Iterative tomographic reconstruction with TV prior for low-dose CBCT dental imaging

Area of Science:

• Medical Imaging
• Computational Imaging
• Radiology

Background:

• Cone-beam computed tomography (CBCT) is increasingly used for 3D dental imaging.
• Iterative reconstruction methods offer improved image quality over the Feldkamp algorithm, particularly for low-dose data, but face challenges in 3D CBCT with truncated projections.
• Optimization algorithms for tomographic problems are often studied in 2D, with limited understanding of their 3D CBCT performance.

Purpose of the Study:

• To compare the performance of different optimization schemes for 3D dental CBCT reconstruction.
• To evaluate novel algorithms like primal-dual hybrid gradient (PDHG) and MLEM-TV in realistic dental imaging scenarios.
• To investigate the effectiveness of these methods in overcoming challenges like truncated projections and metal artifacts.

Main Methods:

• Comparison of iterative algorithms including SIRT-TV, MLEM, PDHG, and a novel MLEM-TV optimizer.
• Experiments conducted on simulated 3D jaw phantom data and real data from a clinical dental scanner.
• Adaptation of algorithms to CBCT specificities and investigation of pre-conditioned versions for improved convergence.

Main Results:

• PDHG and MLEM-TV algorithms demonstrated superior reconstruction quality compared to other methods.
• These improvements were observed when reconstructing low-dose and ultra-low-dose CBCT images.
• Pre-conditioned PDHG and MLEM-TV showed enhanced convergence speeds, crucial for clinical application.

Conclusions:

• PDHG and MLEM-TV, with adaptations for CBCT, are highly effective for 3D dental imaging reconstruction.
• These methods show promise in overcoming challenges like truncated projections and metal artifacts in dental CBCT.
• The study highlights the potential of advanced iterative algorithms for improving low-dose 3D dental imaging.