Vision-Language Guidance in MRI Reconstruction Significantly Improves Image Sharpness

maier_ak · x · 2026-08-13

On fastMRI knee and brain benchmarks, the introduced semantic regularizer lowered LPIPS (better perceptual similarity) and raised Tenengrad (sharper structure), effectively boosting radiologist reader-study scores. Furthermore, text prompts can be used to steer the reconstruction process toward specific anatomy.

Related event: Vision-Language Models Act as Regularizers to Improve MRI Reconstruction(5 posts)→

Original post →

More from Research

Research channel →