(The limits of) eye-tracking with iPads

Area of Science:

• Ophthalmology
• Human-Computer Interaction
• Biomedical Engineering

Background:

• Clinical applications of eye-tracking are often hindered by the need for specialized, costly hardware.
• Mobile devices like iPads offer a potential low-cost alternative for eye-tracking research.

Purpose of the Study:

• To compare the accuracy and reliability of eye-tracking using an Apple iPad Pro (3rd generation) against a dedicated Tobii 4c infrared eye-tracker.
• To evaluate the performance of iPad-based eye-tracking for various oculomotor tasks.

Main Methods:

• 28 participants performed tasks while their gaze was tracked using both an iPad Pro's built-in sensors and a Tobii 4c eye-tracker.
• Gaze location, fixation accuracy, saccade characteristics (count, speed, amplitude), and smooth pursuit were analyzed.

Main Results:

• iPad eye-tracking showed lower accuracy and precision for fixation estimation compared to the Tobii 4c (3.2° ± 2.0° vs. 0.75° ± 0.43°).
• Correlations were found for fixation stability, saccade counts, saccade speed, and amplitude, particularly for larger saccades (>8°).
• Significant variations in smooth pursuit estimation and tracking failures (5-20%) were observed with the iPad.

Conclusions:

• While iPad eye-tracking shows moderate correlation for certain saccadic measures, its accuracy limitations necessitate careful data validation.
• Researchers should exercise caution when using iPads for eye-tracking, implementing robust artifact detection and outlier removal strategies.