[A deep blur learning-based motion artifact reduction algorithm for dental cone-beam computed tomography images].

Area of Science:

• Medical Imaging
• Radiology
• Computer Vision

Context:

• Motion artifacts significantly degrade the quality of dental cone-beam computed tomography (CBCT) images.
• Accurate diagnosis relies on high-resolution CBCT imaging, which is often compromised by patient movement during scans.

Purpose:

• To introduce a deep blur learning algorithm (DMBL) for effective motion artifact correction in dental CBCT.
• To model and remove artifacts caused by spatially varying and random motion patterns.

Summary:

• The DMBL algorithm utilizes a blur encoder to extract motion-related degradation features.
• A joint learning framework within the artifact correction module addresses image blur removal and simulation.
• Experiments on simulated and clinical datasets validated the algorithm's performance.

Impact:

• DMBL significantly improved image quality metrics, with PSNR increasing by 2.88% and SSIM by 0.89% on simulated data.
• Clinical datasets showed superior subjective image quality with a score of 4.417 out of 5.
• The algorithm effectively restores high-quality dental CBCT images, aiding in more accurate diagnoses.