Multi-Contrast MRI Motion Correction via Parameter-Informed Disentanglement and Adaptive Experts
Researchers have developed a new deep learning framework called ScanCLIP to address motion artifacts in MRI scans. This method uses parameter-informed contrast disentanglement and adaptive experts to correct artifacts across different MRI modalities and severity levels. ScanCLIP leverages contrast embeddings derived from acquisition parameters and a Vision Transformer to route features through a Mixture-of-Experts network for targeted correction, demonstrating improved performance and robust generalization on various benchmarks. AI
IMPACT This novel approach could improve the diagnostic accuracy of MRI scans by reducing motion-related distortions.