Development of an Eye-Tracking Method for Diagnosing Delirium: Assessing Cognitive Function and Autonomic Nervous System Activity

Delirium is a severe neuro-psychiatric condition characterized by a global deterioration of cognitive functions. The Confusion Assessment Method for the Intensive Care Unit scale, a test that evaluates a patient's cognitive abilities, is currently the most recommended method for assessing suspected delirium in intensive care units. However, this method faces challenges such as inadequate staff training, patient cooperation difficulties, and a lack of measurable quantitative metrics. While there are existing tests in the literature that use eye tracking to evaluate cognitive functions and assess the autonomic nervous system, there is currently no specific test available based on wearable eye tracking for diagnosing delirium. In this study, we investigated the application of a test to evaluate cognitive function and autonomic nervous system activity using eye tracking. The proposed protocol consists of four tests: the blank test, pro- and anti-saccade tests, and pupillometry test. In the first three tests, visual stimuli are presented to the subjects, and an algorithm has been developed to calculate their reaction times, which is the time interval from the presentation of the stimulus to the completion of the task. The extended uncertainty with a 95% confidence level introduced by the algorithm has been estimated as 12 ms using the Monte Carlo method. In the pupillometry test, a temporary light stimulus is provided, and parameters related to the pupil light reflex are calculated. The tests were conducted on a sample of 17 healthy male subjects, aged between 21 and 28 years, without visual or cognitive impairments. The results provide an overview of the cognitive functions and autonomic nervous system activity in the examined subjects. Additionally, the data collected from healthy subjects can be compared with individuals affected by delirium in future studies. To validate the reaction time algorithm, the Monte Carlo method was used, and a statistical analysis was conducted on the output data. The evaluation of the algorithm for detecting reaction times shows a root mean square error of 6 ms. The results of the pro- and anti-saccade tests show a statistically significant difference in detected reaction times (p-value ! 0.05), with the latter being higher due to the increased cognitive load required. The pupillometry test measured pupil light reflex parameters and compared them with those reported in the literature.