Arbitrary scale super-resolution diffusion model for brain MRI images

Han, Zhitao; Huang, Wenhui

doi:10.1016/j.compbiomed.2024.108003

Cited by 5 publications

(1 citation statement)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In this work, we propose to integrate the conditional variable discrete diffusion process into the HR prostate MRI synthesis. Rather than directly synthesizing the HR MRI, our approach utilizes residual prediction (Han et al 2024, Wang et al 2023, Zheng et al 2021, Chen et al 2024 to estimate the difference between the HR MRI x H and the up-sampled LR MRI up(x L ) with bilinear interpolation. This difference is referred to as the input residual image x r .…”

Section: Methodsmentioning

confidence: 99%

Discrete residual diffusion model for high-resolution prostate MRI synthesis

Han,

Huang

2024

Phys. Med. Biol.

Self Cite

View full text Add to dashboard Cite

Objective. High-resolution magnetic resonance imaging (HR MRI) is an effective tool for diagnosing PCa, but it requires patients to remain immobile for extended periods, increasing chances of image distortion due to motion. One solution is to utilize super-resolution (SR) techniques to process low-resolution (LR) images and create a higher-resolution version. However, existing medical SR models suffer from issues such as excessive smoothness and mode collapse. In this paper, we propose a novel generative model avoiding the problems of existing models, called Discrete Residual Diffusion Model (DR-DM). Approach. First, the forward process of DR-DM gradually disrupts the input via a fixed Markov chain, producing a sequence of latent variables with increasing noise. The backward process learns the conditional transit distribution and gradually match the target data distribution. By optimizing a variant of the variational lower bound, training diffusion models effectively address the issue of mode collapse. Second, to focus DR-DM on recovering high-frequency details, we synthesize residual images instead of synthesizing HR MRI directly. The residual image represents the difference between the HR and LR up-sampled MR image, and we convert residual image into discrete image tokens with a shorter sequence length by a vector quantized variational autoencoder (VQ-VAE), which reduced the computational complexity. Third, transformer architecture is integrated to model the relationship between LR MRI and residual image, which can capture the long-range dependencies between LR MRI and the synthesized imaging and improve the fidelity of reconstructed images.Main results. Extensive experimental validations have been performed on two popular yet challenging magnetic resonance image super-resolution tasks and compared to five state-of-the-art methods. Significance. Our experiments on the Prostate-Diagnosis and PROSTATEx datasets demonstrate that the DR-DM model significantly improves the signal-to-noise ratio of MRI for prostate cancer, resulting in greater clarity and improved diagnostic accuracy for patients.

show abstract

Section: Methodsmentioning

confidence: 99%