A Benchmark X-ray Dataset for Pediatric Supracondylar Humerus Fractures with Improved YOLOv11-Based Detection

Area of Science:

• Orthopedic Surgery
• Artificial Intelligence in Medicine
• Medical Imaging Analysis

Background:

• Supracondylar humerus fractures (SHFs) are the most common pediatric elbow fractures, requiring precise diagnosis to prevent complications.
• Current diagnostic methods for pediatric SHFs can be challenging, necessitating advanced tools for accuracy and timeliness.

Purpose of the Study:

• To introduce PediaSHF-DX, a comprehensive dataset of pediatric elbow X-rays for AI development.
• To develop and evaluate an AI model for automated and accurate detection of pediatric SHFs.

Main Methods:

• Curated PediaSHF-DX dataset: 10,325 de-identified pediatric elbow X-rays (5,163 patients), with 2,015 images expertly annotated.
• Developed an improved YOLOv11-based detection model with LocalAttention and optimized structure for enhanced fracture detection.
• Validated the model on a separate test set of 8,310 images.

Main Results:

• The AI model achieved a precision of 0.96 on the test dataset.
• The model demonstrated high performance, strong generalization, and robustness across diverse imaging conditions.
• PediaSHF-DX dataset is publicly available for research.

Conclusions:

• PediaSHF-DX dataset and the proposed AI model offer a valuable resource for advancing AI-driven diagnosis of pediatric SHFs.
• This work supports the development of AI tools to aid orthopedic surgeons in pediatric fracture care.
• The findings highlight the potential of AI in improving the accuracy and efficiency of diagnosing common pediatric fractures.