In Cephalalgia : an international journal of headache
INTRODUCTION : Retinal vessel dynamics analysis has proven to be a viable, non-invasive surrogate marker for increased intracranial pressure. We aimed to test this method in patients with suspected idiopathic intracranial hypertension.
METHODS : Patients with suspected idiopathic intracranial hypertension were prospectively enrolled for hand-held fundus-videography during diagnostic lumbar puncture. After extracting optic disc images, peripapillary arteriole-to-venule-ratios were measured using machine-learning algorithms with manual identification control. A general linear model was applied to arteriole-to-venule-ratios and corresponding lumbar opening pressures to estimate cerebrospinal fluid pressure.
RESULTS : Twenty-five patients were included with a significant difference in arteriole-to-venule-ratio between patients with (n = 17) and without (n = 8) idiopathic intracranial hypertension (0.78 ± 0.10 vs 0.90 ± 0.08, p = 0.006). Arteriole-to-venule-ratio correlated inversely with lumbar opening pressure (slope regression estimate -0.0043 (95% CI -0.0073 to -0.0023), p = 0.002) and the association was stronger when lumbar opening pressure exceeded 15 mm Hg (20 cm H2O) (slope regression estimate -0.0080 (95% CI -0.0123 to -0.0039), p < 0.001). Estimated cerebrospinal fluid pressure predicted increased lumbar opening pressure >20 mm Hg (27 cm H2O) with 78% sensitivity and 92% specificity (AUC 0.81, p = 0.02). A stand-alone arteriole-to-venule-ratio measurement predicting lumbar opening pressure >20 mm Hg (27 cm H2O) was inferior with a 48% sensitivity and 92% specificity (AUC 0.73, p = 0.002).
CONCLUSION : Retinal vessel dynamics analysis with the described model for estimating cerebrospinal fluid pressure is a promising non-invasive method with a high sensitivity and specificity for detecting elevated intracranial pressure at follow-up assessments of patients with confirmed idiopathic intracranial hypertension if initial lumbar opening pressure and arteriole-to-venule-ratio data are available.
Hagen Snorre Malm, Wibroe Elisabeth Arnberg, Korsbæk Johanne Juhl, Andersen Mikkel Schou, Nielsen Asger Bjørnær, Nortvig Mathias Just, Beier Dagmar, Poulsen Frantz Rom, Jensen Rigmor Højland, Hamann Steffen
2023-Mar
Retinal arteriole-to-venule ratio, cerebrospinal fluid pressure, deep learning, non-invasive diagnostics, optic disc imaging, papilledema