 and 18 months (0.151 and 0.277 mm, p=0.0039; 0.251 and 0.396 mm, p=0.039, respectively) and RPE loss (0.147 and 0.287 mm, p=0.0008; 0.242 and 0.410 mm, p=0.00809). PEOM showed significantly slower mean change of RPE loss compared with sham at 12 months (p=0.0313). Intact macular areas were preserved in PM compared with sham at 12 and 18 months (p=0.0095 and p=0.044). PRD in isolation and intact macula areas was predictive of reduced cRORA growth at 12 months (coefficient 0.0195, p=0.01 and 0.00752, p=0.02, respectively) CONCLUSION: The OCT evidence suggests that pegcetacoplan slows progression of cRORA overall and RPE loss specifically while protecting the remaining photoreceptors and slowing the progression of healthy retina to iRORA.

Concerning Syfovre Approval for Geographic Atrophy.

This Viewpoint discusses concerns about Syfovre being approved as a treatment to slow the rate of geographic atrophy.

LIGHTSITE III: 13-Month Efficacy and Safety Evaluation of Multiwavelength Photobiomodulation in Nonexudative (Dry) Age-Related Macular Degeneration Using the Lumithera Valeda Light Delivery System.

PURPOSE: The LIGHTSITE III study evaluated multiwavelength photobiomodulation (PBM) therapy in nonexudative (dry) age-related macular degeneration (AMD) using the LumiThera Valeda Light Delivery System. METHODS: LIGHTSITE III is a randomized, controlled trial to assess the safety and effectiveness of PBM in dry AMD. Subjects were given multiwavelength PBM (590, 660, and 850 nm) or Sham treatment delivered in a series of nine sessions over 3 to 5 weeks every four months over 24 months. Subjects were assessed for efficacy and safety outcomes. Data from the 13-month analysis are presented in this report. RESULTS: A total of 100 subjects (148 eyes) with dry AMD were randomized. LIGHTSITE III met the primary efficacy best-corrected visual acuity endpoint with a significant difference between PBM (n = 91 eyes) and Sham (n = 54 eyes) groups (Between group difference: 2.4 letters (SE 1.15), CI: -4.7 to -0.1, P = 0.02) (PBM alone: 5.4 letters (SE 0.96), CI: 3.5 to 7.3, P < 0.0001; Sham alone: 3.0 letters (SE 1.13), CI: 0.7-5.2, P < 0.0001). The PBM group showed a significant decrease in new onset geographic atrophy ( P = 0.024, Fisher exact test, odds ratio 9.4). A favorable safety profile was observed. CONCLUSION: LIGHTSITE III provides a prospective, randomized, controlled trial showing improved clinical and anatomical outcomes in intermediate dry AMD following PBM therapy.

CHARACTERISTICS PREDICTIVE OF FELLOW-EYE GEOGRAPHIC ATROPHY WITHOUT NEOVASCULARIZATION IN UNILATERAL TYPE 3 MACULAR NEOVASCULARIZATION.

PURPOSE: To evaluate the predictive characteristics of fellow-eye geographic atrophy (GA) without neovascularization in patients with unilateral Type 3 macular neovascularization. METHODS: This retrospective study included 84 patients who were diagnosed with unilateral Type 3 macular neovascularization. Patients who developed fellow-eye neovascularization and those exhibiting GA without neovascularization at the final follow-up were included in the neovascularization and GA groups, respectively. The patient demographics and baseline fellow-eye characteristics were compared between the two groups. RESULTS: The mean follow-up period was 40.5 ± 11.5 months after diagnosis. Patients included in the GA group (n = 28) were significantly older (mean 77.4 ± 5.2 years vs. 74.2 ± 5.8 years, P = 0.016), had significantly thinner subfoveal choroidal thickness (mean 109.4 ± 36.8 µ m vs. 173.1 ± 77.6 µ m, P < 0.001), and had a significantly higher incidence of baseline GA (39.3% vs. 16.1%, P = 0.019) than those included in the neovascularization group (n = 56). In the multivariate analysis, subfoveal choroidal thickness showed a close negative association with the risk of GA rather than neovascularization ( P = 0.004, ß = 0.982, 95% confidence interval = 0.970-0.994). CONCLUSION: In patients with unilateral Type 3 macular neovascularization, older age, the presence of GA, and a thin choroid in the fellow eye were found to be indicative of a higher probability of progression toward fellow-eye GA instead of neovascularization may be potential candidates for future complement inhibitor treatments targeting fellow-eye GA.

Macular lesions associated with age-related macular degeneration in pachydrusen eyes.

BACKGROUND: To investigate the prevalence of macular lesions associated with age-related macular degeneration (AMD) in eyes with pachydrusen. METHODS: Clinical records and multimodal imaging data of patients over 50 years old with drusen or drusenoid deposits were retrospectively assessed, and eyes with pachydrusen were included in this study. The presence of AMD features, including drusen or drusenoid deposits, macular pigmentary abnormalities, geographic atrophy (GA), and macular neovascularization (MNV), were evaluated. RESULTS: Out of 967 eyes of 494 patients with drusen or drusenoid deposits, 330 eyes of 183 patients had pachydrusen (34.1%). The mean age was 66.1 ± 9.3 years, and the subfoveal choroidal thickness (SFCT) was 292.7 ± 100.1 µm. The mean number of pachydrusen per eye was 2.22 ± 1.73. The majority of eyes with pachydrusen had no other drusen or drusenoid deposits (95.2%). Only 16 eyes (4.8%) had other deposits, including soft drusen (10 eyes, 3.0%), cuticular drusen (3 eyes, 0.9%), and reticular pseudodrusen (RPD; 3 eyes, 0.9%). Macular pigmentary abnormalities accompanied pachydrusen in 68 eyes (27.4%). None of the eyes had GA, and 82 eyes (24.8%) had MNV. The majority of MNV was polypoidal choroidal vasculopathy (PCV; 65 eyes, 19.7%), followed by type 1 (10 eyes, 3.0%), type 2 (5 eyes, 1.5%), and type 3 MNV (2 eyes, 0.6%). CONCLUSIONS: Eyes with pachydrusen in Korean population have several characteristic AMD lesions in low frequencies. These findings indicate that pachydrusen might have diagnostic and prognostic values that are different from those of other drusen or drusenoid deposits.

Critical Dependence on Area in Relationship between ARMS2/HTRA1 Genotype and Faster Geographic Atrophy Enlargement: Age-Related Eye Disease Study 2 Report Number 33.

PURPOSE: To analyze ARMS2/HTRA1 as a risk factor for faster geographic atrophy (GA) enlargement according to (1) GA area and (2) contiguous enlargement versus progression to multifocality. DESIGN: Age-Related Eye Disease Study 2 (AREDS2) cohort analysis. PARTICIPANTS: Eyes with GA: 546 eyes of 406 participants. METHODS: Geographic atrophy area was measured from color fundus photographs at annual visits. Mixed-model regression of square root of GA area and proportional hazards regression of progression to multifocality were analyzed by ARMS2 genotype. MAIN OUTCOME MEASURES: Change in square root GA area and progression to multifocality. RESULTS: Geographic atrophy enlargement was significantly faster with ARMS2 risk alleles (P < 0.0001) at 0.224 mm/year (95% CI, 0.195-0.252 mm/year), 0.298 mm/year (95% CI, 0.271-0.324 mm/year), and 0.317 mm/year (95% CI, 0.279-0.355 mm/year), for 0 to 2 risk alleles, respectively. However, a significant interaction (P = 0.011) was observed between genotype and baseline area. In eyes with very small area (< 1.9 mm2), enlargement was significantly faster with ARMS2 risk alleles (P < 0.0001) at 0.193 mm/year (95% CI, 0.162-0.225 mm/year) versus 0.304 mm/year (95% CI, 0.280-0.329 mm/year) for 0 versus 1 to 2 risk alleles, respectively. With moderately small (1.9-3.8 mm2) or medium to large (≥ 3.8 mm2) area, enlargement was not significantly faster with ARMS2 risk alleles (P = 0.66 and P = 0.70, respectively). In nonmultifocal GA, enlargement was significantly faster with ARMS2 risk alleles (P = 0.001) at 0.175 mm/year (95% CI, 0.142-0.209 mm/year), 0.226 mm/year (95% CI, 0.193-0.259 mm/year), and 0.287 mm/year (95% CI, 0.237-0.337 mm/year) with 0 to 2 risk alleles, respectively. ARMS2 genotype was not associated significantly with progression to multifocal GA. CONCLUSIONS: The relationship between ARMS2/HTRA1 genotype and faster GA enlargement depends critically on GA area: risk alleles represent a strong risk factor for faster enlargement only in eyes with very small area. They increase the growth rate more through contiguous enlargement than progression to multifocality. ARMS2/HTRA1 genotype is more important in increasing risk of progression to GA and initial GA enlargement (contiguously) than in subsequent enlargement or progression to multifocality. These findings may explain some discrepancies between previous studies and have implications for both research and clinical practice. FINANCIAL DISCLOSURE(S): Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

A Cost-Effectiveness Analysis of Pegcetacoplan for the Treatment of Geographic Atrophy.

PURPOSE: To evaluate the cost-effectiveness of the treatment of geography atrophy (GA) with intravitreal pegcetacoplan and to identify utility-measurement surrogates. DESIGN: Cost analysis based on data from a published study. SUBJECTS: None; based on data from published sham control compared with 2 treatment groups in the index study. METHODS: Costs were based on 2022 Medicare reimbursement data. Specific outcomes were extrapolated from the DERBY and OAKS trials. Assumptions were made for the lifetime analysis based on a theoretical logistic growth model of the atrophy. OUTCOME MEASURES: Cost, cost utility, cost per quality-adjusted life-year, and cost per area of GA (in US$). RESULTS: The costs to treat GA in every month (EM) and every-other-month (EOM) treatment groups over the 2 years as reported were $70 000 and $34 600, respectively. The costs per area of delaying GA for 2 years in all patients were $87 300/mm2 (EM) and $49 200/mm2 (EOM), and in initially extrafoveal patients, $53 900/mm2 (EM) and $32 100/mm2 (EOM). The costs per day of delaying GA for 2 years were $295 (EM) and $170 (EOM); the marginal cost (EM vs. EOM) per retinal pigment epithelium cell saved was $30. The modeled lifetime costs were $350 000 (EM) and $172 000 (EOM), or $309 000/mm2 (EM) and $180 000 (EOM) /mm2. The modeled time to 95% atrophy at 13 years was delayed by 2.5 years (EM) and 2.1 years (EOM). The costs/quality-adjusted life-year gained based on modeled visual loss with 95% atrophy were $706 000 (EM) and $397 000 (EOM). CONCLUSION: Treatment of GA with intravitreal pegcetacoplan EOM was more cost effective than EM. Treatment of extrafoveal lesions yielded greater utility than the treatment of the entire group. As atrophy progression approaches an upper limit, the marginal cost/benefit ratios increase. FINANCIAL DISCLOSURE(S): Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

Fundus autofluorescence of retinal atrophy progression between fundus flavimaculatus and extensive macular atrophy with pseudo-drusen.

OBJECTIVE: To analyze the progression of macular atrophy in Fundus Flavimaculatus (FFM) versus Extensive Macular Atrophy with Pseudo-drusen (EMAP), using Spectralis® RegionFinder™ tool. METHODS: Retrospective review of patients diagnosed with FFM and EMAP. Ophthalmic imaging features were reviewed by retina specialists for each patient in both eyes. The atrophic zones were measured on fundus autofluorescence acquisitions using the RegionFinder™ tool. RESULTS: FFM group included 16 eyes of 8 patients, whose mean age was 61.42 ± 10.76 years, with a mean 4.54 ± 2.73 years of follow-up. EMAP group contained 16 eyes of 8 patients, whose mean age was 67.81 ± 3.03 years (p = 0.12), with a mean 3.62 ± 2.49 years of follow-up (P = 0.63). The atrophy progression rates were 3.73 ± 6.75 and 0.70 ± 0.98 mm2/year, for EMAP and FFM respectively. The yearly rate of progression of the atrophic areas in EMAP was 5.3 times higher than in FFM (mm2/year) (p = 0.03). CONCLUSION: The progression of the atrophy in eyes with Extensive Macular Atrophy with Pseudo-drusen (EMAP) is significantly more rapid than in eyes with Fundus Flavimaculatus (FFM).

Diagnostic accuracy of community optometrists for age-related macular degeneration using colour fundus photographs: A pilot evaluation.

PURPOSE: The accurate diagnosis of age-related macular degeneration (AMD) represents an important step in delaying and preventing vision loss and achieving optimal patient care. Therefore, this pilot study aimed to estimate the diagnostic accuracy of community optometrists for identifying AMD using colour fundus photographs (CFPs) to support sample size calculations for subsequent definitive studies. METHODS: Five practising community optometrists were invited to classify a total of 1023 CFPs for the (1) presence of AMD, and, if applicable, (2) stage of AMD (early/intermediate/late geographic atrophy/late neovascular AMD). Diagnosis by referral centre clinicians formed the reference standard. Diagnostic accuracy was assessed by the area under the receiver operating characteristic curve (aROC). Sensitivity, specificity, positive and negative predictive values were also calculated. RESULTS: Of the 1023 CFPs included in the study, 226 images were of AMD and 797 images were of other ocular conditions or no abnormal findings. Participating community optometrists had a mean (SD) age of 30.2 (8.9) years, 60.0% (3/5) were female and the mean number of years practising in primary eye care was 5.4 (5.4) years. Community optometrists demonstrated excellent performance for diagnosing AMD, with an aROC of 0.86 (95% CI 0.83 to 0.89), sensitivity of 84.5% (95% CI 79.1 to 89.0) and specificity of 88.0% (95% CI 85.5 to 90.1). The aROC (95% CI) for diagnosing early, intermediate, late geographic atrophy and late neovascular AMD was 0.82 (0.73 to 0.91), 0.76 (0.72 to 0.81), 0.69 (0.49 to 0.90) and 0.55 (0.34 to 0.75), respectively. CONCLUSIONS: These results justify the need for an appropriately powered definitive study to assess community clinicians' diagnostic accuracy for AMD.

Explainable artificial intelligence model for the detection of geographic atrophy using colour retinal photographs.

OBJECTIVE: To develop and validate an explainable artificial intelligence (AI) model for detecting geographic atrophy (GA) via colour retinal photographs. METHODS AND ANALYSIS: We conducted a prospective study where colour fundus images were collected from healthy individuals and patients with retinal diseases using an automated imaging system. All images were categorised into three classes: healthy, GA and other retinal diseases, by two experienced retinologists. Simultaneously, an explainable learning model using class activation mapping techniques categorised each image into one of the three classes. The AI system's performance was then compared with manual evaluations. RESULTS: A total of 540 colour retinal photographs were collected. Data was divided such that 300 images from each class trained the AI model, 120 for validation and 120 for performance testing. In distinguishing between GA and healthy eyes, the model demonstrated a sensitivity of 100%, specificity of 97.5% and an overall diagnostic accuracy of 98.4%. Performance metrics like area under the receiver operating characteristic (AUC-ROC, 0.988) and the precision-recall (AUC-PR, 0.952) curves reinforced the model's robust achievement. When differentiating GA from other retinal conditions, the model preserved a diagnostic accuracy of 96.8%, a precision of 90.9% and a recall of 100%, leading to an F1-score of 0.952. The AUC-ROC and AUC-PR scores were 0.975 and 0.909, respectively. CONCLUSIONS: Our explainable AI model exhibits excellent performance in detecting GA using colour retinal images. With its high sensitivity, specificity and overall diagnostic accuracy, the AI model stands as a powerful tool for the automated diagnosis of GA.