Comparison of automated ultrasound pupillometry assessment vs. infrared pupillary in critically ill patients

Background The pupillary light reflex (PLR) is a critical indicator of brainstem function in neurocritical care, but traditional assessments are often subjective, inconsistent, and prone to inter-examiner variability. Automated infrared pupillometry (IPA) provides objective metrics and is increasingly used in clinical practice, yet its accuracy may be compromised by eyelid swelling or ocular trauma. Automated ultrasound pupillometry (Auto-UPA) offers a non-invasive, transpalpebral alternative that can overcome these limitations. This study aimed to evaluate the agreement and reliability of Auto-UPA compared with IPA in ICU patients. Methods In this prospective observational study, consecutive adult ICU patients (February 1 to September 25, 2025) underwent paired Auto-UPA and IPA assessments under standardized conditions. PLR metrics included initial (INIT) and end (END) pupil diameters, latency (LAT), constriction ratio (DELTA), average constriction velocity (ACV), and average dilation velocity (ADV). Agreement was assessed using linear regression, Bland-Altman plots, and limits of agreement (LoA). Reliability was evaluated using intraclass correlation coefficients (ICC) for intra- and inter-observer variability. Results Twenty patients (40 eyes) were enrolled. High agreement was observed for pupil diameters and most dynamic indices (INIT: R 2 = 0.989, bias = 0.01 mm, LoA = −0.11 to 0.12 mm; END: R 2 = 0.992, bias = 0.02 mm, LoA = −0.07 to 0.11 mm), DELTA ( R 2 = 0.980, bias = −0.5%, LoA = −2.1 to 1.1%), ACV ( R 2 = 0.960, bias = −0.19 mm/s, LoA = −0.31 to −0.06 mm/s), ADV ( R 2 = 0.899, bias = −0.04 mm/s, LoA = −0.15 to 0.06 mm/s). LAT demonstrated only moderate concordance ( R 2 = 0.550, bias = 0.12 s, LoA = 0.08–0.16 s). Inter-observer ICCs ranged from 0.940 (LAT) to 0.999 (END), while intra-observer ICCs ranged from 0.890 (LAT) to 0.985 (END), indicating good-to-excellent reliability across all parameters. Conclusions Auto-UPA shows strong agreement with IPA in most PLR metrics and provides robust reliability, supporting its role as a feasible and practical alternative in settings where IPA is unavailable or impractical.