
Purpose: Retinopathy of prematurity (ROP) is a leading cause of preventable childhood blindness. Although the North-American (NA)/American Academy of Ophthalmology (AAO) and Growth and ROP (G-ROP) criteria are highly sensitive, they may trigger unnecessary examinations. This study evaluated whether an interpretable machine-learning (ML) model could reduce screen-positive classifications while identifying all documented Early Treatment for ROP (ETROP) type 1 ROP cases. Methods: A retrospective study included infants admitted to Hadassah Medical Center neonatal intensive care units (January 2023-February 2025). Of 2159 infants, 387 met local screening criteria (gestational age ≤32 weeks or birth weight ≤1500 g); 189 had complete G-ROP predictor data, and 5 were excluded because ROP status remained unavailable. ML models used gestational age, birth weight, and postnatal weight gain. L1-regularized logistic regression models used either gestational age and birth weight alone or these variables plus G-ROP postnatal weight-gain measures. Leave-one-out cross-validation assessed performance. Results: Among 184 infants, 21 had type 1 ROP; all approaches identified all 21 cases. Local, NA/AAO, modified G-ROP, and birth-based criteria classified 184 (100.0%), 174 (94.6%), 137 (74.5%), and 94 (51.1%) infants as screen-positive, respectively. The 5-variable ML model classified 58 (31.5%; specificity = 77.3%) as screen-positive, a 68.5% reduction versus local screening and 57.7% versus G-ROP (area under the receiver operating characteristic curve [AUROC] = 0.914). The birth-only ML model performed similarly (55/184 [29.9%]; specificity = 79.1%; AUROC = 0.909), indicating no incremental screening benefit from postnatal weight gain. Conclusions: Interpretable ML substantially reduced screen-positive classifications while identifying all documented type 1 ROP cases. Similar birth-only performance supports prospective validation of locally adapted screening. Translational Relevance: Routinely collected data and ML framework may enable more efficient type 1 ROP screening and neonatal ophthalmic care.
Purpose: To characterize the dynamic evolution of aqueous humor cytokine profiles after glaucoma filtration surgery (GFS) and to identify molecular signatures associated with bleb survival outcomes. Methods: GFS was performed on 26 rabbits' right eyes. Intraocular pressure (IOP), bleb morphology, and aqueous humor samples were collected before and after surgery. Thirty cytokines were quantified using RayBio Rabbit Cytokine Array Q30. Statistical analyses included Kaplan-Meier survival analysis, repeated-measures analysis of variance, independent samples t-tests or nonparametric tests, and Spearman correlation. Results: IOP decreased significantly in the early postoperative period and then gradually increased and stabilized. The mean bleb survival time was 15.19 ± 2.47 days. Among the 30 cytokines, 25 showed significant time-dependent changes, with peak times ranging from postoperative day 1 to day 21. After surgery, 12 cytokines were significantly higher in the early bleb failure group compared to the late failure group, with the greatest number and strongest negative correlations with bleb survival time observed on postoperative day 7, among which Contactin-1 (r = -0.6018), follistatin-related gene (FLRG; r = -0.5906), and thrombospondin-2 (TSP-2; r = -0.5847) exhibited the strongest correlations. Conclusions: GFS can induce time-dependent elevations in multiple cytokines in the aqueous humor, with high expression of factors (such as Contactin-1, FLRG, and TSP-2) on postoperative day 7 closely associated with earlier bleb failure. Translational Relevance: This study profiles post-GFS cytokine dynamics in rabbit aqueous humor, suggesting day 7 as a potential key window where Contactin-1, FLRG, and TSP-2 may act as scarring contributors, providing potential molecular targets and timing for future targeted interventions to improve the long-term success of filtration surgery.
Purpose: The silent substitution principle enables a selective modulation of individual photoreceptor classes while maintaining constant excitation of all others. We describe a five-primary video projector and present pilot photoreceptor-directed contrast modulation flicker perimetry data from four normal observers. Methods: The Multi-Channel Visual Stimulator (MCVS; Cambridge Research Systems Limited, Rochester, UK) is a projector that uses a digital micromirror device to create images based on the emission from five primaries (red, amber, green, blue, and deep blue) that are driven by pulse-width modulation. L-, M-, and S-cone-directed circular stimuli (0.86) were presented at the central locations of the 10-2 pattern (±8° horizontally, ±6° vertically). The photoreceptor excitation was modulated sinusoidally around the background at temporal frequencies of 2 Hz and 20 Hz. Four normal observers (two male, two female; aged 35, 37, 47, and 65 years) were tested. Results: Temporal contrast sensitivities showed expected photoreceptor pathway characteristics, comparable with earlier experiments using a larger, annular test field. Specifically, L- and M-cone-driven sensitivities were very similar at 2 Hz (red-green opponent pathway), but L-cone-driven sensitivities were higher than M-cone-driven sensitivities at 20 Hz (luminance pathway). S-cone-driven sensitivities were lower near fixation but decreased less toward the periphery. Conclusions: The MCVS enables spatial mapping of photoreceptor-directed temporal contrast sensitivities across the central retina. This approach facilitates pathway-specific investigation of retinal processing and holds promise as a functional outcome measure in clinical trials targeting neuroretinal disorders. Translational Relevance: Photoreceptor-directed perimetry may provide sensitive outcome measures for neuroretinal disease diagnosis and clinical trials.
Purpose: Variable expressivity is well recognized in autosomal-dominant inherited retinal diseases (IRDs) but is increasingly observed and less well defined in recessive IRDs. Despite the traditional view of recessive IRDs as monogenic disorders with high penetrance, phenotypic differences among individuals sharing identical biallelic variants suggest additional disease modifiers. We quantified genotype-phenotype concordance in EYS- and USH2A-associated retinitis pigmentosa (RP) and characterized phenotypic variability among individuals with identical primary genotypes. Methods: Participants underwent standardized ophthalmic assessment with multimodal imaging. Longitudinal ellipsoid zone (EZ) widths were measured from optical coherence tomography (OCT) scans obtained ≥12 months apart. Baseline EZ width and age at OCT assessment were compared among individuals sharing identical primary genotypes. Genotype-matched comparisons in which younger individuals demonstrated smaller EZ widths than older counterparts were classified as age-discordant structural comparisons. Results: Ninety patients (41 EYS, 49 USH2A) were included. Recurrent variants EYS c.6416G>A (p.Cys2139Tyr) and USH2A c.2802T>G (p.Cys934Trp) were identified in 46% and 35% of probands, respectively. Females showed greater baseline EZ preservation, although differences were not statistically significant. Marked variability was observed within matched genotypes. Several genotype-matched comparisons demonstrated age-discordant structural variability, including younger individuals with substantially smaller EZ widths than older genotype-matched counterparts. In EYS-associated disease, the largest discordant comparison demonstrated 85.4% less EZ preservation in the younger individuals despite younger age, and the corresponding value in USH2A-associated disease was 81.8%. Intrafamilial discordance further supported the possibility of modifying influences beyond the primary genotype. Conclusions: Disease severity varied substantially among individuals sharing identical primary genotypes, supporting contributions from additional genetic or environmental modifiers. Translational Relevance: Recognition of variability within recessive IRDs has implications for prognosis, trial stratification, and interpretation of treatment response beyond genotype alone.
Purpose: Automatic extraction of retinal vascular biomarkers from color fundus images (CFIs) is crucial for large-scale studies of the retinal vasculature. We present VascX, an open-source software toolbox that extracts biomarkers from CFI artery-vein segmentations. Methods: From a color fundus image, VascX computes artery and vein segmentation masks, extracts their centerlines, builds undirected and directed vessel graphs, and resolves vessel segments into longer vessels. A comprehensive set of biomarkers is derived, including vascular density, central retinal equivalents (CREs), and tortuosity. Spatially localized biomarkers may be calculated over grids placed relative to the fovea and optic disc. Results: Our test-retest reproducibility analysis on repeat imaging of the same eye by different devices shows that most VascX biomarkers have moderate to excellent agreement (intraclass correlation coefficient > 0.5), with important differences in the level of robustness of different biomarkers. Our analyses of biomarker sensitivity to image perturbations and heuristic parameter values support these differences and further characterize VascX biomarkers. Conclusions: VascX provides an explainable and flexible biomarker extraction toolbox that complements segmentation to produce reliable retinal vascular biomarkers. Our graph-based biomarker computation stages support reproducible, region-aware measurements suited for large-scale clinical and epidemiological research. Translational Relevance: VascX supports oculomics research by enabling easy extraction of existing biomarkers and rapid experimentation with new biomarkers. Its robustness and computational efficiency facilitate scalable deployment in large databases. VascX is released via GitHub and PyPI with comprehensive documentation and examples to facilitate adoption by ophthalmic researchers and clinicians.