The pupillary light reflex (PLR) is an important indicator of cognitive capability that is widely utilized in medicine today. Reflex is a mobile solution to assess PLR using only the camera and flash present in an iOS device. This paper presents a full analytical analysis of the operability of Reflex and how it can be utilized to assist in the management of neurological disturbances, injury, and indicators.
The Challenge of Visible-Light Pupillometry
Historically, quantitative PLR (qPLR) systems have relied on infrared cameras to distinguish the pupil from the iris — particularly important for individuals with dark irises, which make up the majority of the global population. Existing hardware systems cost between $4,500 and $9,000 and require dedicated equipment. Reflex is the world's first fully functional mobile pupillometer that requires no additional hardware, operating in the visible light spectrum with accuracy competitive with infrared systems.
The key distinguishing factor is Reflex's ability to operate with a low error rate in ambient, visible-spectrum conditions — overcoming the core challenge that has prevented prior smartphone-based systems from achieving clinical reliability.
The Reflex Pipeline
Reflex records video of a subject's eye at 30 Hz, which is parsed into images and passed into a trained object detector. Reflex uses a custom detection architecture for rapid identification of an iris, including partial irises caused by eyelid occlusion. The detector classifies each frame as either containing an iris or not, then supplies rectangular coordinates for the iris when detected.
After a quality threshold check, outlier detection using first and second order statistics cleans the iris location predictions. A registration phase stabilizes frames to a single reference axis. Then a proprietary algorithm determines pupil diameter for each frame.
Accuracy Results
Reflex's custom iris detector achieved 96.9% accuracy and a mean intersection over union (IoU) of 0.814, outperforming the standard Haar detector (89.2% accuracy, IoU 0.790). Manual annotations were conducted for 600 images by a computer vision professional to provide ground-truth pupil diameter values. Reflex's predicted values trended closely with ground-truth values, with a mean average error of 2.9% on scaled values and 0.75mm in millimeter measurements.
qPLR Clinical Applications
TBI and Concussion
Nearly 4 million sports-related traumatic brain injuries occur in the United States each year, most being mild TBIs (concussions). Changes in pupil response have been correlated with brain stem and hypothalamus injuries, blast-induced mTBIs, and non-blast induced mTBIs. qPLR provides an objective, repeatable method for evaluating concussion and monitoring recovery.
Fatigue
Excessive daytime sleepiness in adults can be characterized by PLR latency and constriction velocity, which closely compare to self-reported fatigue levels. Sleep deprivation and its impact on pupillary activity has been correlated with increased reactions to negative imagery.
Neurodegeneration
qPLR has been statistically validated as a trending biomarker relating to the progression of neurodegeneration, specifically for Alzheimer's disease. Significant predictors of Parkinson's disease include latency and maximum constriction velocity.
Anesthesiology
Oxyhemoglobin levels have been distinctly quantified to trend with the pupillary light reflex, demonstrating the recovery monitoring capability qPLR offers in anesthetic contexts.
Cognition
Cognitive load can influence the autonomic nervous system through perception mechanisms including sound, sight, and touch. PLR correlations to memory load diminish with age, and qPLR has been shown to suffer as a result of backwards masking tasks that pull cognitive processing power away from reactionary allocations.
Conclusion
qPLR has been established as a breakthrough biomarker that can reliably indicate cognitive ability, injury, or neurodegeneration. Reflex, as a qPLR tool, can greatly influence current trends in clinical acceptance of qPLR due to its convenience and high accuracy. Utilization of these tools will add a significant amount of autonomic neurological data that otherwise would go unnoticed, while taking the subjectivity out of PLR analysis by providing data representative of the ground-truth pupillary response.