Evaluation of dose reduction potential in scatter-corrected bedside chest radiography using U-net

Area of Science:

• Medical Imaging
• Radiography
• Deep Learning

Background:

• Bedside radiography enables immediate diagnosis but faces challenges with X-ray tube and detector alignment.
• Scattered X-rays degrade image quality, necessitating scatter removal grids.
• Current methods for scatter reduction in digital radiography are complex and may increase radiation dose.

Purpose of the Study:

• To develop and evaluate a scatter-correction processing method for bedside chest radiography.
• To reduce radiation dose while maintaining diagnostic image quality.
• To enable accurate segmentation of mass regions using deep learning with low-dose images.

Main Methods:

• A scatter-correction processing method was proposed for digital radiography using wireless flat-panel detectors (FPDs).
• Chest phantoms and simulated mass regions were used to assess image quality with reduced radiation doses.
• Quantitative assessment involved peak signal-to-noise ratio (PSNR) and structural similarity (SSIM).
• U-net deep learning model was employed for mass region segmentation.

Main Results:

• Scatter correction achieved image quality equivalent to grid radiography with approximately 20% lower radiation dose.
• Image contrast was improved with scatter correction compared to traditional scatter removal grids.
• The method facilitated accurate mass region segmentation on low-dose, scatter-corrected chest X-rays.

Conclusions:

• Scatter-correction processing is a viable alternative to scatter removal grids in bedside chest radiography.
• This technique allows for significant radiation dose reduction without compromising image quality.
• The developed method supports the advancement of safer and more effective bedside imaging systems.