Automatic Video Analysis Framework for Exposure Region Recognition in X-Ray Imaging Automation.

Area of Science:

• Medical Imaging Automation
• Deep Learning
• Radiography

Background:

• Deep learning shows promise for medical imaging automation, but research in X-ray imaging is limited.
• Current X-ray imaging processes are labor-intensive for radiographers and can be optimized.

Purpose of the Study:

• To develop an automated system for recognizing the exposure moment and region in X-ray imaging.
• To address key challenges in X-ray imaging automation using a hybrid deep learning model.

Main Methods:

• Proposed a hybrid deep learning framework with three components: Body Structure Detection, Motion State Tracing, and Body Modeling.
• Body Structure Detection identifies keypoints and bounding boxes for spatial information.
• Motion State Tracing determines the optimal exposure moment, and Body Modeling calculates the exposure region.

Main Results:

• A large-scale dataset of X-ray examination scenes was created for validation.
• Extensive experiments confirmed the method's superiority in automatically recognizing exposure moments and regions.
• The framework achieves near real-time performance.

Conclusions:

• The developed framework enables automatic and accurate recognition of the exposure region in X-ray imaging without radiographer assistance.
• This represents a significant advancement in medical imaging automation, potentially improving workflow and image quality.