Self-supervised denoising of projection data for low-dose cone-beam CT.

Our self-supervised learning strategy is capable of restoring anatomical information while efficiently removing noise in CBCT projection data.