Self-supervised adversarial diffusion models for fast MRI reconstruction

Mojtaba Safari

doi:10.1002/mp.17675

Self-supervised adversarial diffusion models for fast MRI reconstruction

2025-02-09

0

OA

AI

Xiao-Feng Yang (Xiaofeng Yang) *

DOI：10.1002/mp.17675

原文链接

原文求助

摘要

En 中文

BackgroundMagnetic resonance imaging (MRI) offers excellent soft tissue contrast essential for diagnosis and treatment, but its long acquisition times can cause patient discomfort and motion artifacts.PurposeTo propose a self-supervised deep learning-based compressed sensing MRI method named Self-Supervised Adversarial Diffusion for MRI Accelerated Reconstruction (SSAD-MRI) to accelerate data acquisition without requiring fully sampled datasets.Materials and MethodsWe used the fastMRI multi-coil brain axial T2$\text{T}_{2}$-weighted (T2$\text{T}_{2}$-w) dataset from 1376 cases and single-coil brain quantitative magnetization prepared 2 rapid acquisition gradient echoes T1$\text{T}_{1}$ maps from 318 cases to train and test our model. Robustness against domain shift was evaluated using two out-of-distribution (OOD) datasets: multi-coil brain axial postcontrast T1$\text{T}_{1}$-weighted (T1c$\text{T}_{1}\text{c}$) dataset from 50 cases and axial T1-weighted (T1-w) dataset from 50 patients. Data were retrospectively subsampled at acceleration rates R is an element of{2x,4x,8x}$ R \in \lbrace 2\times, 4\times, 8\times \rbrace $. SSAD-MRI partitions a random sampling pattern into two disjoint sets, ensuring data consistency during training. We compared our method with ReconFormer Transformer and SS-MRI, assessing performance using normalized mean squared error (NMSE), peak signal-to-noise ratio (PSNR), and structural similarity index (SSIM). Statistical tests included one-way analysis of variance and multi-comparison Tukey's honesty significant difference (HSD) tests.ResultsSSAD-MRI preserved fine structures and brain abnormalities visually better than comparative methods at R=8x$ R=8\times$ for both multi-coil and single-coil datasets. It achieved the lowest NMSE at R is an element of{4x,8x}$ R \in \lbrace 4\times, 8\times \rbrace $, and the highest PSNR and SSIM values at all acceleration rates for the multi-coil dataset. Similar trends were observed for the single-coil dataset, though SSIM values were comparable to ReconFormer at R is an element of{2x,8x}$ R \in \lbrace 2\times, 8\times \rbrace $. These results were further confirmed by the voxel-wise correlation scatter plots. OOD results showed significant (p << 10-5$ \ll 10<^>{-5}$) improvements in undersampled image quality after reconstruction.ConclusionsSSAD-MRI successfully reconstructs fully sampled images without utilizing them in the training step, potentially reducing imaging costs and enhancing image quality crucial for diagnosis and treatment.

Keyword:

accelerated MRI

adaptive partitioning

fastMRI

k-space sampling

reconstruction

Self-supervised adversarial diffusion models for fast MRI reconstruction

摘要

期刊

机构

引用论文