FFA-GPT: an automated pipeline for fundus fluorescein angiography interpretation and question-answer.

Fundus fluorescein angiography (FFA) is a crucial diagnostic tool for chorioretinal diseases, but its interpretation requires significant expertise and time. Prior studies have used Artificial Intelligence (AI)-based systems to assist FFA interpretation, but these systems lack user interaction and comprehensive evaluation by ophthalmologists. Here, we used large language models (LLMs) to develop an automated interpretation pipeline for both report generation and medical question-answering (QA) for FFA images. The pipeline comprises two parts: an image-text alignment module (Bootstrapping Language-Image Pre-training) for report generation and an LLM (Llama 2) for interactive QA. The model was developed using 654,343 FFA images with 9392 reports. It was evaluated both automatically, using language-based and classification-based metrics, and manually by three experienced ophthalmologists. The automatic evaluation of the generated reports demonstrated that the system can generate coherent and comprehensible free-text reports, achieving a BERTScore of 0.70 and F1 scores ranging from 0.64 to 0.82 for detecting top-5 retinal conditions. The manual evaluation revealed acceptable accuracy (68.3%, Kappa 0.746) and completeness (62.3%, Kappa 0.739) of the generated reports. The generated free-form answers were evaluated manually, with the majority meeting the ophthalmologists' criteria (error-free: 70.7%, complete: 84.0%, harmless: 93.7%, satisfied: 65.3%, Kappa: 0.762-0.834). This study introduces an innovative framework that combines multi-modal transformers and LLMs, enhancing ophthalmic image interpretation, and facilitating interactive communications during medical consultation.

Association of age at diagnosis of diabetes with subsequent risk of age-related ocular diseases and vision acuity.

BACKGROUND: The importance of age on the development of ocular conditions has been reported by numerous studies. Diabetes may have different associations with different stages of ocular conditions, and the duration of diabetes may affect the development of diabetic eye disease. While there is a dose-response relationship between the age at diagnosis of diabetes and the risk of cardiovascular disease and mortality, whether the age at diagnosis of diabetes is associated with incident ocular conditions remains to be explored. It is unclear which types of diabetes are more predictive of ocular conditions. AIM: To examine associations between the age of diabetes diagnosis and the incidence of cataract, glaucoma, age-related macular degeneration (AMD), and vision acuity. METHODS: Our analysis was using the UK Biobank. The cohort included 8709 diabetic participants and 17418 controls for ocular condition analysis, and 6689 diabetic participants and 13378 controls for vision analysis. Ocular diseases were identified using inpatient records until January 2021. Vision acuity was assessed using a chart. RESULTS: During a median follow-up of 11.0 years, 3874, 665, and 616 new cases of cataract, glaucoma, and AMD, respectively, were identified. A stronger association between diabetes and incident ocular conditions was observed where diabetes was diagnosed at a younger age. Individuals with type 2 diabetes (T2D) diagnosed at < 45 years [HR (95%CI): 2.71 (1.49-4.93)], 45-49 years [2.57 (1.17-5.65)], 50-54 years [1.85 (1.13-3.04)], or 50-59 years of age [1.53 (1.00-2.34)] had a higher risk of AMD independent of glycated haemoglobin. T2D diagnosed < 45 years [HR (95%CI): 2.18 (1.71-2.79)], 45-49 years [1.54 (1.19-2.01)], 50-54 years [1.60 (1.31-1.96)], or 55-59 years of age [1.21 (1.02-1.43)] was associated with an increased cataract risk. T2D diagnosed < 45 years of age only was associated with an increased risk of glaucoma [HR (95%CI): 1.76 (1.00-3.12)]. HRs (95%CIs) for AMD, cataract, and glaucoma associated with type 1 diabetes (T1D) were 4.12 (1.99-8.53), 2.95 (2.17-4.02), and 2.40 (1.09-5.31), respectively. In multivariable-adjusted analysis, individuals with T2D diagnosed < 45 years of age [ß 95%CI: 0.025 (0.009,0.040)] had a larger increase in LogMAR. The ß (95%CI) for LogMAR associated with T1D was 0.044 (0.014, 0.073). CONCLUSION: The younger age at the diagnosis of diabetes is associated with a larger relative risk of incident ocular diseases and greater vision loss.

Shared genetic aetiology of Alzheimer's disease and age-related macular degeneration by APOC1 and APOE genes.

Background: Alzheimer's disease (AD) and age-related macular degeneration (AMD) share similar pathological features, suggesting common genetic aetiologies between the two. Investigating gene associations between AD and AMD may provide useful insights into the underlying pathogenesis and inform integrated prevention and treatment for both diseases. Methods: A stratified quantile-quantile (QQ) plot was constructed to detect the pleiotropy among AD and AMD based on genome-wide association studies data from 17 008 patients with AD and 30 178 patients with AMD. A Bayesian conditional false discovery rate-based (cFDR) method was used to identify pleiotropic genes. UK Biobank was used to verify the pleiotropy analysis. Biological network and enrichment analysis were conducted to explain the biological reason for pleiotropy phenomena. A diagnostic test based on gene expression data was used to predict biomarkers for AD and AMD based on pleiotropic genes and their regulators. Results: Significant pleiotropy was found between AD and AMD (significant leftward shift on QQ plots). APOC1 and APOE were identified as pleiotropic genes for AD-AMD (cFDR <0.01). Network analysis revealed that APOC1 and APOE occupied borderline positions on the gene co-expression networks. Both APOC1 and APOE genes were enriched on the herpes simplex virus 1 infection pathway. Further, machine learning-based diagnostic tests identified that APOC1, APOE (areas under the curve (AUCs) >0.65) and their upstream regulators, especially ZNF131, ADNP2 and HINFP, could be potential biomarkers for both AD and AMD (AUCs >0.8). Conclusion: In this study, we confirmed the genetic pleiotropy between AD and AMD and identified APOC1 and APOE as pleiotropic genes. Further, the integration of multiomics data identified ZNF131, ADNP2 and HINFP as novel diagnostic biomarkers for AD and AMD.

Myopia and Rate of Peripapillary Retinal Nerve Fiber Layer Thickness in Diabetic Patients Without Retinopathy: A 2-Year Longitudinal Study.

PURPOSE: The aim of this study was to investigate the association between myopia and longitudinal changes in peripapillary retinal nerve fiber layer (pRNFL) thickness in type 2 diabetic patients without diabetic retinopathy (DR). METHODS: A total of 1069 participants with a median follow-up time of 1.9 years were included in this study. The participants were categorized into four groups based on the presence of myopia (≤ -0.5 diopter [D]) and diabetes without DR, including a control group (n = 412), diabetes group (n = 416), myopia group (n = 115), and diabetes + myopia group (n = 126). Peripapillary average and sectoral RNFL measurements were obtained using 6 × 6 mm swept-source optical coherence tomography (SS-OCT) scans centered at the optic disc. The change rate of pRNFL, adjusted for age and sex, was calculated and compared among the four groups to investigate the impact of myopia and diabetes. RESULTS: The baseline estimated pRNFL thickness after adjustment for covariates was 113.7 µm, 116.2 µm, 108.0 µm, and 105.6 µm in the control, diabetes, myopia, and diabetes + myopia group, respectively (diabetes > control > myopia = diabetes + myopia, p < 0.001). The respective average pRNFL loss in the four groups was -0.48 µm/year, -1.11 µm/year, -1.23 µm/year, and -2.62 µm/year (all p < 0.01). The diabetes + myopia group exhibited a greater rate of average pRNFL reduction compared to the other groups (all p < 0.001). Multivariate analysis using a linear mixed-effects model showed that age, diabetes, axial length (AL), and baseline pRNFL thickness were significantly associated with the rate of average pRNFL reduction. CONCLUSIONS: The diabetes group showed a faster rate of average pRNFL thickness reduction compared to healthy controls, regardless of the presence of myopia. The average pRNFL thickness decreased more rapidly when diabetes and myopia were present simultaneously than in the individual diabetes or myopia group. Both diabetes and myopia were associated with accelerated pRNFL loss.

Role of lens in early refractive development: evidence from a large cohort of Chinese children.

AIMS: To document longitudinal changes in spherical equivalent refraction (SER) and related biometric factors during early refractive development. METHODS: This was a prospective cohort study of Chinese children, starting in 2018 with annual follow-ups. At each visit, children received cycloplegic autorefraction and ocular biometry measurements. Lens power (LP) was calculated using Bennett's formula. Children were divided into eight groups based on baseline age: the 3-year-old (n=426, 49.77% girls), 4-year-old (n=834, 47.36% girls), 6-year-old (n=292, 46.58% girls), 7-year-old (n=964, 43.46% girls), 9-year-old (n=981, 46.18% girls), 10-year-old (n=1181, 46.32% girls), 12-year-old (n=504, 49.01%) and 13-year-old (n=644, 42.70%) age groups. RESULTS: This study included right-eye data from 5826 children. The 3-year-old and 4-year-old age groups demonstrated an inflection point in longitudinal SER changes at a mild hyperopic baseline SER (+1 to +2 D), with children with more myopic SER showing hyperopic refractive shifts while those with more hyperopic SER showing myopic shifts. The hyperopic shift in SER was mainly attributed to rapid LP loss and was rarely seen in the older age groups. Axial elongation accelerated in the premyopia stage, accompanied by a partially counter-balancing acceleration of LP loss. For children aged 3-7 years, those with annual SER changes <0.25 D were all mildly hyperopic at baseline (mean: 1.23 D, 95% CI 1.20 to 1.27 D). CONCLUSION: Our findings suggest that during early refractive development, refractions cluster around or above +1.00 D. There is a pushback process in which increases in the rate of LP occur in parallel with increases in axial elongation.

Seasonal Variations in Vitamin D Levels and the Incident Dementia Among Older Adults Aged ≥60 Years in the UK Biobank.

Background: Limited knowledge exists regarding the association between dementia incidence and vitamin D insufficiency/deficiency across seasons. Objective: This study aimed to evaluate the impact of seasonal serum vitamin D (25(OH)D) levels on dementia and its subtypes, considering potential modifiers. Methods: We analyzed 193,003 individuals aged 60-73 at baseline (2006-2010) from the UK Biobank cohort, with follow-up until 2018. 25(OH)D were measured at baseline, and incident dementia cases were identified through hospital records, death certificates, and self-reports. Results: Out of 1,874 documented all-cause dementia cases, the median follow-up duration was 8.9 years. Linear and nonlinear associations between 25(OH)D and dementia incidence across seasons were observed. In multivariable-adjusted analysis, 25(OH)D deficiency was associated with a 1.5-fold (95% CIs: 1.2-2.0), 2.2-fold (1.5-3.0), 2.0-fold (1.5-2.7), and 1.7-fold (1.3-2.3) increased incidence of all-cause dementia in spring, summer, autumn, and winter, respectively. Adjusting for seasonal variations, 25(OH)D insufficiency and deficiency were associated with a 1.3-fold (1.1-1.4) and 1.8-fold (1.6-2.2) increased dementia incidence, respectively. This association remained significant across subgroups, including baseline age, gender, and education levels. Furthermore, 25(OH)D deficiency was associated with a 1.4-fold (1.1-1.8) and 1.5-fold (1.1-2.0) higher incidence of Alzheimer's disease and vascular dementia, respectively. These associations remained significant across all subgroups. Conclusions: 25(OH)D deficiency is associated with an increased incidence of dementia and its subtypes throughout the year.

Longitudinal Changes in Choroidal Thickness Varied With Refractive Progression in Myopic and Non-Myopic Children: A Two-Year Cohort Study.

Purpose: To evaluate the longitudinal changes in subfoveal choroidal thickness (SFCT) in children with different refractive status. Methods: A total of 2290 children 3 to 14 years old who attended the first year of kindergarten (G0), first year of primary school (G1), fourth year of primary school (G4), or first year of junior high school (G7) in Guangzhou, China, were recruited and followed up for 2 years. All participants received cycloplegic autorefraction, axial length measurement and SFCT measurement using a CIRRUS HD-OCT device. Children were divided into groups of persistent non-myopia (PNM), persistent myopia (PM), or newly developed myopia (NDM). Children in the PNM and PM groups were further divided into subgroups of stable refraction (absolute mean annual spherical equivalent refraction [SER] change < 0.5 D) and refractive progression (absolute mean annual SER change ≥ 0.5 D). Results: The mean ± SD ages for the G1 to G7 cohorts were 3.89 ± 0.30, 6.79 ± 0.47, 9.71 ± 0.34, and 12.54 ± 0.38, years, respectively. SFCT consistently decreased in the NDM group across the G1 to G7 cohorts (all P < 0.001) and exhibited variability across different age cohorts in the PNM and PM groups. Further subgroup analysis revealed significant thickening of SFCT in the PNM-stable group among the G0, G1, and G7 cohorts (all P < 0.05), whereas it remained stable among all cohorts in the PM-stable group (all P > 0.05). Conversely, SFCT exhibited thinning in the G4 and G7 cohorts in the PM-progressive group (both P < 0.01) and for the entire cohort of children in the PNM-progressive group (P = 0.012). Conclusions: SFCT increased in nonmyopic children with stable refraction, remained stable in myopic children maintained stable refraction, and decreased in those with refractive progression, whether they were myopic or not.

Diagnosing glaucoma in primary eye care and the role of Artificial Intelligence applications for reducing the prevalence of undetected glaucoma in Australia.

Glaucoma is the commonest cause of irreversible blindness worldwide, with over 70% of people affected remaining undiagnosed. Early detection is crucial for halting progressive visual impairment in glaucoma patients, as there is no cure available. This narrative review aims to: identify reasons for the significant under-diagnosis of glaucoma globally, particularly in Australia, elucidate the role of primary healthcare in glaucoma diagnosis using Australian healthcare as an example, and discuss how recent advances in artificial intelligence (AI) can be implemented to improve diagnostic outcomes. Glaucoma is a prevalent disease in ageing populations and can have improved visual outcomes through appropriate treatment, making it essential for general medical practice. In countries such as Australia, New Zealand, Canada, USA, and the UK, optometrists serve as the gatekeepers for primary eye care, and glaucoma detection often falls on their shoulders. However, there is significant variation in the capacity for glaucoma diagnosis among eye professionals. Automation with Artificial Intelligence (AI) analysis of optic nerve photos can help optometrists identify high-risk changes and mitigate the challenges of image interpretation rapidly and consistently. Despite its potential, there are significant barriers and challenges to address before AI can be deployed in primary healthcare settings, including external validation, high quality real-world implementation, protection of privacy and cybersecurity, and medico-legal implications. Overall, the incorporation of AI technology in primary healthcare has the potential to reduce the global prevalence of undiagnosed glaucoma cases by improving diagnostic accuracy and efficiency.

Impact of Acute Ocular Hypertension on Retinal Ganglion Cell Loss in Mice.

Purpose: To assess the correlation between intraocular pressure (IOP) levels and retinal ganglion cell (RGC) loss across different fixed-duration episodes of acute ocular hypertension (AOH). Methods: AOH was induced in Thy1-YFP-H transgenic mice by inserting a needle connected to a saline solution container into the anterior chamber. Thirty-one groups were tested, each comprising three to five mice exposed to IOP levels ranging from 50 to 110 mm Hg in 5/10 mm Hg increments for 60/90/120 minutes and a sham control group. The YFP-expressing RGCs were quantified by confocal scanning laser ophthalmoscopy, whereas peripapillary ganglion cell complex thickness was measured using spectral-domain optical coherence tomography. Changes in RGC count and GCCT were determined from values measured 30 days after AOH relative to baseline (before AOH). Results: In the 60-minute AOH groups, RGC loss varied even when IOP was increased up to 110 mm Hg (36.8%-68.2%). However, for longer durations (90 and 120 minutes), a narrow range of IOP levels (60-70 mm Hg for 90-minute duration; 55-65 mm Hg for 120-minute duration) produced a significant difference in RGC loss, ranging from <25% to >90%. Additionally, loss of YFP-expressing RGCs was comparable to that of total RGCs in the same retinas. Conclusions: Reproducible RGC loss during AOH depends on precise durations and IOP thresholds. In the current study, the optimal choice is an AOH protocol set at 70 mm Hg for a duration of 90 minutes. Translational Relevance: This study can assist in determining the optimal duration and intensity of IOP for the effective utilization of AOH models.

Real-world application of a smartphone-based visual acuity test (WHOeyes) with automatic distance calibration.

BACKGROUND: To develop and assess the usability of a smartphone-based visual acuity (VA) test with an automatic distance calibration (ADC) function, the iOS version of WHOeyes. METHODS: The WHOeyes was an upgraded version with a distinct feature of ADC of an existing validated VA testing app called V@home. Three groups of Chinese participants with different ages (≤20, 20-40, >40 years) were recruited for distance and near VA testing using both an Early Treatment Diabetic Retinopathy Study (ETDRS) chart and the WHOeyes. The ADC function would determine the testing distance. Infrared rangefinder was used to determine the testing distance for the ETDRS, and actual testing distance for the WHOeyes. A questionnaire-based interview was administered to assess the satisfaction. RESULTS: The actual testing distance determined by the WHOeyes ADC showed an overall good agreement with the desired testing distance in all three age groups (p>0.50). Regarding the distance and near VA testing, the accuracy of WHOeyes was equivalent to ETDRS. The mean difference between the WHOeyes and ETDRS ranged from -0.084 to 0.012 logMAR, and the quadratic weighted kappa (QWK) values were >0.75 across all groups. The test-retest reliability of WHOeyes was high for both near and distance VA, with a mean difference ranging from -0.040 to 0.004 logMAR and QWK all >0.85. The questionnaire revealed an excellent user experience and acceptance of WHOeyes. CONCLUSIONS: WHOeyes could provide accurate measurement of the testing distance as well as the distance and near VA when compared to the gold standard ETDRS chart.

Diagnostic Accuracy of Artificial Intelligence-Based Automated Diabetic Retinopathy Screening in Real-World Settings: A Systematic Review and Meta-Analysis.

PURPOSE: To evaluate the diagnostic accuracy of artificial intelligence (AI)-based automated diabetic retinopathy (DR) screening in real-world settings. DESIGN: Systematic review and meta-analysis METHODS: We conducted a systematic review of relevant literature from January 2012 to August 2022 using databases including PubMed, Scopus and Web of Science. The quality of studies was evaluated using Quality Assessment for Diagnostic Accuracy Studies 2 (QUADAS-2) checklist. We calculated pooled accuracy, sensitivity, specificity, and diagnostic odds ratio (DOR) as summary measures. The study protocol was registered with the International Prospective Register of Systematic Reviews (PROSPERO - CRD42022367034). RESULTS: We included 34 studies which utilized AI algorithms for diagnosing DR based on real-world fundus images. Quality assessment of these studies indicated a low risk of bias and low applicability concern. Among gradable images, the overall pooled accuracy, sensitivity, specificity, and DOR were 81%, 94% (95% CI: 92.0-96.0), 89% (95% CI: 85.0-92.0) and 128 (95% CI: 80-204) respectively. Sub-group analysis showed that, when acceptable quality imaging could be obtained, non-mydriatic fundus images had a better DOR of 143 (95% CI: 82-251) and studies using 2 field images had a better DOR of 161 (95% CI 74-347). Our meta-regression analysis revealed a statistically significant association between DOR and variables such as the income status, and the type of fundus camera. CONCLUSION: Our findings indicate that AI algorithms have acceptable performance in screening for DR using fundus images compared to human graders. Implementing a fundus camera with AI-based software has the potential to assist ophthalmologists in reducing their workload and improving the accuracy of DR diagnosis.

Evaluation of the Observational Associations and Shared Genetics Between Glaucoma With Depression and Anxiety.

Purpose: Glaucoma, a leading cause of blindness worldwide, is suspected to exhibit a notable association with psychological disturbances. This study aimed to investigate epidemiological associations and explore shared genetic architecture between glaucoma and mental traits, including depression and anxiety. Methods: Multivariable logistic regression and Cox proportional hazards regression models were employed to investigate longitudinal associations based on UK Biobank. A stepwise approach was used to explore the shared genetic architecture. First, linkage disequilibrium score regression inferred global genetic correlations. Second, MiXeR analysis quantified the number of shared causal variants. Third, specific shared loci were detected through conditional/conjunctional false discovery rate (condFDR/conjFDR) analysis and characterized for biological insights. Finally, two-sample Mendelian randomization (MR) was conducted to investigate bidirectional causal associations. Results: Glaucoma was significantly associated with elevated risks of hospitalized depression (hazard ratio [HR] = 1.54; 95% confidence interval [CI], 1.01-2.34) and anxiety (HR = 2.61; 95% CI, 1.70-4.01) compared to healthy controls. Despite the absence of global genetic correlations, MiXeR analysis revealed 300 variants shared between glaucoma and depression, and 500 variants shared between glaucoma and anxiety. Subsequent condFDR/conjFDR analysis discovered 906 single-nucleotide polymorphisms (SNPs) jointly associated with glaucoma and depression and two associated with glaucoma and anxiety. The MR analysis did not support robust causal associations but indicated the existence of pleiotropic genetic variants influencing both glaucoma and depression. Conclusions: Our study enhances the existing epidemiological evidence and underscores the polygenic overlap between glaucoma and mental traits. This observation suggests a correlation shaped by pleiotropic genetic variants rather than being indicative of direct causal relationships.

ICGA-GPT: report generation and question answering for indocyanine green angiography images.

BACKGROUND: Indocyanine green angiography (ICGA) is vital for diagnosing chorioretinal diseases, but its interpretation and patient communication require extensive expertise and time-consuming efforts. We aim to develop a bilingual ICGA report generation and question-answering (QA) system. METHODS: Our dataset comprised 213 129 ICGA images from 2919 participants. The system comprised two stages: image-text alignment for report generation by a multimodal transformer architecture, and large language model (LLM)-based QA with ICGA text reports and human-input questions. Performance was assessed using both qualitative metrics (including Bilingual Evaluation Understudy (BLEU), Consensus-based Image Description Evaluation (CIDEr), Recall-Oriented Understudy for Gisting Evaluation-Longest Common Subsequence (ROUGE-L), Semantic Propositional Image Caption Evaluation (SPICE), accuracy, sensitivity, specificity, precision and F1 score) and subjective evaluation by three experienced ophthalmologists using 5-point scales (5 refers to high quality). RESULTS: We produced 8757 ICGA reports covering 39 disease-related conditions after bilingual translation (66.7% English, 33.3% Chinese). The ICGA-GPT model's report generation performance was evaluated with BLEU scores (1-4) of 0.48, 0.44, 0.40 and 0.37; CIDEr of 0.82; ROUGE of 0.41 and SPICE of 0.18. For disease-based metrics, the average specificity, accuracy, precision, sensitivity and F1 score were 0.98, 0.94, 0.70, 0.68 and 0.64, respectively. Assessing the quality of 50 images (100 reports), three ophthalmologists achieved substantial agreement (kappa=0.723 for completeness, kappa=0.738 for accuracy), yielding scores from 3.20 to 3.55. In an interactive QA scenario involving 100 generated answers, the ophthalmologists provided scores of 4.24, 4.22 and 4.10, displaying good consistency (kappa=0.779). CONCLUSION: This pioneering study introduces the ICGA-GPT model for report generation and interactive QA for the first time, underscoring the potential of LLMs in assisting with automated ICGA image interpretation.

Cross-modality Labeling Enables Noninvasive Capillary Quantification as a Sensitive Biomarker for Assessing Cardiovascular Risk.

Purpose: We aim to use fundus fluorescein angiography (FFA) to label the capillaries on color fundus (CF) photographs and train a deep learning model to quantify retinal capillaries noninvasively from CF and apply it to cardiovascular disease (CVD) risk assessment. Design: Cross-sectional and longitudinal study. Participants: A total of 90732 pairs of CF-FFA images from 3893 participants for segmentation model development, and 49229 participants in the UK Biobank for association analysis. Methods: We matched the vessels extracted from FFA and CF, and used vessels from FFA as labels to train a deep learning model (RMHAS-FA) to segment retinal capillaries using CF. We tested the model's accuracy on a manually labeled internal test set (FundusCapi). For external validation, we tested the segmentation model on 7 vessel segmentation datasets, and investigated the clinical value of the segmented vessels in predicting CVD events in the UK Biobank. Main Outcome Measures: Area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, and specificity for segmentation. Hazard ratio (HR; 95% confidence interval [CI]) for Cox regression analysis. Results: On the FundusCapi dataset, the segmentation performance was AUC = 0.95, accuracy = 0.94, sensitivity = 0.90, and specificity = 0.93. Smaller vessel skeleton density had a stronger correlation with CVD risk factors and incidence (P < 0.01). Reduced density of small vessel skeletons was strongly associated with an increased risk of CVD incidence and mortality for women (HR [95% CI] = 0.91 [0.84-0.98] and 0.68 [0.54-0.86], respectively). Conclusions: Using paired CF-FFA images, we automated the laborious manual labeling process and enabled noninvasive capillary quantification from CF, supporting its potential as a sensitive screening method for identifying individuals at high risk of future CVD events. Financial Disclosures: Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

Leading determinants of incident dementia among individuals with and without the apolipoprotein E ε4 genotype: a retrospective cohort study.

BACKGROUND: Little is known regarding the leading risk factors for dementia/Alzheimer's disease (AD) in individuals with and without APOE4. The identification of key risk factors for dementia/Alzheimer's disease (AD) in individuals with and without the APOE4 gene is of significant importance in global health. METHODS: Our analysis included 110,354 APOE4 carriers and 220,708 age- and sex-matched controls aged 40-73 years at baseline (between 2006-2010) from UK Biobank. Incident dementia was ascertained using hospital inpatient, or death records until January 2021. Individuals of non-European ancestry were excluded. Furthermore, individuals without medical record linkage were excluded from the analysis. Moderation analysis was tested for 134 individual factors. RESULTS: During a median follow-up of 11.9 years, 4,764 cases of incident all-cause dementia and 2065 incident AD cases were documented. Hazard ratios (95% CIs) for all-cause dementia and AD associated with APOE4 were 2.70(2.55-2.85) and 3.72(3.40-4.07), respectively. In APOE4 carriers, the leading risk factors for all-cause dementia included low self-rated overall health, low household income, high multimorbidity risk score, long-term illness, high neutrophil percentage, and high nitrogen dioxide air pollution. In non-APOE4 carriers, the leading risk factors included high multimorbidity risk score, low overall self-rated health, low household income, long-term illness, high microalbumin in urine, high neutrophil count, and low greenspace percentage. Population attributable risk for these individual risk factors combined was 65.1%, and 85.8% in APOE4 and non-APOE4 carriers, respectively. For 20 risk factors including multimorbidity risk score, unhealthy lifestyle habits, and particulate matter air pollutants, their associations with incident dementia were stronger in non-APOE4 carriers. For only 2 risk factors (mother's history of dementia, low C-reactive protein), their associations with incident all-cause dementia were stronger in APOE4 carriers. CONCLUSIONS: Our findings provide evidence for personalized preventative approaches to dementia/AD in APOE4 and non-APOE4 carriers. A mother's history of dementia and low levels of C-reactive protein were more important risk factors of dementia in APOE4 carriers whereas leading risk factors including unhealthy lifestyle habits, multimorbidity risk score, inflammation and immune-related markers were more predictive of dementia in non-APOE4 carriers.

Economic evaluation for medical artificial intelligence: accuracy vs. cost-effectiveness in a diabetic retinopathy screening case.

Artificial intelligence (AI) models have shown great accuracy in health screening. However, for real-world implementation, high accuracy may not guarantee cost-effectiveness. Improving AI's sensitivity finds more high-risk patients but may raise medical costs while increasing specificity reduces unnecessary referrals but may weaken detection capability. To evaluate the trade-off between AI model performance and the long-running cost-effectiveness, we conducted a cost-effectiveness analysis in a nationwide diabetic retinopathy (DR) screening program in China, comprising 251,535 participants with diabetes over 30 years. We tested a validated AI model in 1100 different diagnostic performances (presented as sensitivity/specificity pairs) and modeled annual screening scenarios. The status quo was defined as the scenario with the most accurate AI performance. The incremental cost-effectiveness ratio (ICER) was calculated for other scenarios against the status quo as cost-effectiveness metrics. Compared to the status quo (sensitivity/specificity: 93.3%/87.7%), six scenarios were cost-saving and seven were cost-effective. To achieve cost-saving or cost-effective, the AI model should reach a minimum sensitivity of 88.2% and specificity of 80.4%. The most cost-effective AI model exhibited higher sensitivity (96.3%) and lower specificity (80.4%) than the status quo. In settings with higher DR prevalence and willingness-to-pay levels, the AI needed higher sensitivity for optimal cost-effectiveness. Urban regions and younger patient groups also required higher sensitivity in AI-based screening. In real-world DR screening, the most accurate AI model may not be the most cost-effective. Cost-effectiveness should be independently evaluated, which is most likely to be affected by the AI's sensitivity.

Population impact and cost-effectiveness of artificial intelligence-based diabetic retinopathy screening in people living with diabetes in Australia: a cost effectiveness analysis.

Background: We aimed to evaluate the cost-effectiveness of an artificial intelligence-(AI) based diabetic retinopathy (DR) screening system in the primary care setting for both non-Indigenous and Indigenous people living with diabetes in Australia. Methods: We performed a cost-effectiveness analysis between January 01, 2022 and August 01, 2023. A decision-analytic Markov model was constructed to simulate DR progression in a population of 1,197,818 non-Indigenous and 65,160 Indigenous Australians living with diabetes aged ≥20 years over 40 years. From a healthcare provider's perspective, we compared current practice to three primary care AI-based screening scenarios-(A) substitution of current manual grading, (B) scaling up to patient acceptance level, and (C) achieving universal screening. Study results were presented as incremental cost-effectiveness ratio (ICER), benefit-cost ratio (BCR), and net monetary benefits (NMB). A Willingness-to-pay (WTP) threshold of AU$50,000 per quality-adjusted life year (QALY) and a discount rate of 3.5% were adopted in this study. Findings: With the status quo, the non-Indigenous diabetic population was projected to develop 96,269 blindness cases, resulting in AU$13,039.6 m spending on DR screening and treatment during 2020-2060. In comparison, all three intervention scenarios were effective and cost-saving. In particular, if a universal screening program was to be implemented (Scenario C), it would prevent 38,347 blindness cases, gain 172,090 QALYs and save AU$595.8 m, leading to a BCR of 3.96 and NMB of AU$9,200 m. Similar findings were also reported in the Indigenous population. With the status quo, 3,396 Indigenous individuals would develop blindness, which would cost the health system AU$796.0 m during 2020-2060. All three intervention scenarios were cost-saving for the Indigenous population. Notably, universal AI-based DR screening (Scenario C) would prevent 1,211 blindness cases and gain 9,800 QALYs in the Indigenous population, leading to a saving of AU$19.2 m with a BCR of 1.62 and NMB of AU$509 m. Interpretation: Our findings suggest that implementing AI-based DR screening in primary care is highly effective and cost-saving in both Indigenous and non-Indigenous populations. Funding: This project received grant funding from the Australian Government: the National Critical Research Infrastructure Initiative, Medical Research Future Fund (MRFAI00035) and the NHMRC Investigator Grant (APP1175405). The contents of the published material are solely the responsibility of the Administering Institution, a participating institution or individual authors and do not reflect the views of the NHMRC. This work was supported by the Global STEM Professorship Scheme (P0046113), the Fundamental Research Funds of the State Key Laboratory of Ophthalmology, Project of Investigation on Health Status of Employees in Financial Industry in Guangzhou, China (Z012014075). The Centre for Eye Research Australia receives Operational Infrastructure Support from the Victorian State Government. W.H. is supported by the Melbourne Research Scholarship established by the University of Melbourne. The funding source had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Using choroidal thickness to detect myopic macular degeneration.

AIM: To explore the usage of choroidal thickness measured by swept-source optical coherence tomography (SS-OCT) to detect myopic macular degeneration (MMD) in high myopic participants. METHODS: Participants with bilateral high myopia (≤-6 diopters) were recruited from a subset of the Guangzhou Zhongshan Ophthalmic Center-Brien Holden Vision Institute High Myopia Cohort Study. SS-OCT was performed to determine the choroidal thickness, and myopic maculopathy was graded by the International Meta-Analysis for Pathologic Myopia (META-PM) Classification. Presence of MMD was defined as META-PM category 2 or above. RESULTS: A total of 568 right eyes were included for analysis. Eyes with MMD (n=106, 18.7%) were found to have older age, longer axial lengths (AL), higher myopic spherical equivalents (SE), and reduced choroidal thickness in each Early Treatment Diabetic Retinopathy Study (ETDRS) grid sector (P<0.001). The area under the receiver operating characteristic (ROC) curves (AUC) for subfoveal choroidal thickness (0.907) was greater than that of the model, including age, AL, and SE at 0.6249, 0.8208, and 0.8205, respectively. The choroidal thickness of the inner and outer nasal sectors was the most accurate indicator of MMD (AUC of 0.928 and 0.923, respectively). An outer nasal sector choroidal thickness of less than 74 µm demonstrated the highest odds of predicting MMD (OR=33.8). CONCLUSION: Choroidal thickness detects the presence of MMD with high agreement, particularly of the inner and outer nasal sectors of the posterior pole, which appears to be a biometric parameter more precise than age, AL, or SE.

Translating color fundus photography to indocyanine green angiography using deep-learning for age-related macular degeneration screening.

Age-related macular degeneration (AMD) is the leading cause of central vision impairment among the elderly. Effective and accurate AMD screening tools are urgently needed. Indocyanine green angiography (ICGA) is a well-established technique for detecting chorioretinal diseases, but its invasive nature and potential risks impede its routine clinical application. Here, we innovatively developed a deep-learning model capable of generating realistic ICGA images from color fundus photography (CF) using generative adversarial networks (GANs) and evaluated its performance in AMD classification. The model was developed with 99,002 CF-ICGA pairs from a tertiary center. The quality of the generated ICGA images underwent objective evaluation using mean absolute error (MAE), peak signal-to-noise ratio (PSNR), structural similarity measures (SSIM), etc., and subjective evaluation by two experienced ophthalmologists. The model generated realistic early, mid and late-phase ICGA images, with SSIM spanned from 0.57 to 0.65. The subjective quality scores ranged from 1.46 to 2.74 on the five-point scale (1 refers to the real ICGA image quality, Kappa 0.79-0.84). Moreover, we assessed the application of translated ICGA images in AMD screening on an external dataset (n = 13887) by calculating area under the ROC curve (AUC) in classifying AMD. Combining generated ICGA with real CF images improved the accuracy of AMD classification with AUC increased from 0.93 to 0.97 (P < 0.001). These results suggested that CF-to-ICGA translation can serve as a cross-modal data augmentation method to address the data hunger often encountered in deep-learning research, and as a promising add-on for population-based AMD screening. Real-world validation is warranted before clinical usage.