Abstract
PURPOSE: To assess whether automated analysis of retinal arterioles and venules can identify treatment response in papilledema secondary to idiopathic intracranial hypertension (IIH). METHODS: This retrospective analysis used data from a multicenter, randomized, double-blind, placebo-controlled IIH treatment trial. Participants (n = 165) with mild visual loss were assigned to a dietary/lifestyle modification plus acetazolamide (ACZ) or placebo for 6 months. Color fundus photographs, optical coherence tomography (OCT), and clinical metrics were collected at baseline and at multiple follow-up visits. AutoMorph, a deep learning-based pipeline, quantified venule and arteriole diameters, fractal dimensionality, tortuosity, and vessel density. Venular widths were standardized to arteriolar widths to form a venule-to-arteriole (V:A) ratio, which was correlated with Frisén grade, OCT optic nerve head (ONH) parameters, and cerebrospinal fluid (CSF) opening pressure. RESULTS: Baseline vascular OCT metrics and Frisén grades were similar between groups. At month 1, ACZ significantly reduced venule diameters (-4.59 µm; P = 0.02), and placebo showed no change (+1.21 µm; P = 0.54). The V:A ratio was consistently lower in the ACZ group than placebo from month 1 (1.20 vs. 1.24, respectively; P = 0.03) to month 6 (1.16 vs. 1.23, P = 0.02). Higher Frisén grades correlated strongly with increased mean V:A values (R2 = 0.91, P = 0.011). The V:A ratio was significantly associated with CSF opening pressure at month 6 (R2 = 0.47, P < 0.001). CONCLUSIONS: Automated retinal vessel analysis provides a promising, non-invasive method for monitoring treatment response in IIH and may complement traditional imaging and clinical assessments. TRANSLATIONAL RELEVANCE: Deep learning-based retinal vessel metrics may provide an accessible biomarker for monitoring treatment response in papilledema.