b-MAR: bidirectional artifact representations learning framework for metal artifact reduction in dental CBCT

Area of Science:

• Medical Imaging
• Artificial Intelligence in Radiology
• Computational Imaging

Background:

• Metal artifacts in computed tomography (CT) images, particularly in dental CBCT, significantly impede diagnosis and treatment.
• Existing supervised metal artifact reduction (MAR) methods often overlook the underlying metal artifact generation process.
• Dental CBCT is prone to severe artifact contamination due to low tube voltages and high-attenuation dental materials.

Purpose of the Study:

• To propose a novel bidirectional artifact representations learning framework for adaptive metal artifact reduction (MAR) in dental CBCT.
• To model both the generation and elimination of metal artifacts for enhanced MAR performance.
• To improve the accuracy and reliability of dental CBCT imaging.

Main Methods:

• Introduced an efficient artifact encoder to extract multi-scale metal artifact representations.
• Developed a framework embedding artifact representations into both generator and eliminator networks for simultaneous artifact removal and generation learning.
• Implemented artifact consistency loss to align image generation processes.

Main Results:

• The proposed bidirectional MAR (b-MAR) method demonstrated significant improvements on simulated and clinical datasets with diverse dental metal morphologies.
• Achieved >1.4131 dB in PSNR, >0.3473 HU decrements in RMSE, and >0.0025 promotion in structural similarity index measurement compared to state-of-the-art methods.
• Effectively removed artifacts caused by various metal morphologies, restoring dental tissue integrity.

Conclusions:

• The b-MAR method enhances MAR performance by strengthening joint learning of artifact removal and generation processes through bidirectional artifact representation embedding.
• b-MAR robustly and effectively corrects metal artifacts in dental CBCT images caused by different dental metals.
• The proposed framework offers a significant advancement in artifact reduction for dental imaging.