MRI super-resolution reconstruction using efficient diffusion probabilistic model with residual shifting.

Objective: MRI offers superior soft-tissue contrast yet suffers from long acquisition times that can induce patient discomfort and motion artifacts. Super-resolution (SR) methods reconstruct high-resolution (HR) images from low-resolution (LR) scans, but diffusion models typically require numerous sampling steps, hindering real-time use. Here, we introduce a residual error-shifting strategy that reduce sampling steps without compromising anatomical fidelity, thereby improving MRI SR for clinical deployment.

Approach: We propose a novel diffusion-based SR framework called Res-SRDiff, which integrates residual error shifting into the forward diffusion process. This approach enables efficient HR image reconstruction by aligning the degraded HR image distribution with the LR image distribution. Our model was evaluated on two MRI datasets: ultra-high-field brain T1 MP2RAGE maps and T2-weighted prostate images. We compared Res-SRDiff against established methods, including Bicubic, Pix2pix, CycleGAN, and a conventional denoising diffusion probabilistic model with vision transformer backbone (TM-DDPM), using quantitative metrics such as peak signal-to-noise ratio (PSNR), structural similarity index (SSIM), gradient magnitude similarity deviation (GMSD), and learned perceptual image patch similarity (LPIPS).

Main results: Res-SRDiff significantly outperformed all comparative methods in terms of PSNR, SSIM, and GMSD across both datasets, with statistically significant improvements ( $p$ -values ≪ 0.05). The model achieved high-fidelity image restoration with only four sampling steps, drastically reducing computational time to under one second per slice, which is substantially faster than conventional TM-DDPM with around 20 seconds per slice. Qualitative analyses further demonstrated that Res-SRDiff effectively preserved fine anatomical details and lesion morphology in both brain and pelvic MRI images.

Significance: Our findings show that Res-SRDiff is an efficient and accurate MRI SR method, markedly improving computational efficiency and image quality. By integrating residual error shifting into the diffusion process, it allows for rapid and robust HR image reconstruction, enhancing clinical MRI workflows and advancing medical imaging research. The source at: https://github.com/mosaf/Res-SRDiff.