Cycle-guided denoising diffusion probability model for 3D cross-modality MRI synthesis

This study aims to develop a novel Cycle-guided Denoising Diffusion Probability Model (CG-DDPM) for enhanced cross-modality magnetic resonance imaging (MRI) synthesis. The CG-DDPM deploys a pair of mutually conditioned DDPMs to synthesize images from two different MRI pulse sequences. The exchange of random latent noise between the two DDPMs in the reverse processes serves to regularize both DDPMs and ensure the generation of matching images across the two modalities. This improves image-to-image translation accuracy. We evaluated the CG-DDPM quantitatively using mean absolute error (MAE), multi-scale structural similarity index measure (MSSIM), and peak signal-to-noise ratio (PSNR), as well as the network synthesis consistency, on the BraTS2020 dataset. Our proposed method showed superior accuracy and consistency in MRI synthesis. Comparative analyses with several other state-of-the-art networks demonstrated our method’s statistically significant improvements in the image quality of synthetic MRIs. By enhancing the capability of current multimodal MRI synthesis approaches, the CG-DDPM has the potential to improve diagnostic accuracy and optimize treatment planning by enabling the synthesis of additional MRI modalities.