Core Concepts
Deep learning methods successfully reduce artifacts in CBCT scans, addressing various types of artifacts with different architectures.
Abstract
Deep learning approaches have been utilized to enhance image quality in cone-beam computed tomography (CBCT).
Research focuses on reducing artifacts arising from motion, metal objects, or low-dose acquisition.
Various deep learning techniques are applied to mitigate different types of artifacts in 3D and 4D CBCT.
The literature is organized based on the type of artifact addressed, with a primary focus on artifact reduction.
CNNs, U-Nets, GANs, and Cycle-GANs are commonly used architectures for artifact reduction.
Limited availability of open-source code repositories hinders reproducibility and transparency in research.
Stats
Deep learning approaches have been used to improve image quality in CBCT.
Deep learning techniques have successfully reduced artifacts in CBCT scans.
Generative models including GANs have been trending for artifact reduction in CBCT.
Only four papers provided a public code repository for reproducibility.
Quotes
"Deep learning based approaches have been used to improve image quality in cone-beam computed tomography (CBCT)."
"Research focuses on reducing artifacts arising from motion, metal objects, or low-dose acquisition."