AI models can create one type of MRI scan from another, potentially saving time.

A new study compared three leading AI models for generating a T2-weighted MRI scan from a T1-weighted scan. Using over 11,000 training images, researchers found the CycleGAN model produced the most structurally accurate synthetic scans, while a Pix2Pix GAN created scans with the lowest error. This technology could reduce lengthy MRI sessions by synthesizing needed images instead of requiring separate, time-consuming scans.