A fidelity-embedded learning for metal artifact reduction in dental CBCT

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Radiology

Background:

• Dental cone-beam computed tomography (CBCT) is crucial for maxillofacial imaging.
• Metallic objects like implants and braces cause severe artifacts in CBCT, degrading diagnostic accuracy.
• Current methods struggle to accurately reconstruct images with metallic inserts.

Purpose of the Study:

• To develop and evaluate a deep learning method for reducing metal artifacts in dental CBCT.
• To improve the diagnostic performance of CBCT by mitigating image degradation caused by metallic materials.

Main Methods:

• An iterative deep learning approach was proposed to address complex metal artifacts.
• The method enforces data fidelity in the projection domain during iterative learning.
• A realistic training dataset was generated by simulating metal artifacts in metal-free CBCT scans.

Main Results:

• The proposed method significantly reduced metal artifacts in both simulated and clinical CBCT scans.
• Image quality and morphological structures near metallic objects were preserved.
• The fidelity-embedded learning approach outperformed direct image domain learning.

Conclusions:

• Fidelity-embedded learning effectively reduces metal artifacts in dental CBCT.
• The method offers a promising solution for improving CBCT diagnostic performance in patients with metallic restorations.