Analyzing Overlaid Foreign Objects in Chest X-rays-Clinical Significance and Artificial Intelligence Tools

Area of Science:

• Radiology and Medical Imaging
• Artificial Intelligence in Medicine
• Computer-Aided Diagnosis

Background:

• Chest X-rays (CXRs) can contain non-biomedical foreign objects (NBFO) like coins and buttons, and biomedical foreign objects (BFO) such as medical devices.
• These foreign objects can impede accurate interpretation of CXRs, leading to misdiagnosis or overlooking genuine abnormalities.
• NBFOs and BFOs can be misclassified as biological abnormalities by machine learning algorithms, resulting in false positives.

Purpose of the Study:

• To review and discuss the clinical significance of detecting and classifying NBFOs and BFOs in CXRs.
• To examine the application of computer-aided detection (CADe) and diagnosis (CADx) tools for managing foreign objects in CXRs.
• To highlight the importance of accurately isolating and either removing or highlighting foreign objects for effective abnormality analysis.

Main Methods:

• Comprehensive review of 29 peer-reviewed research reports and articles.
• Analysis of clinical reports detailing the impact of foreign objects on CXR interpretation.
• Examination of shallow learning and deep learning algorithms applied in CADe and CADx tools for foreign object detection.

Main Results:

• Foreign objects, both NBFOs and BFOs, significantly complicate CXR interpretation.
• Machine learning algorithms, if not specifically trained, can erroneously identify foreign objects as pathological findings.
• Effective detection, localization, and classification of foreign objects are essential for reliable diagnostic outcomes.

Conclusions:

• Accurate detection, classification, and handling of NBFOs and BFOs are vital for improving diagnostic accuracy in chest radiography.
• Advanced computational tools, including deep learning, show promise in addressing the challenge of foreign objects in CXRs.
• Further research and development in CADe/CADx systems are needed to robustly manage foreign objects and enhance clinical decision-making.