Aims/Purpose: Dynamic Vessel Analysis (DVA) is a non‐invasive imaging modality that evaluates the reactivity of retinal arteriolar and venular vessels in response to flicker light stimulation. This study aims to introduce a novel data‐processing pipeline that involves noise reduction techniques localized to different locations along the vessel wall. Additionally, it employs standardized parameter extraction, allowing for a more precise assessment of retinal vessel responses across these locations.Methods: Previous studies used robust principal component analysis (RPCA) on averaged time series data for arteries and veins, combining one averaged time series per patient into arterial and venous matrices. Our method retains individual location time series for each arteriole and venule, preserving local responses. We combine these individual time series into arteriolar and venular matrices and apply RPCA, resulting in sparse matrices containing noise and low‐rank matrices capturing common features. The low‐rank matrices, containing the noise‐reduced information, are then used for further analysis. This approach maintains the granularity of local vascular responses.Results: We identified and integrated key parameters from existing literature to delineate vasodilation and vasoconstriction profiles. By applying analysis of covariance (ANCOVA) for age, sex, baseline vessel dilation, and mean arterial pressure, we identified significantly lower Baseline‐Corrected Flicker Response of the proximal retinal venules in normal tension glaucoma patients (2.46%, n = 23) compared to healthy controls (4.23%, n = 21, p = 0.049).Conclusions: Our novel pipeline adds to all previously published methods and effectively identified altered neurovascular responses in glaucoma patients after the necessary adjustments. This standardized and automated approach is a promising catalysator for DVA research, offering a more localized and precise understanding of retinal vascular dynamics.
Read full abstract