Experimental system for measurement of radiologists' performance by visual search task

Area of Science:

• Radiology
• Medical Imaging
• Visual Perception

Background:

• Current evaluation of radiologist performance often relies on Receiver Operating Characteristic (ROC) analysis.
• Visual search tasks are considered more appropriate for evaluating performance with obvious targets but are underutilized in radiology research.
• Establishing a feasible visual search task environment is crucial for advancing radiologist performance assessment.

Purpose of the Study:

• To develop and evaluate a feasible visual search task environment for radiology.
• To assess the impact of radiologist experience (board certification) on visual search performance.
• To compare visual search task performance with traditional metrics.

Main Methods:

• Ten radiologists participated in a visual search task in a controlled darkroom setting.
• Images with and without various lesions (bulla, ground-glass nodule, solid nodule) were presented randomly.
• Accuracy and reaction times were recorded, and differences based on board certification were analyzed.

Main Results:

• The visual search task was successfully implemented and proven feasible in a radiology context.
• Radiologists demonstrated high diagnostic accuracy across all evaluated lesion types.
• Board-certified radiologists showed significantly faster reaction times for detecting bullae compared to non-boarded radiologists.

Conclusions:

• An experimental system for conducting visual search experiments in radiology has been successfully developed.
• Radiologist experience, specifically board certification, correlates with improved reaction times for specific lesion detection (bulla).
• This study validates the use of visual search tasks for evaluating radiologist performance.