Skeletal Fracture Detection with Deep Learning: A Comprehensive Review.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Radiology

Background:

• Deep learning shows potential for diagnosing bone fractures from X-rays.
• Current research faces challenges due to unclear task definitions and lack of explainability.
• Existing reviews often lack technical depth or scope.

Purpose of the Study:

• To establish precise definitions for deep learning tasks in bone fracture diagnosis (recognition, classification, detection, localization).
• To analyze and summarize recent research methodologies, datasets, and outcomes.
• To identify key areas for future research in AI-driven fracture diagnosis.

Main Methods:

• Comprehensive literature review of 337 papers from WOS, Scopus, and EI.
• In-depth analysis and evaluation of 40 selected recent studies.
• Development of a generalized processing framework for deep learning in fracture diagnosis.

Main Results:

• Clear definitions provided for bone fracture recognition, classification, detection, and localization tasks.
• Summaries of 40 studies detailing bones, objectives, datasets, methods, and results.
• Identification of critical future research directions, including interpretability and multimodal data integration.

Conclusions:

• This review addresses the need for standardized task definitions in deep learning for fracture diagnosis.
• Findings provide a foundation for advancing AI in radiology with improved interpretability and clinical decision support.
• Future work should focus on explainable AI, multimodal data, and therapeutic recommendations.