Smart chest X-ray worklist prioritization using artificial intelligence: a clinical workflow simulation.

Area of Science:

• Radiology Workflow Optimization
• Artificial Intelligence in Medical Imaging
• Clinical Decision Support Systems

Background:

• Current radiology workflows often use a first-in, first-out (FIFO) system, which may delay critical findings.
• Artificial intelligence (AI) offers potential for intelligent worklist prioritization to improve efficiency.
• False negative predictions by AI could lead to dangerously long report turnaround times (RTATs).

Purpose of the Study:

• To evaluate AI-driven smart worklist prioritization for optimizing radiology workflows.
• To assess the impact of AI on reducing RTATs for critical findings in chest radiographs (CXRs).
• To investigate methods for mitigating risks of prolonged RTATs due to AI false negatives.

Main Methods:

• Development of a realistic clinical workflow simulation framework using hospital-specific data.
• Comparison of standard FIFO worklist processing with AI-based urgency prioritization.
• Implementation of an 'upper limit' for maximum waiting time to assign highest urgency.

Main Results:

• AI prioritization significantly reduced average RTAT for critical findings compared to FIFO (e.g., pneumothorax: 35.6 min vs. 80.1 min).
• Maximum RTAT increased in some cases without an upper limit, highlighting the risk of false negatives.
• The 'upper limit' strategy substantially reduced maximum RTATs across all critical finding classes.

Conclusions:

• AI-powered smart worklist prioritization effectively reduces average RTAT for critical CXR findings.
• An upper limit on waiting time is crucial to prevent excessively long RTATs caused by AI false negatives.
• AI can achieve near-optimal performance, comparable to perfect classification, in reducing critical finding RTATs.