Computational Pupillometry and the Pupil Reactivity (PuRe) Score in Neurocritical Care: Prospects for High-Accuracy, Lighting-Invariant Neuromonitoring.
Patryk Szczęśniewski, Małgorzata Dudzic, Karol Badowski, Michał Świątek, Martyna Płatnikow, Jakub Klawiter, Wiktor Olszewski, Marta Kowalow, Zuzanna Paryzek, Kacper Pawlak, Hugo Chrost, Michal Wlodarski, Marek Dziubiński, Radosław Chrapkiewicz, Artur Drużdż
Abstract
Open AccessIntroduction An evaluation of computational pupillometry, implemented through the pupil reactivity (PuRe) pupillometer smartphone-based device, was prospectively carried out in a neuro-ICU cohort to make a lighting-invariant assessment of pupil reactivity and its clinical utility in critical care. Methods This prospective single-center observational pilot study was conducted in the neuro-ICU of a municipal hospital in Poznań, Poland, in 2024. Adult patients admitted with acute neurological conditions such as hemorrhagic stroke, traumatic brain injury, brain tumor, hydrocephalus, or intracranial hypertension were monitored under varying ambient light conditions. Pupillometry measurements were collected using the PuRe pupillometer, a smartphone-based device that computes standard pupillary light reflex parameters and the lighting-invariant Pupil Reactivity (PuRe) Score. Neurological status was assessed via the Glasgow Coma Scale, and survival outcomes were recorded at discharge. Statistical analyses included comparisons across light conditions using non-parametric tests, correlations with clinical variables via Spearman's correlation, and receiver operating characteristic analysis for predictive thresholds. Results Twelve patients, with a median age of 59 years, contributed 1,331 measurements, at 44 measurements per patient, under ambient light ranging from 4 to 1,200 lux. Standard pupillary light reflex parameters showed high light dependence with variations up to 133% (p<0.001), whereas PuRe Scores remained stable with medians of 2.64 in dim light versus 2.42 in bright light (p=0.449; r=-0.026). PuRe Scores correlated strongly with neurological status (Spearman ρ=0.746, p<0.001). At a threshold below 3.0, the PuRe Score identified severe Glasgow Coma Scale scores (≤8) with 84.3% sensitivity, 90.2% specificity, 86% accuracy, and an area under the curve of 0.940. Non-survivors exhibited markedly lower median PuRe Scores of 0.00 compared to 2.82 in survivors (p<0.001), with 91.7% of non-survivor recordings below the 3.0 threshold. Frame-level analysis demonstrated pupil diameter accuracy of ±0.025 mm. Conclusion Lighting-invariant, artificial intelligence (AI)-driven pupillometry provides a precise, clinically meaningful assessment of pupil reactivity, enabling objective bedside neuromonitoring without strict light control and supporting early detection of severe injury and poor outcome.