Motion artifacts are a persistent problem in routine two-dimensional T2-weighted turbo spin-echo brain MRI, blurring images and lowering diagnostic confidence. In severe cases, scans must be repeated, adding time and cost. A new preprint on arXiv introduces MoSAIK, an AI-based method designed to tackle this issue at the data level.
MoSAIK, which stands for Motion Search with AI in K-space, works directly in the raw frequency domain of the MRI signal rather than on the final image. By identifying and correcting motion-induced errors in k-space, the method aims to restore image quality without requiring additional scan time or hardware changes.
The preprint is an early-stage description and does not yet include clinical validation results. Still, it highlights a practical direction for AI in medical imaging: addressing a common, real-world source of image degradation that affects patient care and workflow efficiency.