Abstract
Contrast sensitivity (CS) is important when assessing functional vision. However, current techniques for assessing CS are not suitable for young children or non-verbal individuals because they require reliable, subjective perceptual reports. This study explored the feasibility of applying eye tracking technology to quantify CS as a first step toward developing a testing paradigm that will not rely on observers' behavioral or language abilities. Using a within-subject design, 27 healthy young adults completed CS measures for three spatial frequencies with best-corrected vision and lens-induced optical blur. Monocular CS was estimated using a five-alternative, forced-choice grating detection task. Thresholds were measured using eye movement responses and conventional key-press responses. CS measured using eye movements compared well with results obtained using key-press responses [Pearson's r<sub>best-corrected</sub> = 0.966, P < 0.001]. Good test-retest variability was evident for the eye-movement-based measures (Pearson's r = 0.916, P < 0.001) with a coefficient of repeatability of 0.377 log CS across different days. This study provides a proof of concept that eye tracking can be used to automatically record eye gaze positions and accurately quantify human spatial vision.
Introduction
While standard visual acuity tests (like the Snellen "Big E" chart) measure the smallest detail a person can resolve, contrast sensitivity (CS) is often a superior predictor of real-world functional vision. CS measures the lowest level of contrast detectable across various spatial frequencies (sizes).
Traditionally, measuring CS requires a subjective report, such as a button press or verbal identification. This requirement creates a significant barrier for "pre-verbal" populations, including infants and individuals with cognitive or speech impairments. Current clinical workarounds, such as "Preferential Looking" (PL) tests, rely on a clinician's subjective judgment of where a patient is looking, which is prone to bias and requires specialized training. This study investigates an automated, objective alternative: using infrared eye tracking to quantify CS without requiring manual responses.
Methodology
The study utilized a within-subject design involving 27 healthy young adults (ages 19–35). Researchers compared two primary measurement methods:
- Behavioral (Beh): The traditional gold standard where participants used arrow keys to indicate the location of a stimulus.
- Eye Tracking (Elk): A novel method using an Eye-link 1000 infrared tracker. A "positive" response was recorded if the subject fixated on a target for more than one second.
Participants performed a five-alternative, forced-choice (5-AFC) task. A Gabor patch (the target) appeared randomly in one of four peripheral circles, or the subject could look at a central blank area if no target was detected. A "three-down-one-up" staircase procedure was used to determine the threshold across three spatial frequencies: 1, 4, and 16 cycles per degree (cpd). To test the system's sensitivity to vision deficits, a sub-group of 10 participants was also tested with lens-induced optical blur.
Key Results
The eye-tracking method (Elk) demonstrated exceptional alignment with traditional behavioral measures (Beh). The correlation between the two methods was remarkably high (Pearson’s $r = 0.966$), indicating that gaze fixation is a robust proxy for manual reporting.
Key findings include:
- Consistency: Individual subject data (N1-N27) showed that the eye-tracking results mirrored behavioral results across all tested spatial frequencies.
- Repeatability: In a sub-group of 9 participants tested on different days, the eye-tracking method remained highly consistent ($r = 0.916$).
- Clinical Sensitivity: The system successfully quantified the expected drop in CS when optical blur was introduced, particularly at medium and high spatial frequencies (4 and 16 cpd).
- Fatigue Effects: CS scores were slightly more conservative (lower) in the eye-tracking condition. The authors suggest this may be due to the eye-tracking test taking approximately twice as long as the behavioral test, leading to increased visual fatigue.
Discussion and Future Directions
This study provides a successful proof of concept for removing the "human-in-the-loop" subjectivity from pediatric and non-verbal vision testing. By automating gaze detection, the system bridges the gap between high-tech eye tracking and the biological reality of how humans naturally look at salient objects.
However, the current 10-minute test duration per spatial frequency is a limitation for clinical use with children. Future work will focus on:
- Algorithm Optimization: Implementing Bayesian adaptive frameworks (like "quick CSF") to reduce the number of trials needed.
- Engaging Stimuli: Using filtered cartoon patterns instead of Gabor patches to maintain the attention of younger patients.
- Hardware Flexibility: Transitioning to "head-free" remote eye tracking to accommodate the natural movements of infants and toddlers.
Summary updated