Changes in Radiologists' Gaze Patterns Against Lung X-rays with Different Abnormalities: a Randomized Experiment

Area of Science:

• Radiology and Medical Imaging
• Human-Computer Interaction
• Cognitive Science

Background:

• Radiologist workload has increased, potentially impacting diagnostic accuracy.
• Diagnostic accuracy declines significantly at the end of shifts.
• Understanding gaze patterns under high workload is crucial for quality assurance.

Purpose of the Study:

• To investigate how radiologists' gaze patterns on chest X-rays change with increasing workload.
• To quantitatively assess the impact of workload on gaze coverage of lung fields.
• To analyze gaze behavior in the presence of various chest abnormalities.

Main Methods:

• A randomized experiment using an eye-tracker on a radiological workstation.
• Four radiologists read 400 chest X-rays with known diagnoses.
• U-Net neural network for automatic lung field segmentation to measure gaze coverage.

Main Results:

• Average lung coverage ranged from 55-65% per radiologist.
• Lung coverage decreased by 1.3-7.6% for every 100 X-rays read.
• Coverage reduction was consistent across all abnormality types, e.g., 3.4% for cardiomegaly, 4.1% for atelectasis.

Conclusions:

• Increased workload leads to reduced gaze coverage of lung fields in chest X-rays.
• This pattern is consistent across different abnormalities and viewing orders.
• The findings quantify changes in X-ray reading patterns due to high workload.