Miga improves speed and quality in 3d multi-echo mri scans
MIGA:Shared-Geometry Gaussian Representation with Implicit Amplitude Modeling for Accelerated 3D Multi-Echo MRI
Computer Vision and Pattern Recognition
Summary
3D multi-echo MRI scans take a long time because they collect lots of detailed data, which can be hard to speed up without losing quality. The authors developed MIGA, a new method that uses shared spatial models and smart signal processing to reconstruct faster scans from incomplete data. MIGA learns the common shapes and echo-specific details together from the scan itself, without needing extra training data. Tests showed MIGA made faster MRI scans possible while keeping or improving image quality, especially when the data was very incomplete.
What this means in practice
- •For medical imaging teams: Accelerate 3D multi-echo MRI reconstruction to reduce patient scan times while maintaining image quality using MIGA’s shared geometry and amplitude modeling.
- •For clinical mri technologists: Use MIGA to improve multi-echo MRI throughput and image clarity in clinical workflows needing rapid volumetric scans.
Authors
Jingran Xu, Yuanyuan Liu, Yanjie Zhu
Abstract
Three-dimensional multi-echo MRI provides rich anatomical and quantitative information, but repeated volumetric encoding prolongs acquisition and motivates k-space undersampling. Reconstructing undersampled multi-echo data requires exploiting shared anatomy while preserving echo-dependent signal variation; full-volume modeling also introduces substantial computational and memory demands. We propose MIGA, a scan-specific framework comprising shared anisotropic Gaussian geometry, a coordinate-conditioned multi-output amplitude network, and explicit echo-specific phase variables. The Gaussian geometry provides common spatial support across echoes, the implicit network models spatially structured amplitude variations, and the phase variables retain echo-specific complex signal information. All components are jointly optimized using only the acquired multi-coil k-space, requiring no fully sampled training data. Experiments showed that MIGA consistently outperformed the comparison methods across imaging tasks and acceleration factors, with larger improvements under stronger undersampling. MIGA also achieved a favorable quality-cost balance among the evaluated full-volume multi-echo methods. These results support the effectiveness of combining shared Gaussian geometry with implicit echo-dependent amplitude modeling for accelerated 3D multi-echo MRI reconstruction.