Abstract
PURPOSE: Intracranial pressure (ICP) control is important to avoid secondary brain injury in patients with intracranial pathologies. Current methods for measuring ICP are invasive and carry risks of infection and hemorrhage. Previously we found correlation between ICP and the arteriole-venous ratio (A/V ratio) of retinal vessels in an outpatient setting. This study investigated the usability of fundoscopy for non-invasive ICP estimation with the addition of intraocular pressure (IOP) in patients in a neuro-intensive care unit (NICU). METHODS: This single-center prospective cohort study was conducted at the NICU of Odense University Hospital from September 2020 to May 2021. Adult patients with a Glasgow Coma Score of 8 or less, who underwent invasive pressure neuromonitoring were included. Fundoscopy videos were captured daily and analyzed using deep learning algorithms. The A/V ratio was calculated and correlated with ICP. The data was analyzed using mixed-effect linear regression models. RESULTS: Forty patients were enrolled. Fifteen were included in the final analysis. ICP ranged from -1 to 31 mmHg (mean: 10.9, SD: 5.7), and IOP ranged from 4 to 13 mmHg (mean: 7.4, SD: 2.1). The A/V ratio showed a significant negative correlation with ICP > 15 mmHg (regression slope: -0.0659, 95%-CI: [-0.0665;-0.0653], p < 0.001). No significant change in A/V ratio was observed for ICP ≤ 15 mmHg. A similar significant correlation was found for ICP > IOP (regression slope: -0.0055, 95%-CI: [-0.0062;-0.0048], p < 0.001). Taking the IOP into account did not improve the model. The sensitivity analysis showed a sensitivity of 80.08% and a specificity of 22.51%, with an AUC of 0.6389. CONCLUSION: In line with our previous work, non-invasive fundoscopy is a potential tool for detecting elevated ICP. However, challenges such as image quality and diagnostic specificity remains. Further research with larger, multi-center studies are needed to validate the utility. Standardization may enhance the technique's clinical applicability.