DECODING RADIOLOGISTS' INTENTIONS: A NOVEL SYSTEM FOR ACCURATE REGION IDENTIFICATION IN CHEST X-RAY IMAGE ANALYSIS

Area of Science:

• Medical Imaging
• Radiology Informatics

Background:

• Chest X-ray (CXR) analysis is critical for diagnosing thoracic abnormalities.
• Diagnostic errors occur, especially with less experienced radiologists, highlighting the need for improved accuracy.
• Understanding radiologist intent and focus areas in CXR reports is key to error correction.

Purpose of the Study:

• To develop a novel system for identifying radiologist intentions and corresponding regions of interest in CXR reports.
• To link textual findings in radiology reports to specific image locations.
• To enhance diagnostic accuracy and provide targeted feedback for medical trainees.

Main Methods:

• The study proposes a system to analyze radiologist reports and CXR images.
• The system aims to correlate textual descriptions with visual elements in the images.
• This facilitates the identification of regions of interest (ROIs) and radiologist intent.

Main Results:

• The developed system can identify primary radiologist intentions from reports.
• It successfully links these intentions to specific regions of interest within CXR images.
• This provides a mechanism for visual context elucidation.

Conclusions:

• The proposed system aids in rectifying diagnostic errors by guiding attention to accurate ROIs.
• It serves as a valuable tool for constructive feedback to inexperienced radiologists and residents.
• The system supports continuous learning and enhances diagnostic accuracy in medical imaging.