MotionDPS: Motion-Compensated 3D Brain MRI Reconstruction
Researchers have developed MotionDPS, a novel Bayesian framework for reconstructing 3D brain MRI scans corrupted by patient motion. This method jointly estimates the anatomical image, motion parameters, and coil sensitivity maps using pretrained diffusion models as image priors. Experiments show MotionDPS outperforms existing techniques, especially with severe motion and high acceleration, and operates without requiring paired motion-free training data. AI
IMPACT This new method could improve the diagnostic quality of brain MRIs by correcting motion artifacts, potentially leading to more accurate diagnoses and treatment plans.