Richard B. Rosen

Background/Objectives: Diabetic retinopathy (DR) is a leading cause of vision loss worldwide, and its pathology is driven by chronic hyperglycemia, elevated oxidative stress, inflammation, and microvascular dysfunction. Polyunsaturated fatty acids (PUFAs), particularly omega-3 and omega-6 fatty acids, influence inflammatory and angiogenic pathways, but their role in DR prevention and progression remains uncertain. This integrative review synthesizes current evidence on the effects of PUFAs in DR. Methods: A comprehensive literature search was conducted and eligible studies included human cell culture, animal model, observational, Mendelian randomization, and interventional studies. A total of 44 studies meeting inclusion criteria were systematically analyzed and categorized by study design. Results: Human trials with whole-diet/complex supplements suggest that higher intake or circulating levels of omega-3 PUFAs, particularly docosahexaenoic acid (DHA), are associated with reduced risk and severity of DR. A Mediterranean diet enriched with omega-3 sources demonstrated consistent protective associations. However, randomized controlled trials using isolated omega-3 supplementation reported mixed findings, with some showing no significant effect on disease progression (e.g., ASCEND). Mechanistic studies indeed support biological plausibility, demonstrating that omega-3 PUFAs reduce oxidative stress, inhibit inflammatory signaling, and modulate angiogenesis by attenuating vascular endothelial growth factor activity. Emerging evidence also suggests the potential protective role of certain omega-6 fatty acids, though findings remain heterogeneous. Conclusions: PUFAs, particularly omega-3 fatty acids, may play a protective role in DR through anti-inflammatory, antioxidant, and anti-angiogenic mechanisms. These outcomes appear influenced by dietary context, dosage, and disease stage, highlighting the need for additional long-term, well-designed studies.

BACKGROUND A new class of AI "decision models" returns a probability for every option in a single pass and is marketed as calibrated, which invites use in clinical triage. Whether those probabilities hold up on clinical diagnosis has not been independently tested. OBJECTIVE To test the diagnostic accuracy, calibration, and probabilistic coherence of a decision model (Jev 1.13, TypeSafe AI) against frontier language models, with and without reasoning. METHODS In a pre-registered study (OSF, registered 26 September 2026), we compared Jev with Claude Opus 5.5 (adaptive reasoning, always on) and Claude Opus 5 (reasoning disabled) on three case sets: the 45 Semigran symptom-checker vignettes, 1,000 stratified synthetic patients from DDXPlus with reference differentials, and 80 simulated ophthalmic complaints (40 diagnoses at two levels of detail). Each model returned a probability for every diagnosis on a fixed list. Outcomes were top-1 and top-3 accuracy, expected calibration error (ECE), Brier score, the rate of exactly zero probability on the true diagnosis, sensitivity to option order, behavior on uninformative inputs, and behavior when the true diagnosis was absent. RESULTS On DDXPlus, top-1 accuracy was 72.0% for Jev, 79.5% for Opus 5 without reasoning, and 87.4% for Opus 5.5 (both comparisons with Jev P<.001). Jev assigned exactly zero probability to the true diagnosis in 62 of 1,000 cases (6.2%); neither comparator did so in any case. Calibration error did not differ between Jev and the registered comparator (ECE difference 0.007, 95% CI -0.026 to 0.038). Offered an explicit "none of these" option, Jev declined to diagnose all 30 uninformative inputs; without that option, it assigned a mean 58% to a single diagnosis. Order sensitivity was small in all models. Given the option list with no patient information, Jev placed up to 96% on a single diagnosis, while the comparators spread probability almost evenly. On the ophthalmic complaints, accuracy did not differ (Jev 88.8%, Opus 5 91.2%, Opus 5.5 88.8%), and Jev was better calibrated than the registered comparator (ECE difference -0.081, 95% CI -0.124 to -0.018). CONCLUSIONS The decision model matched frontier models on short, distinctive ophthalmic complaints but fell well behind on multi-symptom synthetic patients, including behind a comparator without reasoning, and there it excluded the correct diagnosis outright in about 1 in 16 cases. Its probabilities signaled uncertainty only when the question offered a way to abstain. Calibration claims for clinical decision models should be tested on clinical tasks, with explicit abstention options and zero-probability rates reported. CLINICALTRIAL Open Science Framework (OSF) Registries osf.io/fpmg3; https://osf.io/fpmg3

UNSTRUCTURED As artificial intelligence (AI) systems approach and sometimes exceed human diagnostic performance on medical imaging tasks, a foundational question for clinical informatics is what sustains the safety value of human oversight once individual accuracy is no longer the differentiator. We argue that the operative parameter is not accuracy but decorrelation: the degree to which human and AI errors differ on the same cases. The classifier-combination literature establishes that pairwise error correlation (ρ) governs whether additional overseers meaningfully extend the vetting ceiling; when ρ approaches one, the effective number of independent overseers collapses to one regardless of how many systems are nominally deployed. We propose that ρ should be treated as a first-class quantity in the informatics of algorithmic safety. Drawing on published agreement data across 22 modality-task combinations in ophthalmology, radiology, and pathology, we offer illustrative cross-domain patterns suggesting a qualitative association between interpretive complexity (an author-proposed heuristic we call interpretive dimensionality) and concordance, and we observe that concordance appears to rise within fixed tasks over time as AI absorbs the relevant feature space. We call this erosion decorrelation wasting and contrast it with decorrelation injection, in which a novel substrate or task specification transiently lowers ρ. These patterns motivate a multi-lever Strategic Cognitive Reserve combining (1) maintaining clinicians in the diagnostic loop, (2) redesigning human-AI interaction, (3) exploiting the broader clinical context window, and (4) periodically introducing novel substrates. The framework is hypothesis-generating rather than prescriptive; we close with testable predictions requiring paired case-level error data. Multimedia Appendices 1 and 2 provide the task-level source data and the formal dynamics.

Background/Objectives: The aim of this study was to investigate the effects of intraocular pressure (IOP)-lowering drops on the sublayers of the human tear film as assessed by a novel nanometer-resolution Tear Film Imager (TFI, AdOM, Israel). Methods: In a prospective, cross-sectional study, 98 eyes from 56 adult human subjects were imaged using the TFI. The dataset included data from 18 eyes from 12 subjects treated with preserved IOP-lowering drops and 80 eyes from 44 control subjects not under ocular hypotensive therapy. Subjects in the IOP treatment group used a variety of IOP-lowering medications, including prostaglandin analogs, beta-blockers, carbonic anhydrase inhibitors, alpha agonists, and combination drops. A linear mixed effects model was used to assess the association between IOP-lowering therapy and tear film (TF) metrics, controlling for age and intra-individual correlation. The following parameters were measured: muco-aqueous layer thickness (MALT), muco-aqueous layer thinning rate (MALTR), lipid layer thickness (LLT), lipid map uniformity (LMU), inter-blink intervals (IBI), and lipid break-up time (LBUT). Results: Average ages significantly differed (p = 0.013) between the treatment group (66.5 years) and control group (average age 51.5 years), and thus results were adjusted for age accordingly. IOP was 17.1 mmHg in the treatment group and 16.1 mmHg in the control group. When analyzing the sublayers of the TF, MALTR had a significant association with IOP-lowering therapy after adjusting for age, with a difference of −52.68 nm/s; 95% confidence interval [−96.87, −8.48]; p-value = 0.020. Additionally, IBI was significantly associated with IOP-lowering therapy after log transformation (p = 0.049), with shorter IBI in the treatment group. All other metrics (MALT, LLT, LMU, and LBUT) were statistically insignificant (p > 0.05). Conclusions: These pilot results suggest that IOP-lowering drops may accelerate thinning of the TF, specifically the muco-aqueous layer. Longitudinal studies with significantly larger samples are needed to specify the differential impact of various ocular hypotensive therapies on the human TF and the clinical implications of these findings.

Sickle cell disease (SCD) is a single-gene illness which causes painful vaso-occlusion, debilitating organ damage, and early mortality. Its clinical course is variable, but current prognostic tools have large data requirements. Understanding prognosis is important for clinicians and patients in selecting therapies, particularly curative therapies. Retinal imaging may provide a non-invasive indicator of vaso-occlusion and risk of organ damage in SCD. This prospective observational cohort study included 150 individuals (>15 years) living with SCD. Retinal optical coherence tomography-angiography and basic laboratory tests were performed at six-month intervals over median 450 days (IQR 295-744) of follow-up. Primary outcomes were hospitalization within 12 months, kidney damage (urine albumin-to-creatinine ratio, uACR >100 and increased by ≥15 at next study visit), and heart-lung damage (N-terminal pro-Brain Natriuretic Peptide (NTproBNP) >160 pg/mL and increased by ≥15 pg/mL at next study visit). We processed retinal image metrics for perfusion and vascularity using established methods. We extracted retinal image features using a Swin transformer, and created ensemble models to predict outcomes. Random forest models predicted hospitalization within 12 months with area under the receiver operating curve (AUROC) 0.717 (standard deviation, SD = 0.053, 5-fold cross-validation). We predicted future kidney damage, AUROC 0.881 (SD 0.083), and heart-lung damage, AUROC 0.866 (SD 0.106). Ensemble models including transformer-derived image features, processed retinal image metrics, and laboratory tests performed best. Using supervised machine learning on noninvasive retinal imaging and basic laboratory tests we predicted future SCD hospitalizations and organ damage. This method shows promise for prognostication in SCD.

Purpose: To describe a case of minimally symptomatic, insidious systemic-lupus-erythematous (SLE) retinal vasculitis revealed with widefield multimodal imaging. Methods: Report of a case. Results: A 56-year-old woman presented with a vague history of visual blur in her left eye for 3 weeks. While her central acuity was relatively preserved, fundus examination was notable for scattered cotton-wool spots in a Purtscher-like retinopathy presentation. Widefield angiographic imaging revealed profound peripheral non-perfusion with some focal arteriolar vasculitis. Systemic work-up confirmed the diagnosis of SLE and high-dose intravenous steroid therapy was initiated. Immune modulatory therapy with mycophenolate mofetil and rituximab infusions were subsequently added to achieve control. Conclusions: Retinal vasculitis is a rare manifestation of SLE which may present in the absence of, or prior to central nervous system involvement. Visual prognosis is extremely guarded, necessitating aggressive immunosuppression and often pan-retinal photocoagulation. Widefield optical coherence tomography angiography may have a role in screening otherwise asymptomatic SLE patients for retinopathy as well as utility in monitoring treatment response.

Vision-Language Models (VLMs) offer a promising path toward interpretable medical diagnosis by allowing users to ask about clinical explanations alongside predictions and across different modalities. However, training VLMs for detailed reasoning requires large-scale image-text datasets. In many specialized domains, for example in reading Optical Coherence Tomography Angiography (OCTA) images, such precise text with grounded description of pathologies is scarce or even non-existent. To overcome this bottleneck, we introduce Synthetic Vasculature Reasoning (SVR), a framework that controllably synthesizes images and corresponding text, specifically: realistic retinal vasculature with Diabetic Retinopathy (DR) features: capillary dropout, microaneurysms, neovascularization, and tortuosity, while automatically generating granular reasoning texts. Based on this we curate OCTA-100K-SVR, an OCTA image-reasoning dataset with 100,000 pairs. Our experiments show that a general-purpose VLM (Qwen3-VL-8b) trained on the dataset achieves a zero-shot balanced classification accuracy of 89.67% on real OCTA images, outperforming supervised baselines. Through human expert evaluation we also demonstrate that it significantly enhances explanation quality and pathology localization on clinical data.

Purpose: We aimed to analyze the effects of age on human tear film (TF) using a novel nanometer resolution TF imaging device (Tear Film Imager, TFI, AdOM, Israel). Methods: 44 healthy adult subjects (≥18 years of age) without ocular or systemic diseases or prior eye treatments with ages spanning seven decades were enrolled in this prospective cross-sectional study. Subjects underwent a comprehensive ophthalmic examination and completed the Ocular Surface Disease Index questionnaire (OSDI). All study participants underwent TF imaging using the TFI, including assessment of muco-aqueous layer thickness (MALT), lipid-layer thickness (LLT), inter-blink interval, and lipid map uniformity. Associations between TFI parameters and age were tested using linear regression (accounting for multiple eyes). Results: A total of 80 eyes (44 subjects) were imaged: 19 eyes from 10 subjects in the 3rd decade of life (aged 20–29); 10 eyes from 5 subjects in the 4th decade of life (aged 30–39); 5 eyes from 3 subjects in the 5th decade of life (40–49); 12 eyes from 7 subjects in the 6th decade of life (50–59), 19 eyes from 11 subjects in the 7th decade of life (60–69); 11 eyes from 6 subjects in the 8th decade of life (70–79); and 4 eyes from 2 subjects in the 9th decade of life (80–89). With increasing age, MALT significantly decreased (p = 0.024), and LLT significantly increased (p = 0.001). No statistically significant linear age effects were found for the other TFI parameters (p > 0.05) or the OSDI scores of study participants of different ages (p = 0.786). Conclusions: Quantitative TF biomarkers varied significantly with advancing age in healthy individuals, highlighting the importance of accounting for age in TF assessments.