Investigation of nonlinear pupil dynamics by recurrence quantification analysis

Area of Science:

• Neuroscience
• Biomedical Engineering
• Physiology

Background:

• The pupil's size and position are controlled by the autonomous nervous system (ANS).
• Pupil dynamics exhibit complex behavior even under constant stimulation.
• Investigating ANS function non-invasively is crucial for understanding physiological states and pathologies.

Purpose of the Study:

• To explore the potential of low-cost pupil investigation systems for extracting ANS information.
• To evaluate the significance of nonlinear information within pupillograms for ANS assessment.
• To differentiate between subtle ANS stimulations using pupil dynamics.

Main Methods:

• 13 healthy subjects were studied under different stationary conditions, including habitual dental occlusion (HDO) as a mild ANS stimulus.
• Infrared cameras captured pupil images, from which position and size time series were extracted.
• Linear and nonlinear (Recurrence Quantification Analysis - RQA) indexes were calculated from the time series.
• Machine learning algorithms (multilayer perceptrons, support vector machines) were used for data classification.

Main Results:

• Classification performance improved significantly when nonlinear RQA indexes were included as input features.
• Nonlinear indexes provided additional, valuable information on pupil dynamics compared to linear descriptors alone.
• The study successfully discriminated between subtle ANS stimulations, such as HDO versus rest position.

Conclusions:

• Nonlinear analysis of pupillograms, particularly using RQA, offers a powerful tool for assessing ANS function.
• Low-cost pupil investigation systems combined with nonlinear analysis can detect subtle physiological changes.
• This approach holds promise for non-invasive investigation of ANS-related pathologies.