December 6, 2025LifeOpen Access

Multiscale Segmentation-Guided Diffusion Model for CBCT-to-CT Synthesis

View Full Paper

Ask AI

Bookmark

Authors

Discussion

Proposed framework lowers mean absolute error to 61.82 HU in synthetic CT generation, indicating enhanced fidelity in radiotherapy images.

Synthetic CT generation achieves a mean absolute error of 61.82 HU, improving image fidelity.
Evaluated using the SynthRAD2023 brain dataset shows strong performance against baseline models.
Proposed framework applies multiscale segmentation guidance to enhance anatomical consistency in CT images.
Findings support improved radiotherapy applications for cone-beam CT by utilizing structural similarity.

Guo et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark

Proposed framework lowers mean absolute error to 61.82 HU in synthetic CT generation, indicating enhanced fidelity in radiotherapy images.