Quality control of elbow joint radiography using a YOLOv8-based artificial intelligence technology

Area of Science:

• Radiology
• Artificial Intelligence
• Medical Imaging

Background:

• Quality control (QC) of elbow joint radiographs is crucial for accurate disease detection.
• Traditional QC methods can be time-consuming and subjective.
• Exploring advanced technologies like AI for objective and efficient QC is essential.

Purpose of the Study:

• To investigate the efficacy of an artificial intelligence (AI) technology employing YOLOv8 for automated quality control (QC) of elbow joint radiographs.
• To develop and validate AI models for identifying key positioning parameters and assessing image quality.

Main Methods:

• Collected 2643 elbow radiographs from January 2022 to August 2023.
• Developed anteroposterior (AP) and lateral (LAT) YOLOv8 models to detect target boxes and key points.
• Transformed identifications into five quality standards, including positioning coordinates and flexion angle.

Main Results:

• YOLOv8 models achieved high precision, recall, and mean average precision in identifying anatomical landmarks.
• AI demonstrated high agreement with physician assessments across all evaluated quality parameters (ICC > 0.865).
• AI-based QC reduced image processing time by 43% for AP and 45% for LAT views (p < 0.001).

Conclusions:

• YOLOv8-based AI technology is a feasible and high-performing tool for elbow radiography QC.
• The proposed AI models offer objective and efficient solutions for automated quality assessment in clinical settings.
• This AI approach enhances the reliability and efficiency of radiographic quality control.