Role of an Automated Deep Learning Algorithm for Reliable Screening of Abnormality in Chest Radiographs: A...

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence in Healthcare

Background:

• Chest X-rays are essential diagnostic tools, but high workloads and radiologist shortages challenge patient care.
• A reliable computer-aided diagnosis (CAD) system is needed to enhance radiological workflow efficiency.

Purpose of the Study:

• To evaluate the clinical utility of an AI-based chest X-ray screening tool (qXR) in real-world settings.
• To assess AI's accuracy in detecting abnormalities and its impact on reporting turnaround time.

Main Methods:

• A prospective multicenter quality-improvement study involving radiologists using the qXR tool.
• Consecutive chest X-rays were processed using the AI tool in a large Indian radiology network from June 2021 to March 2022.

Main Results:

• The AI system processed 65,604 chest X-rays, showing good overall performance in detecting normal and abnormal findings.
• High negative predicted value (NPV) of 98.9% was achieved, with excellent AUC and NPV for various abnormalities.
• Turnaround time (TAT) for reporting decreased by approximately 40.63% post-implementation of the AI tool.

Conclusions:

• The AI-based qXR solution effectively screens chest X-rays, enabling confident exclusion of normal cases.
• This allows radiologists to dedicate more time to analyzing complex pathologies in abnormal X-rays, optimizing patient management.