Extended Partial Angle Based Motion Compensation for Dental CBCT

Cristina Sarti, Mikhail Mikerov, Claudio Landi

Keywords: Dental CBCT, Motion compensation, Partial-angle reconstruction, Angular range extension

TL;DR: A robust and fast deep learning–based approach to compensate for motion artifacts in dental CBCT

Abstract: Motion artifacts can degrade image quality in dental cone-beam CT and complicate diagnosis. In some cases, an exam retake is necessary, resulting in additional radiation exposure for the patient, without any guarantee of improved image quality. Therefore, motion compensation methods play a crucial role. Many methods are time-consuming since they require several reconstructions. We propose a very efficient method that requires only two partial-angle reconstructions. It assumes that the patient remains still during the acquisition, except for a short interval. In this situation, two motion-free partial-angle reconstructions, one before and one after patient motion, can be reconstructed. Motion compensation is achieved by registering forward projections of the two volumes. To enhance the robustness of the registration step, we simulate an extended angular range covered by the two partial volumes using a conditioned U-Net trained on a target-specific dataset. Qualitative analysis shows that we can significantly reduce the appearance of motion artifacts even in the case of challenging motion patterns.

Changes Summary: We incorporated feedback aimed at improving the readability of the manuscript. However, no additional experiments were performed.

Main Tex File: samplepaper.tex

Confirm Latex Only: true

Authors Changed: false

Submission Number: 8