An interactive eye-tracking system for measuring radiologists' visual fixations in volumetric CT images:...

Area of Science:

• Radiology
• Human-Computer Interaction
• Medical Imaging

Background:

• Eye-tracking in radiology is crucial for understanding visual search.
• Previous CT studies were limited to static images or video playback, not interactive 3D viewing.
• A novel platform integrates eye-tracking with workstation software for realistic reader task simulation.

Purpose of the Study:

• To validate the spatial accuracy of an integrated eye-tracking system.
• To evaluate different eye-tracking data acquisition modes in a clinical context.
• To assess the system's performance in simulating radiologists' interactive 3D image viewing.

Main Methods:

• Integrated an eye-tracker with reader workstation software, capturing eye movements and workstation events at 1000 Hz.
• Compared head-stabilized mode (chinrest) with free-movement modes (general and strict audio biofeedback).
• Evaluated spatial accuracy using digital phantoms and interactive CT angiographic image viewing tasks.

Main Results:

• All three head position constraint conditions showed comparable accuracy with digital phantoms (median offset ~14-19 pixels).
• Interactive viewing of patient CT angiographic images revealed comparable accuracy between chinrest and strict biofeedback modes.
• General biofeedback showed slightly worse accuracy in interactive tasks (median offset ~30 pixels), translating to visual angles of 0.7°-1.3°.

Conclusions:

• The integrated eye-tracker system demonstrates reasonable spatial accuracy for assessing visual fixation during interactive 3D image viewing.
• Head-free movement with audio biofeedback offers a viable alternative to head-stabilized mode.
• This platform enables more realistic studies of radiologists' visual search behavior in diagnostic imaging.