Impact of Deep Learning-Based Computer-Aided Detection and Electronic Notification System for Pneumothorax on Time to...

Area of Science:

• Radiology
• Artificial Intelligence in Medicine
• Clinical Workflow Optimization

Background:

• Pneumothorax (PTX) diagnosis on chest radiography (CXR) requires timely treatment.
• Current clinical workflows may have delays in initiating PTX treatment.
• Deep learning (DL) computer-aided detection (CAD) and electronic notification systems (ENS) offer potential solutions.

Purpose of the Study:

• To evaluate the impact of DL-based CAD combined with ENS on time to treatment (TTT) for suspected PTX.
• To assess the effectiveness of this integrated system in a real-world clinical setting.

Main Methods:

• A commercial DL-based CAD and ENS system was implemented across all CXRs at a large general hospital.
• A difference-in-differences analysis compared TTT between a group using both CAD and ENS and a group using CAD only.
• Data from 603,028 CXRs (140,841 patients) were analyzed, covering periods before and after system implementation.

Main Results:

• The combined CAD and ENS group showed a significant reduction in TTT for supplemental oxygen therapy (-143.8 min; P = .035).
• No significant differences in TTT were observed for aspiration, tube thoracostomy, or surgical consultations.
• The study included a PTX prevalence of 2.0%.

Conclusions:

• Implementing DL-based CAD with ENS effectively reduces the time to initiate oxygen supplementation for PTX patients.
• The integrated system shows promise for improving early management of PTX.